transcribtion_diarization.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "# Whisper's transcription plus Pyannote's Diarization \n",
        "\n",
        "Andrej Karpathy [suggested](https://twitter.com/karpathy/status/1574476200801538048?s=20&t=s5IMMXOYjBI6-91dib6w8g) training a classifier on top of  OpenAI [Whisper](https://openai.com/blog/whisper/) model features to identify the speaker, so we can visualize the speaker in the transcript. But, as [pointed out](https://twitter.com/tarantulae/status/1574493613362388992?s=20&t=s5IMMXOYjBI6-91dib6w8g) by Christian Perone, it seems that features from whisper wouldn't be that great for speaker recognition as its main objective is basically to ignore speaker differences.\n",
        "\n",
        "In the following, I use [**`pyannote-audio`**](https://github.com/pyannote/pyannote-audio), a speaker diarization toolkit by Hervé Bredin, to identify the speakers, and then match it with the transcriptions of Whispr. I do it on the first 30 minutes of  Lex's 2nd [interview](https://youtu.be/SGzMElJ11Cc) with Yann LeCun. Check the result [**here**](https://majdoddin.github.io/lexicap.html). \n",
        "\n",
        "It is tricky to match the transcriptions to diarization segemtns, specially when the speaker changes. To resolve it, Sarah Kaiser [suggested](https://github.com/openai/whisper/discussions/264#discussioncomment-3825375) runnnig the pyannote.audio first and  then just running whisper on the split-by-speaker chunks. \n",
        "For sake of performance (and transcription quality?), we attach the audio segements into a single audio file with a silent spacer as a seperator, and run whisper on it. Enjoy it!"
      ],
      "metadata": {
        "id": "RQyROdrfsvk4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Preparing the audio file"
      ],
      "metadata": {
        "id": "wtljXaTXnowa"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        " Installing `yt-dlp` and downloading the [video](https://)."
      ],
      "metadata": {
        "id": "Ct4adIANpQvX"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "TL7fC4aZdpyH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b1e04802-27db-4ca5-bbd1-97d0375825d6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting yt-dlp\n",
            "  Downloading yt_dlp-2022.10.4-py2.py3-none-any.whl (2.7 MB)\n",
            "\u001b[K     |████████████████████████████████| 2.7 MB 7.6 MB/s \n",
            "\u001b[?25hCollecting pycryptodomex\n",
            "  Downloading pycryptodomex-3.15.0-cp35-abi3-manylinux2010_x86_64.whl (2.3 MB)\n",
            "\u001b[K     |████████████████████████████████| 2.3 MB 55.8 MB/s \n",
            "\u001b[?25hCollecting mutagen\n",
            "  Downloading mutagen-1.46.0-py3-none-any.whl (193 kB)\n",
            "\u001b[K     |████████████████████████████████| 193 kB 73.3 MB/s \n",
            "\u001b[?25hRequirement already satisfied: certifi in /usr/local/lib/python3.7/dist-packages (from yt-dlp) (2022.9.24)\n",
            "Collecting websockets\n",
            "  Downloading websockets-10.3-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (112 kB)\n",
            "\u001b[K     |████████████████████████████████| 112 kB 74.0 MB/s \n",
            "\u001b[?25hCollecting brotli\n",
            "  Downloading Brotli-1.0.9-cp37-cp37m-manylinux1_x86_64.whl (357 kB)\n",
            "\u001b[K     |████████████████████████████████| 357 kB 64.7 MB/s \n",
            "\u001b[?25hInstalling collected packages: websockets, pycryptodomex, mutagen, brotli, yt-dlp\n",
            "Successfully installed brotli-1.0.9 mutagen-1.46.0 pycryptodomex-3.15.0 websockets-10.3 yt-dlp-2022.10.4\n"
          ]
        }
      ],
      "source": [
        "!pip install -U yt-dlp"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "KsedMPN-daEX"
      },
      "outputs": [],
      "source": [
        "!wget -O - -q  https://github.com/yt-dlp/FFmpeg-Builds/releases/download/latest/ffmpeg-master-latest-linux64-gpl.tar.xz | xz -qdc| tar -x"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "2VHzpv72dkvV",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5dc16c9c-e457-4a0b-e9b4-145fc718f7b6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[debug] Command-line config: ['-xv', '--ffmpeg-location', 'ffmpeg-master-latest-linux64-gpl/bin', '--audio-format', 'wav', '-o', 'lecun.wav', '--', 'https://youtu.be/SGzMElJ11Cc']\n",
            "[debug] Encodings: locale UTF-8, fs utf-8, pref UTF-8, out UTF-8, error UTF-8, screen UTF-8\n",
            "[debug] yt-dlp version 2022.10.04 [4e0511f] (pip) API\n",
            "[debug] Python 3.7.14 (CPython 64bit) - Linux-5.10.133+-x86_64-with-Ubuntu-18.04-bionic (glibc 2.26)\n",
            "[debug] Checking exe version: ffmpeg-master-latest-linux64-gpl/bin/ffmpeg -bsfs\n",
            "[debug] Checking exe version: ffmpeg-master-latest-linux64-gpl/bin/ffprobe -bsfs\n",
            "[debug] exe versions: ffmpeg N-108603-ge10e27a2ea-20221010 (setts), ffprobe N-108603-ge10e27a2ea-20221010\n",
            "[debug] Optional libraries: Cryptodome-3.15.0, brotli-1.0.9, certifi-2022.09.24, mutagen-1.46.0, sqlite3-2.6.0, websockets-10.3\n",
            "[debug] Proxy map: {}\n",
            "[debug] Loaded 1690 extractors\n",
            "[debug] [youtube] Extracting URL: https://youtu.be/SGzMElJ11Cc\n",
            "[youtube] SGzMElJ11Cc: Downloading webpage\n",
            "[youtube] SGzMElJ11Cc: Downloading android player API JSON\n",
            "[debug] Sort order given by extractor: quality, res, fps, hdr:12, source, vcodec:vp9.2, channels, acodec, lang, proto\n",
            "[debug] Formats sorted by: hasvid, ie_pref, quality, res, fps, hdr:12(7), source, vcodec:vp9.2(10), channels, acodec, lang, proto, filesize, fs_approx, tbr, vbr, abr, asr, vext, aext, hasaud, id\n",
            "[info] SGzMElJ11Cc: Downloading 1 format(s): 251\n",
            "[download] lecun.wav has already been downloaded\n",
            "[debug] ffmpeg command line: ffmpeg-master-latest-linux64-gpl/bin/ffprobe -show_streams file:lecun.wav\n",
            "[ExtractAudio] Destination: lecun.wav\n",
            "[debug] ffmpeg command line: ffmpeg-master-latest-linux64-gpl/bin/ffmpeg -y -loglevel repeat+info -i file:lecun.wav -vn -movflags +faststart file:lecun.temp.wav\n",
            "Deleting original file lecun.orig.wav (pass -k to keep)\n"
          ]
        }
      ],
      "source": [
        "!yt-dlp -xv --ffmpeg-location ffmpeg-master-latest-linux64-gpl/bin --audio-format wav  -o lecun.wav -- https://youtu.be/SGzMElJ11Cc"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "Cutting the first 20 minutes of the video for further process.\n"
      ],
      "metadata": {
        "id": "Gzp4EVzWpd5I"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ZqznOE7Kw725",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "73877474-b03c-43cf-825b-2cea28ff0c3f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting pydub\n",
            "  Downloading pydub-0.25.1-py2.py3-none-any.whl (32 kB)\n",
            "Installing collected packages: pydub\n",
            "Successfully installed pydub-0.25.1\n"
          ]
        }
      ],
      "source": [
        "!pip install pydub"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "LlB5gZnswvL1",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "085f2142-68da-4a37-b5e1-b24988619276"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<_io.BufferedRandom name='lecun1.wav'>"
            ]
          },
          "metadata": {},
          "execution_count": 7
        }
      ],
      "source": [
        "from pydub import AudioSegment\n",
        "\n",
        "t1 = 0 * 1000 #Works in milliseconds\n",
        "t2 = 20 * 60 * 1000\n",
        "\n",
        "newAudio = AudioSegment.from_wav(\"lecun.wav\")\n",
        "a = newAudio[t1:t2]\n",
        "a.export(\"lecun1.wav\", format=\"wav\") \n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "`pyannote.audio` seems to miss the first 0.5 seconds of the audio, and, therefore, we prepend a spcacer."
      ],
      "metadata": {
        "id": "q7qMLTISFE6M"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "audio = AudioSegment.from_wav(\"lecun1.wav\")\n",
        "spacermilli = 2000\n",
        "spacer = AudioSegment.silent(duration=spacermilli)\n",
        "audio = spacer.append(audio, crossfade=0)\n",
        "\n",
        "audio.export('audio.wav', format='wav')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MaRDsBV1CWi8",
        "outputId": "e8b7eb9d-0908-4c87-f969-bb780e7e1d19"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<_io.BufferedRandom name='audio.wav'>"
            ]
          },
          "metadata": {},
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Pyannote's Diarization"
      ],
      "metadata": {
        "id": "mb5eEOKUooju"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nxNf1l8Ye_U9"
      },
      "source": [
        "[`pyannote.audio`](https://github.com/pyannote/pyannote-audio) is an open-source toolkit written in Python for **speaker diarization**. \n",
        "\n",
        "Based on [`PyTorch`](https://pytorch.org) machine learning framework, it provides a set of trainable end-to-end neural building blocks that can be combined and jointly optimized to build speaker diarization pipelines. \n",
        "\n",
        "`pyannote.audio` also comes with pretrained [models](https://huggingface.co/models?other=pyannote-audio-model) and [pipelines](https://huggingface.co/models?other=pyannote-audio-pipeline) covering a wide range of domains for voice activity detection, speaker segmentation, overlapped speech detection, speaker embedding reaching state-of-the-art performance for most of them. "
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Installing Pyannote and running it on the video to generate the diarizations."
      ],
      "metadata": {
        "id": "S8Ak_OQwqd-3"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "vJGyKTQJqdzq",
        "outputId": "a910865e-252a-4609-f998-32f6d9d59300"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting pyannote.audio\n",
            "  Downloading pyannote.audio-2.0.1-py2.py3-none-any.whl (385 kB)\n",
            "\u001b[K     |████████████████████████████████| 385 kB 8.1 MB/s \n",
            "\u001b[?25hCollecting pyannote.database<5.0,>=4.1.1\n",
            "  Downloading pyannote.database-4.1.3-py3-none-any.whl (41 kB)\n",
            "\u001b[K     |████████████████████████████████| 41 kB 470 kB/s \n",
            "\u001b[?25hCollecting backports.cached-property\n",
            "  Downloading backports.cached_property-1.0.2-py3-none-any.whl (6.1 kB)\n",
            "Collecting torch-audiomentations>=0.11.0\n",
            "  Downloading torch_audiomentations-0.11.0-py3-none-any.whl (47 kB)\n",
            "\u001b[K     |████████████████████████████████| 47 kB 5.4 MB/s \n",
            "\u001b[?25hCollecting pyannote.pipeline<3.0,>=2.3\n",
            "  Downloading pyannote.pipeline-2.3-py3-none-any.whl (30 kB)\n",
            "Requirement already satisfied: networkx<3.0,>=2.6 in /usr/local/lib/python3.7/dist-packages (from pyannote.audio) (2.6.3)\n",
            "Collecting singledispatchmethod\n",
            "  Downloading singledispatchmethod-1.0-py2.py3-none-any.whl (4.7 kB)\n",
            "Collecting semver<3.0,>=2.10.2\n",
            "  Downloading semver-2.13.0-py2.py3-none-any.whl (12 kB)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from pyannote.audio) (4.1.1)\n",
            "Requirement already satisfied: torchaudio<1.0,>=0.10 in /usr/local/lib/python3.7/dist-packages (from pyannote.audio) (0.12.1+cu113)\n",
            "Collecting omegaconf<3.0,>=2.1\n",
            "  Downloading omegaconf-2.2.3-py3-none-any.whl (79 kB)\n",
            "\u001b[K     |████████████████████████████████| 79 kB 7.5 MB/s \n",
            "\u001b[?25hCollecting pyannote.metrics<4.0,>=3.2\n",
            "  Downloading pyannote.metrics-3.2.1-py3-none-any.whl (51 kB)\n",
            "\u001b[K     |████████████████████████████████| 51 kB 197 kB/s \n",
            "\u001b[?25hCollecting torchmetrics<1.0,>=0.6\n",
            "  Downloading torchmetrics-0.10.0-py3-none-any.whl (529 kB)\n",
            "\u001b[K     |████████████████████████████████| 529 kB 67.1 MB/s \n",
            "\u001b[?25hCollecting huggingface-hub<0.9,>=0.7\n",
            "  Downloading huggingface_hub-0.8.1-py3-none-any.whl (101 kB)\n",
            "\u001b[K     |████████████████████████████████| 101 kB 13.0 MB/s \n",
            "\u001b[?25hCollecting pyannote.core<5.0,>=4.4\n",
            "  Downloading pyannote.core-4.5-py3-none-any.whl (60 kB)\n",
            "\u001b[K     |████████████████████████████████| 60 kB 9.0 MB/s \n",
            "\u001b[?25hCollecting pytorch-lightning<1.7,>=1.5.4\n",
            "  Downloading pytorch_lightning-1.6.5-py3-none-any.whl (585 kB)\n",
            "\u001b[K     |████████████████████████████████| 585 kB 66.1 MB/s \n",
            "\u001b[?25hCollecting einops<0.4.0,>=0.3\n",
            "  Downloading einops-0.3.2-py3-none-any.whl (25 kB)\n",
            "Collecting pytorch-metric-learning<2.0,>=1.0.0\n",
            "  Downloading pytorch_metric_learning-1.6.2-py3-none-any.whl (111 kB)\n",
            "\u001b[K     |████████████████████████████████| 111 kB 74.4 MB/s \n",
            "\u001b[?25hRequirement already satisfied: torch>=1.9 in /usr/local/lib/python3.7/dist-packages (from pyannote.audio) (1.12.1+cu113)\n",
            "Collecting speechbrain<0.6,>=0.5.12\n",
            "  Downloading speechbrain-0.5.13-py3-none-any.whl (498 kB)\n",
            "\u001b[K     |████████████████████████████████| 498 kB 68.7 MB/s \n",
            "\u001b[?25hCollecting asteroid-filterbanks<0.5,>=0.4\n",
            "  Downloading asteroid_filterbanks-0.4.0-py3-none-any.whl (29 kB)\n",
            "Collecting soundfile<0.11,>=0.10.2\n",
            "  Downloading SoundFile-0.10.3.post1-py2.py3-none-any.whl (21 kB)\n",
            "Collecting hmmlearn<0.3,>=0.2.7\n",
            "  Downloading hmmlearn-0.2.8-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.whl (217 kB)\n",
            "\u001b[K     |████████████████████████████████| 217 kB 75.5 MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from asteroid-filterbanks<0.5,>=0.4->pyannote.audio) (1.21.6)\n",
            "Requirement already satisfied: scipy>=0.19 in /usr/local/lib/python3.7/dist-packages (from hmmlearn<0.3,>=0.2.7->pyannote.audio) (1.7.3)\n",
            "Requirement already satisfied: scikit-learn>=0.16 in /usr/local/lib/python3.7/dist-packages (from hmmlearn<0.3,>=0.2.7->pyannote.audio) (1.0.2)\n",
            "Requirement already satisfied: packaging>=20.9 in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<0.9,>=0.7->pyannote.audio) (21.3)\n",
            "Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<0.9,>=0.7->pyannote.audio) (5.0.0)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<0.9,>=0.7->pyannote.audio) (4.64.1)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<0.9,>=0.7->pyannote.audio) (2.23.0)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<0.9,>=0.7->pyannote.audio) (6.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<0.9,>=0.7->pyannote.audio) (3.8.0)\n",
            "Collecting antlr4-python3-runtime==4.9.*\n",
            "  Downloading antlr4-python3-runtime-4.9.3.tar.gz (117 kB)\n",
            "\u001b[K     |████████████████████████████████| 117 kB 76.9 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.9->huggingface-hub<0.9,>=0.7->pyannote.audio) (3.0.9)\n",
            "Collecting simplejson>=3.8.1\n",
            "  Downloading simplejson-3.17.6-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl (130 kB)\n",
            "\u001b[K     |████████████████████████████████| 130 kB 73.0 MB/s \n",
            "\u001b[?25hRequirement already satisfied: sortedcontainers>=2.0.4 in /usr/local/lib/python3.7/dist-packages (from pyannote.core<5.0,>=4.4->pyannote.audio) (2.4.0)\n",
            "Requirement already satisfied: typer[all]>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from pyannote.database<5.0,>=4.1.1->pyannote.audio) (0.4.2)\n",
            "Requirement already satisfied: pandas>=0.19 in /usr/local/lib/python3.7/dist-packages (from pyannote.database<5.0,>=4.1.1->pyannote.audio) (1.3.5)\n",
            "Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.19->pyannote.database<5.0,>=4.1.1->pyannote.audio) (2.8.2)\n",
            "Requirement already satisfied: pytz>=2017.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.19->pyannote.database<5.0,>=4.1.1->pyannote.audio) (2022.4)\n",
            "Requirement already satisfied: tabulate>=0.7.7 in /usr/local/lib/python3.7/dist-packages (from pyannote.metrics<4.0,>=3.2->pyannote.audio) (0.8.10)\n",
            "Requirement already satisfied: matplotlib>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from pyannote.metrics<4.0,>=3.2->pyannote.audio) (3.2.2)\n",
            "Requirement already satisfied: sympy>=1.1 in /usr/local/lib/python3.7/dist-packages (from pyannote.metrics<4.0,>=3.2->pyannote.audio) (1.7.1)\n",
            "Collecting docopt>=0.6.2\n",
            "  Downloading docopt-0.6.2.tar.gz (25 kB)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.0.0->pyannote.metrics<4.0,>=3.2->pyannote.audio) (0.11.0)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib>=2.0.0->pyannote.metrics<4.0,>=3.2->pyannote.audio) (1.4.4)\n",
            "Collecting optuna>=1.4\n",
            "  Downloading optuna-3.0.3-py3-none-any.whl (348 kB)\n",
            "\u001b[K     |████████████████████████████████| 348 kB 46.3 MB/s \n",
            "\u001b[?25hCollecting alembic>=1.5.0\n",
            "  Downloading alembic-1.8.1-py3-none-any.whl (209 kB)\n",
            "\u001b[K     |████████████████████████████████| 209 kB 66.5 MB/s \n",
            "\u001b[?25hCollecting importlib-metadata\n",
            "  Downloading importlib_metadata-4.13.0-py3-none-any.whl (23 kB)\n",
            "Requirement already satisfied: sqlalchemy>=1.3.0 in /usr/local/lib/python3.7/dist-packages (from optuna>=1.4->pyannote.pipeline<3.0,>=2.3->pyannote.audio) (1.4.41)\n",
            "Collecting colorlog\n",
            "  Downloading colorlog-6.7.0-py2.py3-none-any.whl (11 kB)\n",
            "Collecting cliff\n",
            "  Downloading cliff-3.10.1-py3-none-any.whl (81 kB)\n",
            "\u001b[K     |████████████████████████████████| 81 kB 10.0 MB/s \n",
            "\u001b[?25hCollecting cmaes>=0.8.2\n",
            "  Downloading cmaes-0.8.2-py3-none-any.whl (15 kB)\n",
            "Requirement already satisfied: importlib-resources in /usr/local/lib/python3.7/dist-packages (from alembic>=1.5.0->optuna>=1.4->pyannote.pipeline<3.0,>=2.3->pyannote.audio) (5.9.0)\n",
            "Collecting Mako\n",
            "  Downloading Mako-1.2.3-py3-none-any.whl (78 kB)\n",
            "\u001b[K     |████████████████████████████████| 78 kB 8.6 MB/s \n",
            "\u001b[?25hRequirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->huggingface-hub<0.9,>=0.7->pyannote.audio) (3.8.1)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas>=0.19->pyannote.database<5.0,>=4.1.1->pyannote.audio) (1.15.0)\n",
            "Requirement already satisfied: fsspec[http]!=2021.06.0,>=2021.05.0 in /usr/local/lib/python3.7/dist-packages (from pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (2022.8.2)\n",
            "Collecting pyDeprecate>=0.3.1\n",
            "  Downloading pyDeprecate-0.3.2-py3-none-any.whl (10 kB)\n",
            "Requirement already satisfied: protobuf<=3.20.1 in /usr/local/lib/python3.7/dist-packages (from pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (3.17.3)\n",
            "Requirement already satisfied: tensorboard>=2.2.0 in /usr/local/lib/python3.7/dist-packages (from pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (2.8.0)\n",
            "Requirement already satisfied: aiohttp!=4.0.0a0,!=4.0.0a1 in /usr/local/lib/python3.7/dist-packages (from fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (3.8.3)\n",
            "Requirement already satisfied: async-timeout<5.0,>=4.0.0a3 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (4.0.2)\n",
            "Requirement already satisfied: aiosignal>=1.1.2 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.2.0)\n",
            "Requirement already satisfied: asynctest==0.13.0 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (0.13.0)\n",
            "Requirement already satisfied: attrs>=17.3.0 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (22.1.0)\n",
            "Requirement already satisfied: frozenlist>=1.1.1 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.3.1)\n",
            "Requirement already satisfied: yarl<2.0,>=1.0 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.8.1)\n",
            "Requirement already satisfied: multidict<7.0,>=4.5 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (6.0.2)\n",
            "Requirement already satisfied: charset-normalizer<3.0,>=2.0 in /usr/local/lib/python3.7/dist-packages (from aiohttp!=4.0.0a0,!=4.0.0a1->fsspec[http]!=2021.06.0,>=2021.05.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (2.1.1)\n",
            "Requirement already satisfied: torchvision in /usr/local/lib/python3.7/dist-packages (from pytorch-metric-learning<2.0,>=1.0.0->pyannote.audio) (0.13.1+cu113)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.16->hmmlearn<0.3,>=0.2.7->pyannote.audio) (1.2.0)\n",
            "Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from scikit-learn>=0.16->hmmlearn<0.3,>=0.2.7->pyannote.audio) (3.1.0)\n",
            "Requirement already satisfied: cffi>=1.0 in /usr/local/lib/python3.7/dist-packages (from soundfile<0.11,>=0.10.2->pyannote.audio) (1.15.1)\n",
            "Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.0->soundfile<0.11,>=0.10.2->pyannote.audio) (2.21)\n",
            "Collecting sentencepiece\n",
            "  Downloading sentencepiece-0.1.97-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "\u001b[K     |████████████████████████████████| 1.3 MB 45.4 MB/s \n",
            "\u001b[?25hCollecting hyperpyyaml\n",
            "  Downloading HyperPyYAML-1.0.1.tar.gz (14 kB)\n",
            "Requirement already satisfied: greenlet!=0.4.17 in /usr/local/lib/python3.7/dist-packages (from sqlalchemy>=1.3.0->optuna>=1.4->pyannote.pipeline<3.0,>=2.3->pyannote.audio) (1.1.3)\n",
            "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.7/dist-packages (from sympy>=1.1->pyannote.metrics<4.0,>=3.2->pyannote.audio) (1.2.1)\n",
            "Requirement already satisfied: setuptools>=41.0.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (57.4.0)\n",
            "Requirement already satisfied: tensorboard-data-server<0.7.0,>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (0.6.1)\n",
            "Requirement already satisfied: werkzeug>=0.11.15 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.0.1)\n",
            "Requirement already satisfied: google-auth<3,>=1.6.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.35.0)\n",
            "Requirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.8.1)\n",
            "Requirement already satisfied: absl-py>=0.4 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.2.0)\n",
            "Requirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (0.4.6)\n",
            "Requirement already satisfied: grpcio>=1.24.3 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.49.1)\n",
            "Requirement already satisfied: markdown>=2.6.8 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (3.4.1)\n",
            "Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.7/dist-packages (from tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (0.37.1)\n",
            "Requirement already satisfied: rsa<5,>=3.1.4 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (4.9)\n",
            "Requirement already satisfied: pyasn1-modules>=0.2.1 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (0.2.8)\n",
            "Requirement already satisfied: cachetools<5.0,>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from google-auth<3,>=1.6.3->tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (4.2.4)\n",
            "Requirement already satisfied: requests-oauthlib>=0.7.0 in /usr/local/lib/python3.7/dist-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (1.3.1)\n",
            "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /usr/local/lib/python3.7/dist-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (0.4.8)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->huggingface-hub<0.9,>=0.7->pyannote.audio) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->huggingface-hub<0.9,>=0.7->pyannote.audio) (1.24.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->huggingface-hub<0.9,>=0.7->pyannote.audio) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->huggingface-hub<0.9,>=0.7->pyannote.audio) (2022.9.24)\n",
            "Requirement already satisfied: oauthlib>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard>=2.2.0->pytorch-lightning<1.7,>=1.5.4->pyannote.audio) (3.2.1)\n",
            "Requirement already satisfied: librosa>=0.6.0 in /usr/local/lib/python3.7/dist-packages (from torch-audiomentations>=0.11.0->pyannote.audio) (0.8.1)\n",
            "Collecting torch-pitch-shift>=1.2.2\n",
            "  Downloading torch_pitch_shift-1.2.2-py3-none-any.whl (5.0 kB)\n",
            "Collecting julius<0.3,>=0.2.3\n",
            "  Downloading julius-0.2.7.tar.gz (59 kB)\n",
            "\u001b[K     |████████████████████████████████| 59 kB 6.1 MB/s \n",
            "\u001b[?25hRequirement already satisfied: decorator>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (4.4.2)\n",
            "Requirement already satisfied: resampy>=0.2.2 in /usr/local/lib/python3.7/dist-packages (from librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (0.4.2)\n",
            "Requirement already satisfied: pooch>=1.0 in /usr/local/lib/python3.7/dist-packages (from librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (1.6.0)\n",
            "Requirement already satisfied: numba>=0.43.0 in /usr/local/lib/python3.7/dist-packages (from librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (0.56.2)\n",
            "Requirement already satisfied: audioread>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (3.0.0)\n",
            "Requirement already satisfied: llvmlite<0.40,>=0.39.0dev0 in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (0.39.1)\n",
            "Requirement already satisfied: appdirs>=1.3.0 in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa>=0.6.0->torch-audiomentations>=0.11.0->pyannote.audio) (1.4.4)\n",
            "Collecting primePy>=1.3\n",
            "  Downloading primePy-1.3-py3-none-any.whl (4.0 kB)\n",
            "Requirement already satisfied: click<9.0.0,>=7.1.1 in /usr/local/lib/python3.7/dist-packages (from typer[all]>=0.2.1->pyannote.database<5.0,>=4.1.1->pyannote.audio) (7.1.2)\n",
            "Collecting colorama<0.5.0,>=0.4.3\n",
            "  Downloading colorama-0.4.5-py2.py3-none-any.whl (16 kB)\n",
            "Collecting shellingham<2.0.0,>=1.3.0\n",
            "  Downloading shellingham-1.5.0-py2.py3-none-any.whl (9.3 kB)\n",
            "Collecting stevedore>=2.0.1\n",
            "  Downloading stevedore-3.5.0-py3-none-any.whl (49 kB)\n",
            "\u001b[K     |████████████████████████████████| 49 kB 7.3 MB/s \n",
            "\u001b[?25hCollecting pbr!=2.1.0,>=2.0.0\n",
            "  Downloading pbr-5.10.0-py2.py3-none-any.whl (112 kB)\n",
            "\u001b[K     |████████████████████████████████| 112 kB 74.6 MB/s \n",
            "\u001b[?25hCollecting cmd2>=1.0.0\n",
            "  Downloading cmd2-2.4.2-py3-none-any.whl (147 kB)\n",
            "\u001b[K     |████████████████████████████████| 147 kB 63.3 MB/s \n",
            "\u001b[?25hCollecting autopage>=0.4.0\n",
            "  Downloading autopage-0.5.1-py3-none-any.whl (29 kB)\n",
            "Requirement already satisfied: PrettyTable>=0.7.2 in /usr/local/lib/python3.7/dist-packages (from cliff->optuna>=1.4->pyannote.pipeline<3.0,>=2.3->pyannote.audio) (3.4.1)\n",
            "Requirement already satisfied: wcwidth>=0.1.7 in /usr/local/lib/python3.7/dist-packages (from cmd2>=1.0.0->cliff->optuna>=1.4->pyannote.pipeline<3.0,>=2.3->pyannote.audio) (0.2.5)\n",
            "Collecting pyperclip>=1.6\n",
            "  Downloading pyperclip-1.8.2.tar.gz (20 kB)\n",
            "Collecting ruamel.yaml>=0.17.8\n",
            "  Downloading ruamel.yaml-0.17.21-py3-none-any.whl (109 kB)\n",
            "\u001b[K     |████████████████████████████████| 109 kB 70.9 MB/s \n",
            "\u001b[?25hCollecting ruamel.yaml.clib>=0.2.6\n",
            "  Downloading ruamel.yaml.clib-0.2.6-cp37-cp37m-manylinux1_x86_64.whl (546 kB)\n",
            "\u001b[K     |████████████████████████████████| 546 kB 59.8 MB/s \n",
            "\u001b[?25hRequirement already satisfied: MarkupSafe>=0.9.2 in /usr/local/lib/python3.7/dist-packages (from Mako->alembic>=1.5.0->optuna>=1.4->pyannote.pipeline<3.0,>=2.3->pyannote.audio) (2.0.1)\n",
            "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.7/dist-packages (from torchvision->pytorch-metric-learning<2.0,>=1.0.0->pyannote.audio) (7.1.2)\n",
            "Building wheels for collected packages: antlr4-python3-runtime, docopt, julius, pyperclip, hyperpyyaml\n",
            "  Building wheel for antlr4-python3-runtime (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for antlr4-python3-runtime: filename=antlr4_python3_runtime-4.9.3-py3-none-any.whl size=144575 sha256=e7e3581a6975328ac6349ee7ca394efa4ab5de89eaf7d156febbd3852ee9f5a9\n",
            "  Stored in directory: /root/.cache/pip/wheels/8b/8d/53/2af8772d9aec614e3fc65e53d4a993ad73c61daa8bbd85a873\n",
            "  Building wheel for docopt (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for docopt: filename=docopt-0.6.2-py2.py3-none-any.whl size=13723 sha256=50e04f7d744d89bb271ee40399345ea7fddf39c462d0db05456f95f7031459c7\n",
            "  Stored in directory: /root/.cache/pip/wheels/72/b0/3f/1d95f96ff986c7dfffe46ce2be4062f38ebd04b506c77c81b9\n",
            "  Building wheel for julius (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for julius: filename=julius-0.2.7-py3-none-any.whl size=21894 sha256=100073a66e83972c2496ad035c8fca399f733ff06edbd9813882409862f9d788\n",
            "  Stored in directory: /root/.cache/pip/wheels/44/52/2c/7dd069f82c7f905f40b190a8039ec2a17fdd4bb009c57c6664\n",
            "  Building wheel for pyperclip (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for pyperclip: filename=pyperclip-1.8.2-py3-none-any.whl size=11137 sha256=449d044e6613042e8af7e0194266161c2ad43209ca5ca112b1fe7849b4aab1de\n",
            "  Stored in directory: /root/.cache/pip/wheels/9f/18/84/8f69f8b08169c7bae2dde6bd7daf0c19fca8c8e500ee620a28\n",
            "  Building wheel for hyperpyyaml (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for hyperpyyaml: filename=HyperPyYAML-1.0.1-py3-none-any.whl size=15192 sha256=68f45a0d6c96cfc2450bb743e8c823fc574ef231b5b08b5292cc7dd3d8e452d0\n",
            "  Stored in directory: /root/.cache/pip/wheels/6c/87/65/266d722c3932f81f16332ce842e972be8421e3a9cd3771766b\n",
            "Successfully built antlr4-python3-runtime docopt julius pyperclip hyperpyyaml\n",
            "Installing collected packages: pyperclip, pbr, importlib-metadata, stevedore, simplejson, shellingham, ruamel.yaml.clib, Mako, colorama, cmd2, autopage, soundfile, ruamel.yaml, pyannote.core, primePy, colorlog, cmaes, cliff, alembic, torchmetrics, torch-pitch-shift, sentencepiece, pyDeprecate, pyannote.database, optuna, julius, hyperpyyaml, huggingface-hub, docopt, antlr4-python3-runtime, torch-audiomentations, speechbrain, singledispatchmethod, semver, pytorch-metric-learning, pytorch-lightning, pyannote.pipeline, pyannote.metrics, omegaconf, hmmlearn, einops, backports.cached-property, asteroid-filterbanks, pyannote.audio\n",
            "  Attempting uninstall: importlib-metadata\n",
            "    Found existing installation: importlib-metadata 5.0.0\n",
            "    Uninstalling importlib-metadata-5.0.0:\n",
            "      Successfully uninstalled importlib-metadata-5.0.0\n",
            "  Attempting uninstall: soundfile\n",
            "    Found existing installation: soundfile 0.11.0\n",
            "    Uninstalling soundfile-0.11.0:\n",
            "      Successfully uninstalled soundfile-0.11.0\n",
            "Successfully installed Mako-1.2.3 alembic-1.8.1 antlr4-python3-runtime-4.9.3 asteroid-filterbanks-0.4.0 autopage-0.5.1 backports.cached-property-1.0.2 cliff-3.10.1 cmaes-0.8.2 cmd2-2.4.2 colorama-0.4.5 colorlog-6.7.0 docopt-0.6.2 einops-0.3.2 hmmlearn-0.2.8 huggingface-hub-0.8.1 hyperpyyaml-1.0.1 importlib-metadata-4.13.0 julius-0.2.7 omegaconf-2.2.3 optuna-3.0.3 pbr-5.10.0 primePy-1.3 pyDeprecate-0.3.2 pyannote.audio-2.0.1 pyannote.core-4.5 pyannote.database-4.1.3 pyannote.metrics-3.2.1 pyannote.pipeline-2.3 pyperclip-1.8.2 pytorch-lightning-1.6.5 pytorch-metric-learning-1.6.2 ruamel.yaml-0.17.21 ruamel.yaml.clib-0.2.6 semver-2.13.0 sentencepiece-0.1.97 shellingham-1.5.0 simplejson-3.17.6 singledispatchmethod-1.0 soundfile-0.10.3.post1 speechbrain-0.5.13 stevedore-3.5.0 torch-audiomentations-0.11.0 torch-pitch-shift-1.2.2 torchmetrics-0.10.0\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.colab-display-data+json": {
              "pip_warning": {
                "packages": [
                  "pydevd_plugins"
                ]
              }
            }
          },
          "metadata": {}
        }
      ],
      "source": [
        "!pip install   pyannote.audio"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 273,
          "referenced_widgets": [
            "546a16efa53e462f88890ebe3977a3bc",
            "1f7c6499be4846e1bc674761f3b875f2",
            "268c730678bb4af6baf7bb4f185ddb1e",
            "350678a2f548480b9d6acceda5d78949",
            "8f9cafa27efc43c4bb3f566f87df523c",
            "e4802c94ccaf404180f3760e3b859a3a",
            "71d0b4b532e649829bc813d0f09b2c3b",
            "5fb9f7acb4c7491a986f6d674f467028",
            "40efa9c5b79e40cab21dcdcf98533caa",
            "766c1a1b8c1248e79dfa03ec89dbc4dd",
            "15e62f4a65104bdea233a96c4dbb6db8",
            "327bebaf45eb46b289615c8deb29a69d",
            "a6919e4b1e2c4510958bf343ef2524a8",
            "309cc6bf3b1e41b1a558fbdb3745c301",
            "ec474373198046d299e09b4823a2766d",
            "88186bbd5a1c4425b254d608ca85cd8a",
            "458d4d31cd87402c97c4038553c06574",
            "d6d8814501a34ef391c565ba4748e365",
            "a9a5de2de75643048ba821d73a00d5c7",
            "51422049fa5846bb9e009b530f0a7526",
            "ae3785b55e8c4b7e963dc357fb255f05",
            "7e8227f2dac34d83a3c670e4d4e0a249",
            "fe9e7728c09946c5908d7502c3f77432",
            "b67a926cc4904c2f84efd509098e103c",
            "fd98d75dc2424c2faecfb4b5f6a94aa2",
            "187e2d909bfc4a41b2502b964a03fcad",
            "bf18adc120174a8bb6cd6f09d853911f",
            "0c4de4008c2e45e5a31d21cdb954f0f2",
            "28ef8f3ac2e44445ac9e40d140e513d3",
            "d137a72df0a849a8913fd62924c1848a",
            "105cf78a882a4cc3bbdaaa2b80217110",
            "b2791832f51f4794acfdbded875bc9b5",
            "28c4d707b0764f51ba2d06eeed5dad96",
            "0b616fa3a46a4502b14f93270fcb1ca7",
            "fd8fd5f045414ab199d3f37c9397ad54",
            "9652595d01cc478c952a368a1f86940d",
            "3a648dda02c9485e8e975e7a1ad2c27e",
            "eb2e878c6f0d413b9e82b93b6057e66f",
            "a6df649cf752475c9540bf1ba2fc7199",
            "e491634079f44e65a6edb5d799a2b74e",
            "c9389e064ac54435a3fd356b13b86f4e",
            "098d718a941542b99b772068e442e9fa",
            "4a1e581661b84dc6b9c52c1c6efa006b",
            "3d1b76d21f6349ccbeefd795974321f7",
            "bbacac6c5aff4686a31130560e9cf86b",
            "ade8c72f0eb64de58c4a1351c0b5cdbd",
            "ea54cd59edc64f95958f62a9b6e182c9",
            "5a1b7cd142e94f22a9abc9e64b11d633",
            "9403474d6e774f4dbb36c843620ba666",
            "68eb916bf9494da6a1740e404da1c602",
            "1c72713c82fa4ab89cb8fbe84089828f",
            "a730cdaad4e24736b872723240a094dd",
            "a0b10ced5ddc47b5867d864bcc786e0e",
            "7c74e70870304e83826f2a322341a000",
            "5a98ade616cd441da36ed51b85f4b61c",
            "36ecee3e086a4a34bd6f6fe435b20a6e",
            "f638d0f51ac548e8a36a0b8858082e29",
            "41359dbbceea458b9137b8cb8fb2e7e1",
            "a9b8f75e82274a4e8cfa1026fe8408de",
            "b71b33109b124e30b0004634467e64dc",
            "32792ea893a140dab321565df126e461",
            "e9b04004efb944c6ad7d53763d426626",
            "5d5fa4a478844c798c91a274361cda2b",
            "323bd202a5a540b4b8805b88138468ad",
            "0831e94760fa421384e2b33c97532d09",
            "b7073c903e1e4701b68516dc8a867fc1",
            "28e3b80e86634a32aaef5b67345c217c",
            "3a1c92eec125401ca3184d713b27f4a5",
            "500472e00812437390d5caf9bc0ee67b",
            "41aa0f5eaddd4907802992b380e85ded",
            "030169a7a2594b5a8646f327a0566dfe",
            "55d89d284453466c8013f2a884d55426",
            "47506b0858534c6ebc53e8e500ef56b6",
            "feaf363bb8c548fcae1cd13fd1074581",
            "fc80e8542e34422b86874eee0b20b3f8",
            "b8cdfacd7bee4c398fc513ed2a3311e3",
            "85c433e956794588b166c9ed6fa68c20",
            "36c91974947c4bf7aef8585de061a5a7",
            "a7d1f042f6124c51aa69ac6985e7c83c",
            "3fbcb6d197ab4c44930a2e15e439459b",
            "6b7194d233cd47428f58bf987f9712fe",
            "135ef8a71e4041ad80129cc5dbc51354",
            "5d46ef19e68c4974a5001a48121047f3",
            "65b31b2e94b04dd39298f004f9493e29",
            "e73e393e01f4487d86dfbe5e401b4c09",
            "a93270536a094ed694b825ea38aa36a2",
            "dbbb60965fb24cd9b2dc1ea510e7e480",
            "d06a71fe937b4ee1a44764f020bf3d57"
          ]
        },
        "id": "jKG14DGYbwku",
        "outputId": "799c6a8b-16b1-49a7-b87a-ad6d340cf375"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/598 [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "546a16efa53e462f88890ebe3977a3bc"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/17.7M [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "327bebaf45eb46b289615c8deb29a69d"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/318 [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "fe9e7728c09946c5908d7502c3f77432"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/1.92k [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "0b616fa3a46a4502b14f93270fcb1ca7"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/83.3M [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "bbacac6c5aff4686a31130560e9cf86b"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/1.92k [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "36ecee3e086a4a34bd6f6fe435b20a6e"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/5.53M [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "28e3b80e86634a32aaef5b67345c217c"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "Downloading:   0%|          | 0.00/129k [00:00<?, ?B/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "36c91974947c4bf7aef8585de061a5a7"
            }
          },
          "metadata": {}
        }
      ],
      "source": [
        "from pyannote.audio import Pipeline\n",
        "\n",
        "pipeline = Pipeline.from_pretrained('pyannote/speaker-diarization')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "yA4xiEefft9Z"
      },
      "outputs": [],
      "source": [
        "DEMO_FILE = {'uri': 'blabal', 'audio': 'audio.wav'}\n",
        "dz = pipeline(DEMO_FILE)  \n",
        "\n",
        "with open(\"diarization.txt\", \"w\") as text_file:\n",
        "    text_file.write(str(dz))"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(*list(dz.itertracks(yield_label = True))[:10], sep=\"\\n\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GHIY2MB3Vz3e",
        "outputId": "11faa3bf-cd94-4753-dffb-1be3c061f8b4"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(<Segment(2.03344, 36.8128)>, 0, 'SPEAKER_00')\n",
            "(<Segment(38.1122, 51.3759)>, 0, 'SPEAKER_00')\n",
            "(<Segment(51.8653, 90.2053)>, 1, 'SPEAKER_01')\n",
            "(<Segment(91.2853, 92.9391)>, 1, 'SPEAKER_01')\n",
            "(<Segment(94.8628, 116.497)>, 0, 'SPEAKER_00')\n",
            "(<Segment(116.497, 124.124)>, 1, 'SPEAKER_01')\n",
            "(<Segment(124.192, 151.597)>, 1, 'SPEAKER_01')\n",
            "(<Segment(152.018, 179.12)>, 1, 'SPEAKER_01')\n",
            "(<Segment(180.318, 194.037)>, 1, 'SPEAKER_01')\n",
            "(<Segment(195.016, 207.385)>, 0, 'SPEAKER_00')\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def millisec(timeStr):\n",
        "  spl = timeStr.split(\":\")\n",
        "  s = (int)((int(spl[0]) * 60 * 60 + int(spl[1]) * 60 + float(spl[2]) )* 1000)\n",
        "  return s"
      ],
      "metadata": {
        "id": "KPGOaVpOH7pZ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import re\n",
        "dz = open('diarization.txt').read().splitlines()\n",
        "dzList = []\n",
        "for l in dz:\n",
        "  start, end =  tuple(re.findall('[0-9]+:[0-9]+:[0-9]+\\.[0-9]+', string=l))\n",
        "  start = millisec(start) - spacermilli\n",
        "  end = millisec(end)  - spacermilli\n",
        "  lex = not re.findall('SPEAKER_01', string=l)\n",
        "  dzList.append([start, end, lex])\n",
        "\n",
        "print(*dzList[:10], sep='\\n')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5uIUo68M0g7j",
        "outputId": "b957f4eb-9c00-45e9-a738-c5f3a37f6539"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[33, 34812, True]\n",
            "[36112, 49375, True]\n",
            "[49865, 88205, False]\n",
            "[89285, 90939, False]\n",
            "[92862, 114496, True]\n",
            "[114496, 122124, False]\n",
            "[122191, 149596, False]\n",
            "[150018, 177119, False]\n",
            "[178317, 192037, False]\n",
            "[193015, 205385, True]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Preparing audio file from the diarization"
      ],
      "metadata": {
        "id": "Wp36eMedRkR0"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Attaching audio segements according to the diarization, with a spacer as the delimiter."
      ],
      "metadata": {
        "id": "VEEuVPj6H0J_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from pydub import AudioSegment\n",
        "import re \n",
        "\n",
        "sounds = spacer\n",
        "segments = []\n",
        "\n",
        "dz = open('diarization.txt').read().splitlines()\n",
        "for l in dz:\n",
        "  start, end =  tuple(re.findall('[0-9]+:[0-9]+:[0-9]+\\.[0-9]+', string=l))\n",
        "  start = int(millisec(start)) #milliseconds\n",
        "  end = int(millisec(end))  #milliseconds\n",
        "  \n",
        "  segments.append(len(sounds))\n",
        "  sounds = sounds.append(audio[start:end], crossfade=0)\n",
        "  sounds = sounds.append(spacer, crossfade=0)\n",
        "\n",
        "sounds.export(\"dz.wav\", format=\"wav\") #Exports to a wav file in the current path."
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pQqlyfY4Ciqh",
        "outputId": "323d3a88-a803-4bdf-b64d-d28d54be6349"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<_io.BufferedRandom name='dz.wav'>"
            ]
          },
          "metadata": {},
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "segments[:8]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cacmB1ggm51s",
        "outputId": "4ac0d27a-d215-4ba6-8704-7cdfa4416412"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[2000, 38779, 54042, 94382, 98036, 121670, 131297, 160702]"
            ]
          },
          "metadata": {},
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Freeing up some memory"
      ],
      "metadata": {
        "id": "rv2GYZCsLKBJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "del   sounds, DEMO_FILE, pipeline, spacer,  audio, dz, a, newAudio"
      ],
      "metadata": {
        "id": "cONumKWUjfus"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Whisper's Transcriptions"
      ],
      "metadata": {
        "id": "AmxtB0k4n8lY"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Installing Open AI whisper.\n",
        "\n",
        "**Important:** There is a version conflict with pyannote.audio resulting in an error (see this RP). Our workaround is to first run Pyannote and then whisper. You can safely ignore the error.\n"
      ],
      "metadata": {
        "id": "swPVuqWaakkH"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install git+https://github.com/openai/whisper.git "
      ],
      "metadata": {
        "id": "kUd7I__FUZVc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 927
        },
        "outputId": "fa0da6d8-a6c4-4fd7-9878-300264bb89b7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting git+https://github.com/openai/whisper.git\n",
            "  Cloning https://github.com/openai/whisper.git to /tmp/pip-req-build-wg93n19_\n",
            "  Running command git clone -q https://github.com/openai/whisper.git /tmp/pip-req-build-wg93n19_\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from whisper==1.0) (1.21.6)\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (from whisper==1.0) (1.12.1+cu113)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from whisper==1.0) (4.64.1)\n",
            "Requirement already satisfied: more-itertools in /usr/local/lib/python3.7/dist-packages (from whisper==1.0) (8.14.0)\n",
            "Collecting transformers>=4.19.0\n",
            "  Downloading transformers-4.23.0-py3-none-any.whl (5.3 MB)\n",
            "\u001b[K     |████████████████████████████████| 5.3 MB 6.8 MB/s \n",
            "\u001b[?25hCollecting ffmpeg-python==0.2.0\n",
            "  Downloading ffmpeg_python-0.2.0-py3-none-any.whl (25 kB)\n",
            "Requirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from ffmpeg-python==0.2.0->whisper==1.0) (0.16.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.7/dist-packages (from transformers>=4.19.0->whisper==1.0) (3.8.0)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.7/dist-packages (from transformers>=4.19.0->whisper==1.0) (2022.6.2)\n",
            "Requirement already satisfied: importlib-metadata in /usr/local/lib/python3.7/dist-packages (from transformers>=4.19.0->whisper==1.0) (4.13.0)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.7/dist-packages (from transformers>=4.19.0->whisper==1.0) (21.3)\n",
            "Collecting tokenizers!=0.11.3,<0.14,>=0.11.1\n",
            "  Downloading tokenizers-0.13.1-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (7.6 MB)\n",
            "\u001b[K     |████████████████████████████████| 7.6 MB 42.6 MB/s \n",
            "\u001b[?25hRequirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from transformers>=4.19.0->whisper==1.0) (2.23.0)\n",
            "Collecting huggingface-hub<1.0,>=0.10.0\n",
            "  Downloading huggingface_hub-0.10.1-py3-none-any.whl (163 kB)\n",
            "\u001b[K     |████████████████████████████████| 163 kB 57.4 MB/s \n",
            "\u001b[?25hRequirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.7/dist-packages (from transformers>=4.19.0->whisper==1.0) (6.0)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.7/dist-packages (from huggingface-hub<1.0,>=0.10.0->transformers>=4.19.0->whisper==1.0) (4.1.1)\n",
            "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.7/dist-packages (from packaging>=20.0->transformers>=4.19.0->whisper==1.0) (3.0.9)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.7/dist-packages (from importlib-metadata->transformers>=4.19.0->whisper==1.0) (3.8.1)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->transformers>=4.19.0->whisper==1.0) (3.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->transformers>=4.19.0->whisper==1.0) (2022.9.24)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->transformers>=4.19.0->whisper==1.0) (2.10)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->transformers>=4.19.0->whisper==1.0) (1.24.3)\n",
            "Building wheels for collected packages: whisper\n",
            "  Building wheel for whisper (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for whisper: filename=whisper-1.0-py3-none-any.whl size=1175147 sha256=cc42c95c9d7ecbfbfac99b877a859b712a22e5af1d9e1f13577ce6fb1acaf9e8\n",
            "  Stored in directory: /tmp/pip-ephem-wheel-cache-3s2prqw9/wheels/16/15/89/1c7bb31bd0006793a95549d04785121a8a36daad9158e1e43a\n",
            "Successfully built whisper\n",
            "Installing collected packages: tokenizers, huggingface-hub, transformers, ffmpeg-python, whisper\n",
            "  Attempting uninstall: huggingface-hub\n",
            "    Found existing installation: huggingface-hub 0.8.1\n",
            "    Uninstalling huggingface-hub-0.8.1:\n",
            "      Successfully uninstalled huggingface-hub-0.8.1\n",
            "\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
            "pyannote-audio 2.0.1 requires huggingface-hub<0.9,>=0.7, but you have huggingface-hub 0.10.1 which is incompatible.\u001b[0m\n",
            "Successfully installed ffmpeg-python-0.2.0 huggingface-hub-0.10.1 tokenizers-0.13.1 transformers-4.23.0 whisper-1.0\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "application/vnd.colab-display-data+json": {
              "pip_warning": {
                "packages": [
                  "huggingface_hub"
                ]
              }
            }
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Running Open AI whisper on the prepared audio file. [link text](https://) It writes the transcription into a file."
      ],
      "metadata": {
        "id": "8AleOXL7p7aM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!whisper dz.wav --language en --model large"
      ],
      "metadata": {
        "id": "WhzMeNgMVNFc",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2a2f768d-613f-4d1d-fc7e-698fe13cf30f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The cache for model files in Transformers v4.22.0 has been updated. Migrating your old cache. This is a one-time only operation. You can interrupt this and resume the migration later on by calling `transformers.utils.move_cache()`.\n",
            "Moving 0 files to the new cache system\n",
            "0it [00:00, ?it/s]\n",
            "100%|█████████████████████████████████████| 2.87G/2.87G [01:08<00:00, 45.1MiB/s]\n",
            "tcmalloc: large alloc 3087007744 bytes == 0x6162000 @  0x7fc7a07ea1e7 0x4b2150 0x5ac2ec 0x5dc6af 0x58ee9b 0x5901a3 0x5e3f6b 0x4d18aa 0x51b31c 0x58f2a7 0x51740e 0x5b41c5 0x58f49e 0x51837f 0x58f2a7 0x51740e 0x5b41c5 0x604133 0x606e06 0x606ecc 0x609aa6 0x64d332 0x64d4de 0x7fc7a03e7c87 0x5b561a\n",
            "[00:00.000 --> 00:04.720]  The following is a conversation with Yann LeCun,\n",
            "[00:04.720 --> 00:06.560]  his second time on the podcast.\n",
            "[00:06.560 --> 00:11.160]  He is the chief AI scientist at Meta, formerly Facebook,\n",
            "[00:11.160 --> 00:15.040]  professor at NYU, touring award winner,\n",
            "[00:15.040 --> 00:17.600]  one of the seminal figures in the history\n",
            "[00:17.600 --> 00:20.460]  of machine learning and artificial intelligence,\n",
            "[00:20.460 --> 00:23.940]  and someone who is brilliant and opinionated\n",
            "[00:23.940 --> 00:25.400]  in the best kind of way,\n",
            "[00:25.400 --> 00:27.960]  and so is always fun to talk to.\n",
            "[00:27.960 --> 00:29.960]  This is the Lex Friedman podcast.\n",
            "[00:29.960 --> 00:31.920]  To support it, please check out our sponsors\n",
            "[00:31.920 --> 00:33.200]  in the description.\n",
            "[00:33.200 --> 00:36.940]  And now, here's my conversation with Yann LeCun.\n",
            "[00:38.820 --> 00:40.160]  You co-wrote the article,\n",
            "[00:40.160 --> 00:41.520]  \"'Self-Supervised Learning,\n",
            "[00:41.520 --> 00:43.560]  the Dark Matter of Intelligence.\"\n",
            "[00:43.560 --> 00:46.380]  Great title, by the way, with Ishan Mizra.\n",
            "[00:46.380 --> 00:49.280]  So let me ask, what is self-supervised learning,\n",
            "[00:49.280 --> 00:52.200]  and why is it the dark matter of intelligence?\n",
            "[00:54.080 --> 00:55.880]  I'll start by the dark matter part.\n",
            "[00:55.880 --> 00:59.880]  There is obviously a kind of learning\n",
            "[00:59.880 --> 01:04.080]  that humans and animals are doing\n",
            "[01:04.080 --> 01:07.040]  that we currently are not reproducing properly\n",
            "[01:07.040 --> 01:08.880]  with machines or with AI, right?\n",
            "[01:08.880 --> 01:10.480]  So the most popular approaches\n",
            "[01:10.480 --> 01:12.680]  to machine learning today are,\n",
            "[01:12.680 --> 01:13.880]  or paradigms, I should say,\n",
            "[01:13.880 --> 01:16.920]  are supervised learning and reinforcement learning.\n",
            "[01:16.920 --> 01:19.340]  And they are extremely inefficient.\n",
            "[01:19.340 --> 01:21.840]  Supervised learning requires many samples\n",
            "[01:21.840 --> 01:23.960]  for learning anything,\n",
            "[01:23.960 --> 01:26.000]  and reinforcement learning requires\n",
            "[01:26.000 --> 01:29.420]  a ridiculously large number of trial and errors\n",
            "[01:29.420 --> 01:31.480]  for a system to learn anything.\n",
            "[01:34.420 --> 01:36.720]  And that's why we don't have self-driving cars.\n",
            "[01:38.120 --> 01:39.960]  That was a big leap from one to the other.\n",
            "[01:39.960 --> 01:43.940]  Okay, so to solve difficult problems,\n",
            "[01:43.940 --> 01:47.520]  you have to have a lot of human annotation\n",
            "[01:47.520 --> 01:49.240]  for supervised learning to work.\n",
            "[01:49.240 --> 01:50.700]  And to solve those difficult problems\n",
            "[01:50.700 --> 01:51.860]  with reinforcement learning,\n",
            "[01:51.860 --> 01:55.400]  you have to have some way to maybe simulate that problem\n",
            "[01:55.400 --> 01:57.880]  such that you can do that large-scale kind of learning\n",
            "[01:57.880 --> 01:59.680]  that reinforcement learning requires.\n",
            "[02:01.680 --> 02:05.180]  Right, so how is it that most teenagers\n",
            "[02:05.180 --> 02:08.840]  can learn to drive a car in about 20 hours of practice?\n",
            "[02:11.400 --> 02:16.400]  Whereas even with millions of hours of simulated practice,\n",
            "[02:16.600 --> 02:18.320]  a self-driving car can't actually learn\n",
            "[02:18.320 --> 02:19.820]  to drive itself properly.\n",
            "[02:19.820 --> 02:21.740]  And so, obviously, we're missing something, right?\n",
            "[02:21.740 --> 02:23.460]  And it's quite obvious for a lot of people\n",
            "[02:23.460 --> 02:27.620]  that the immediate response you get from many people is,\n",
            "[02:27.620 --> 02:30.620]  well, humans use their background knowledge\n",
            "[02:30.620 --> 02:33.580]  to learn faster, and they're right.\n",
            "[02:33.580 --> 02:36.060]  Now, how was that background knowledge acquired?\n",
            "[02:36.060 --> 02:37.260]  And that's the big question.\n",
            "[02:39.460 --> 02:41.820]  So now you have to ask,\n",
            "[02:41.820 --> 02:44.460]  how do babies in the first few months of life\n",
            "[02:44.460 --> 02:46.580]  learn how the world works?\n",
            "[02:46.580 --> 02:47.580]  Mostly by observation,\n",
            "[02:47.580 --> 02:50.020]  because they can hardly act in the world.\n",
            "[02:50.020 --> 02:51.700]  And they learn an enormous amount\n",
            "[02:51.700 --> 02:52.980]  of background knowledge about the world\n",
            "[02:52.980 --> 02:57.180]  that may be the basis of what we call common sense.\n",
            "[02:57.180 --> 03:00.420]  This type of learning is not learning a task,\n",
            "[03:00.420 --> 03:02.700]  it's not being reinforced for anything,\n",
            "[03:02.700 --> 03:06.340]  it's just observing the world and figuring out how it works.\n",
            "[03:08.220 --> 03:11.180]  Building world models, learning world models.\n",
            "[03:11.180 --> 03:12.020]  How do we do this?\n",
            "[03:12.020 --> 03:14.500]  And how do we reproduce this in machines?\n",
            "[03:14.500 --> 03:17.500]  So self-supervised learning is a very, very important thing\n",
            "[03:17.500 --> 03:21.780]  is one instance or one attempt\n",
            "[03:21.780 --> 03:23.980]  at trying to reproduce this kind of learning.\n",
            "[03:25.580 --> 03:28.860]  Okay, so you're looking at just observation,\n",
            "[03:28.860 --> 03:31.180]  so not even the interacting part of a child.\n",
            "[03:31.180 --> 03:34.060]  It's just sitting there watching mom and dad walk around,\n",
            "[03:34.060 --> 03:35.940]  pick up stuff, all of that.\n",
            "[03:35.940 --> 03:38.140]  That's what we mean by background knowledge.\n",
            "[03:41.980 --> 03:43.980]  Perhaps not even watching mom and dad,\n",
            "[03:43.980 --> 03:48.980]  just watching the world go by.\n",
            "[03:49.980 --> 03:51.900]  Just having eyes open or having eyes closed\n",
            "[03:51.900 --> 03:54.460]  or the very act of opening and closing eyes\n",
            "[03:54.460 --> 03:56.260]  that the world appears and disappears,\n",
            "[03:56.260 --> 03:57.820]  all that basic information.\n",
            "[03:59.820 --> 04:03.180]  And you're saying in order to learn to drive,\n",
            "[04:05.060 --> 04:07.780]  like the reason humans are able to learn to drive quickly,\n",
            "[04:07.780 --> 04:09.300]  some faster than others,\n",
            "[04:09.300 --> 04:10.620]  is because of the background knowledge\n",
            "[04:10.620 --> 04:15.100]  that we're able to watch cars operate in the world\n",
            "[04:15.100 --> 04:16.980]  in the many years leading up to it,\n",
            "[04:16.980 --> 04:19.060]  the physics of basic objects, all that.\n",
            "[04:21.100 --> 04:21.940]  That's right.\n",
            "[04:21.940 --> 04:22.780]  I mean, the basic physics of objects,\n",
            "[04:22.780 --> 04:26.220]  you don't even need to know how a car works\n",
            "[04:26.220 --> 04:27.860]  because that you can learn fairly quickly.\n",
            "[04:27.860 --> 04:29.180]  I mean, the example I use very often\n",
            "[04:29.180 --> 04:31.140]  is you're driving next to a cliff.\n",
            "[04:35.900 --> 04:39.820]  And you know in advance because of your understanding\n",
            "[04:39.820 --> 04:42.420]  of intuitive physics that if you turn the wheel\n",
            "[04:42.420 --> 04:44.300]  to the right, the car will veer to the right,\n",
            "[04:44.300 --> 04:46.820]  will run off the cliff, fall off the cliff,\n",
            "[04:46.820 --> 04:49.660]  and nothing good will come out of this, right?\n",
            "[04:49.660 --> 04:53.020]  But if you are a sort of tabularized\n",
            "[04:53.020 --> 04:54.340]  reinforcement learning system\n",
            "[04:54.340 --> 04:56.300]  that doesn't have a model of the world,\n",
            "[04:57.420 --> 04:59.740]  you have to repeat falling off this cliff\n",
            "[04:59.740 --> 05:02.060]  thousands of times before you figure out it's a bad idea.\n",
            "[05:02.060 --> 05:03.820]  And then a few more thousand times\n",
            "[05:03.820 --> 05:06.340]  before you figure out how to not do it.\n",
            "[05:06.340 --> 05:08.460]  And then a few more million times before you figure out\n",
            "[05:08.460 --> 05:11.260]  how to not do it in every situation you ever encounter.\n",
            "[05:12.740 --> 05:16.020]  So self-supervised learning still has to have\n",
            "[05:16.020 --> 05:20.780]  some source of truth being told to it by somebody.\n",
            "[05:20.780 --> 05:24.780]  So you have to figure out a way without human assistance\n",
            "[05:24.780 --> 05:26.820]  or without significant amount of human assistance\n",
            "[05:26.820 --> 05:29.300]  to get that truth from the world.\n",
            "[05:29.300 --> 05:34.180]  So the mystery there is how much signal is there,\n",
            "[05:34.180 --> 05:36.500]  how much truth is there that the world gives you,\n",
            "[05:36.500 --> 05:39.340]  whether it's the human world, like you watch YouTube\n",
            "[05:39.340 --> 05:42.680]  or something like that, or it's the more natural world.\n",
            "[05:44.500 --> 05:46.060]  So how much signal is there?\n",
            "[05:47.940 --> 05:51.020]  So here's the trick, there is way more signal\n",
            "[05:53.500 --> 05:55.540]  in sort of a self-supervised setting\n",
            "[05:55.540 --> 05:57.460]  than there is in either a supervised\n",
            "[05:57.460 --> 05:59.460]  or reinforcement setting.\n",
            "[05:59.460 --> 06:03.340]  And this is going to my analogy of the cake.\n",
            "[06:03.340 --> 06:07.300]  The luck cake, as someone has called it,\n",
            "[06:07.300 --> 06:10.980]  where when you try to figure out how much information\n",
            "[06:10.980 --> 06:13.660]  you ask the machine to predict and how much feedback\n",
            "[06:13.660 --> 06:15.500]  you give the machine at every trial.\n",
            "[06:17.820 --> 06:18.660]  In reinforcement learning,\n",
            "[06:18.660 --> 06:20.100]  you give the machine a single scalar.\n",
            "[06:20.100 --> 06:22.180]  You tell the machine you did good, you did bad.\n",
            "[06:22.180 --> 06:26.420]  And you only tell this to the machine once in a while.\n",
            "[06:26.420 --> 06:28.220]  When I say you, it could be the universe\n",
            "[06:28.220 --> 06:29.620]  telling the machine, right?\n",
            "[06:30.900 --> 06:32.620]  But it's just one scalar.\n",
            "[06:32.620 --> 06:33.940]  And so as a consequence,\n",
            "[06:33.940 --> 06:36.380]  you cannot possibly learn something very complicated\n",
            "[06:36.380 --> 06:37.900]  without many, many, many trials\n",
            "[06:37.900 --> 06:41.540]  where you get many, many feedbacks of this type.\n",
            "[06:41.540 --> 06:45.660]  Supervised learning, you give a few bits to the machine\n",
            "[06:45.660 --> 06:47.000]  at every sample.\n",
            "[06:48.860 --> 06:51.340]  Let's say you're training a system\n",
            "[06:51.340 --> 06:53.940]  on recognizing images on ImageNet,\n",
            "[06:53.940 --> 06:55.300]  there is 1,000 categories,\n",
            "[06:55.300 --> 06:58.540]  that's a little less than 10 bits of information per sample.\n",
            "[07:00.400 --> 07:02.060]  But self-supervised learning, here is the setting.\n",
            "[07:02.060 --> 07:04.580]  You ideally, we don't know how to do this yet,\n",
            "[07:04.580 --> 07:09.580]  but ideally you would show a machine a segment of a video\n",
            "[07:09.880 --> 07:12.420]  and then stop the video and ask the machine to predict\n",
            "[07:12.420 --> 07:13.820]  what's going to happen next.\n",
            "[07:15.500 --> 07:16.620]  And so you let the machine predict\n",
            "[07:16.620 --> 07:19.300]  and then you let time go by\n",
            "[07:19.300 --> 07:22.260]  and show the machine what actually happened\n",
            "[07:22.260 --> 07:25.840]  and hope the machine will learn to do a better job\n",
            "[07:25.840 --> 07:27.300]  at predicting next time around.\n",
            "[07:27.300 --> 07:29.500]  There's a huge amount of information you give the machine\n",
            "[07:29.500 --> 07:31.720]  because it's an entire video clip.\n",
            "[07:31.720 --> 07:35.860]  It's of the future after the video clip\n",
            "[07:35.860 --> 07:38.020]  you fed it in the first place.\n",
            "[07:39.860 --> 07:42.460]  So both for language and for vision,\n",
            "[07:42.460 --> 07:46.540]  there's a subtle, seemingly trivial construction,\n",
            "[07:46.540 --> 07:48.140]  but maybe that's representative\n",
            "[07:48.140 --> 07:50.240]  of what is required to create intelligence,\n",
            "[07:50.240 --> 07:52.420]  which is filling the gap.\n",
            "[07:54.320 --> 07:55.420]  So. Filling the gaps.\n",
            "[07:57.060 --> 07:57.980]  Filling the gaps.\n",
            "[07:57.980 --> 08:02.980]  It sounds dumb, but can you,\n",
            "[08:03.620 --> 08:06.460]  it is possible that you could solve\n",
            "[08:06.460 --> 08:08.300]  all of intelligence in this way.\n",
            "[08:08.300 --> 08:10.560]  Just for both language,\n",
            "[08:10.560 --> 08:14.080]  just give a sentence and continue it\n",
            "[08:14.080 --> 08:16.440]  or give a sentence and there's a gap in it.\n",
            "[08:17.360 --> 08:20.980]  Some words blanked out and you fill in what words go there.\n",
            "[08:20.980 --> 08:24.460]  For vision, you give a sequence of images\n",
            "[08:24.460 --> 08:26.220]  and predict what's going to happen next\n",
            "[08:26.220 --> 08:29.100]  or you fill in what happened in between.\n",
            "[08:29.100 --> 08:32.140]  Do you think it's possible that formulation alone\n",
            "[08:34.100 --> 08:36.460]  as a signal for self-supervised learning\n",
            "[08:36.460 --> 08:39.140]  can solve intelligence for vision and language?\n",
            "[08:40.980 --> 08:43.660]  I think that's our best shot at the moment.\n",
            "[08:43.660 --> 08:46.500]  So whether this will take us all the way\n",
            "[08:46.500 --> 08:49.140]  to human level intelligence or something\n",
            "[08:49.140 --> 08:52.220]  or just cat level intelligence is not clear,\n",
            "[08:52.220 --> 08:54.700]  but among all the possible approaches\n",
            "[08:54.700 --> 08:56.880]  that people have proposed, I think it's our best shot.\n",
            "[08:56.880 --> 09:01.880]  So I think this idea of an intelligence system\n",
            "[09:02.020 --> 09:06.240]  filling in the blanks, either predicting the future,\n",
            "[09:06.240 --> 09:09.560]  inferring the past, filling in missing information.\n",
            "[09:11.140 --> 09:14.040]  I'm currently filling the blank of what is behind your head\n",
            "[09:14.040 --> 09:17.980]  and what your head looks like from the back\n",
            "[09:17.980 --> 09:21.140]  because I have basic knowledge about how humans are made.\n",
            "[09:21.140 --> 09:22.900]  And I don't know if you're gonna,\n",
            "[09:22.900 --> 09:23.740]  what word you're gonna say,\n",
            "[09:23.740 --> 09:24.660]  which point you're gonna speak,\n",
            "[09:24.660 --> 09:26.340]  whether you're gonna move your head this way or that way,\n",
            "[09:26.340 --> 09:27.640]  which way you're gonna look,\n",
            "[09:27.640 --> 09:29.460]  but I know you're not gonna just dematerialize\n",
            "[09:29.460 --> 09:32.220]  and reappear three meters down the hall\n",
            "[09:33.640 --> 09:36.900]  because I know what's possible and what's impossible\n",
            "[09:36.900 --> 09:40.300]  according to intuitive physics.\n",
            "[09:40.300 --> 09:42.020]  So you have a model of what's possible,\n",
            "[09:42.020 --> 09:43.900]  what's impossible and then you'd be very surprised\n",
            "[09:43.900 --> 09:46.940]  if it happens and then you'll have to reconstruct your model.\n",
            "[09:50.940 --> 09:52.700]  Right, so that's the model of the world.\n",
            "[09:52.700 --> 09:55.380]  It's what tells you, you know, what fills in the blanks.\n",
            "[09:55.380 --> 09:57.580]  So given your partial information\n",
            "[09:57.580 --> 10:00.280]  about the state of the world, given by your perception,\n",
            "[10:01.180 --> 10:04.460]  your model of the world fills in the missing information\n",
            "[10:04.460 --> 10:06.860]  and that includes predicting the future,\n",
            "[10:06.860 --> 10:09.300]  re-predicting the past, you know,\n",
            "[10:09.300 --> 10:11.700]  filling in things you don't immediately perceive.\n",
            "[10:13.180 --> 10:17.060]  And that doesn't have to be purely generic vision\n",
            "[10:17.060 --> 10:19.140]  or visual information or generic language.\n",
            "[10:19.140 --> 10:23.700]  You can go to specifics like predicting\n",
            "[10:23.700 --> 10:25.080]  what control decision you make\n",
            "[10:25.080 --> 10:26.420]  when you're driving in a lane.\n",
            "[10:26.420 --> 10:30.420]  You have a sequence of images from a vehicle\n",
            "[10:30.420 --> 10:34.460]  and then you have information if you record it on video\n",
            "[10:34.460 --> 10:38.460]  where the car ended up going so you can go back in time\n",
            "[10:38.460 --> 10:40.320]  and predict where the car went\n",
            "[10:40.320 --> 10:41.480]  based on the visual information.\n",
            "[10:41.480 --> 10:43.580]  That's very specific, domain specific.\n",
            "[10:45.460 --> 10:47.500]  Right, but the question is whether we can come up\n",
            "[10:47.500 --> 10:52.420]  with sort of a generic method for, you know,\n",
            "[10:52.420 --> 10:54.560]  training machines to do this kind of prediction\n",
            "[10:54.560 --> 10:55.880]  or filling in the blanks.\n",
            "[10:55.880 --> 10:59.300]  So right now, this type of approach\n",
            "[10:59.300 --> 11:01.620]  has been unbelievably successful\n",
            "[11:01.620 --> 11:04.220]  in the context of natural language processing.\n",
            "[11:04.220 --> 11:05.740]  Every modern natural language processing\n",
            "[11:05.740 --> 11:07.660]  is pre-trained in self-supervised manner\n",
            "[11:10.020 --> 11:11.460]  to fill in the blanks.\n",
            "[11:11.460 --> 11:14.140]  You show it a sequence of words, you remove 10% of them\n",
            "[11:14.140 --> 11:15.660]  and then you train some gigantic neural net\n",
            "[11:15.660 --> 11:18.060]  to predict the words that are missing.\n",
            "[11:18.060 --> 11:20.460]  And once you've pre-trained that network,\n",
            "[11:20.460 --> 11:24.340]  you can use the internal representation learned by it\n",
            "[11:24.340 --> 11:27.500]  as input to, you know, something that you train,\n",
            "[11:27.500 --> 11:29.940]  supervised or whatever.\n",
            "[11:29.940 --> 11:31.120]  That's been incredibly successful.\n",
            "[11:31.120 --> 11:35.340]  Not so successful in images, although it's making progress\n",
            "[11:35.340 --> 11:40.180]  and it's based on sort of manual data augmentation.\n",
            "[11:40.180 --> 11:41.300]  We can go into this later,\n",
            "[11:41.300 --> 11:43.420]  but what has not been successful yet\n",
            "[11:43.420 --> 11:44.940]  is training from video.\n",
            "[11:44.940 --> 11:46.260]  So getting a machine to learn,\n",
            "[11:46.260 --> 11:48.900]  to represent the visual world, for example,\n",
            "[11:50.780 --> 11:52.040]  by just watching video.\n",
            "[11:52.040 --> 11:54.140]  Nobody has really succeeded in doing this.\n",
            "[11:55.900 --> 11:58.660]  Okay, well, let's kind of give a high-level overview.\n",
            "[11:58.660 --> 12:03.500]  What's the difference in kind and in difficulty\n",
            "[12:03.500 --> 12:05.060]  between vision and language?\n",
            "[12:05.060 --> 12:07.340]  So you said people haven't been able\n",
            "[12:07.340 --> 12:11.580]  to really kind of crack the problem of vision open\n",
            "[12:11.580 --> 12:13.060]  in terms of self-supervised learning,\n",
            "[12:13.060 --> 12:14.900]  but that may not be necessarily\n",
            "[12:14.900 --> 12:16.960]  because it's fundamentally more difficult.\n",
            "[12:16.960 --> 12:19.820]  Maybe like when we're talking about achieving,\n",
            "[12:19.820 --> 12:23.420]  like passing the Turing test in the full spirit\n",
            "[12:23.420 --> 12:26.020]  of the Turing test in language might be harder than vision.\n",
            "[12:26.020 --> 12:27.500]  That's not obvious.\n",
            "[12:27.500 --> 12:30.520]  So in your view, which is harder,\n",
            "[12:30.520 --> 12:32.740]  or perhaps are they just the same problem?\n",
            "[12:35.860 --> 12:37.520]  The farther we get to solving each,\n",
            "[12:37.520 --> 12:39.380]  the more we realize it's all the same thing.\n",
            "[12:39.380 --> 12:40.340]  It's all the same cake.\n",
            "[12:40.340 --> 12:44.860]  I think what I'm looking for are methods\n",
            "[12:44.860 --> 12:48.260]  that make them look essentially like the same cake,\n",
            "[12:48.260 --> 12:49.460]  but currently they're not.\n",
            "[12:49.460 --> 12:53.160]  And the main issue with learning world models\n",
            "[12:53.160 --> 12:57.820]  or learning predictive models is that the prediction\n",
            "[12:57.820 --> 13:00.580]  is never a single thing\n",
            "[13:00.580 --> 13:03.900]  because the world is not entirely predictable.\n",
            "[13:03.900 --> 13:05.380]  It may be deterministic or stochastic.\n",
            "[13:05.380 --> 13:07.620]  We can get into the philosophical discussion about it,\n",
            "[13:07.620 --> 13:09.940]  but even if it's deterministic,\n",
            "[13:09.940 --> 13:12.100]  it's not entirely predictable.\n",
            "[13:12.100 --> 13:16.420]  And so if I play a short video clip\n",
            "[13:16.420 --> 13:18.820]  and then I ask you to predict what's going to happen next,\n",
            "[13:18.820 --> 13:21.020]  there's many, many plausible continuations\n",
            "[13:21.020 --> 13:23.020]  for that video clip.\n",
            "[13:23.020 --> 13:25.220]  And the number of continuation grows\n",
            "[13:25.220 --> 13:28.580]  with the interval of time that you're asking the system\n",
            "[13:28.580 --> 13:31.140]  to make a prediction for.\n",
            "[13:31.140 --> 13:34.560]  And so one big question with self-supervised learning\n",
            "[13:34.560 --> 13:37.020]  is how you represent this uncertainty,\n",
            "[13:37.020 --> 13:39.900]  how you represent multiple discrete outcomes,\n",
            "[13:39.900 --> 13:41.780]  how you represent a sort of continuum\n",
            "[13:41.780 --> 13:45.100]  of possible outcomes, et cetera.\n",
            "[13:45.100 --> 13:49.900]  And if you are sort of a classical machine learning person,\n",
            "[13:49.900 --> 13:52.700]  you say, oh, you just represent a distribution, right?\n",
            "[13:54.460 --> 13:55.300]  Yeah.\n",
            "[13:56.700 --> 14:00.140]  And that we know how to do when we're predicting words,\n",
            "[14:00.140 --> 14:01.280]  missing words in the text,\n",
            "[14:01.280 --> 14:04.420]  because you can have a neural net give a score\n",
            "[14:04.420 --> 14:06.180]  for every word in the dictionary.\n",
            "[14:06.180 --> 14:08.380]  It's a big list of numbers,\n",
            "[14:08.380 --> 14:10.060]  maybe 100,000 or so.\n",
            "[14:10.060 --> 14:12.820]  And you can turn them into a probability distribution\n",
            "[14:12.820 --> 14:15.220]  that tells you when I say a sentence,\n",
            "[14:17.420 --> 14:19.920]  the cat is chasing the blank in the kitchen.\n",
            "[14:22.380 --> 14:25.080]  There are only a few words that make sense there.\n",
            "[14:25.080 --> 14:27.620]  It could be a mouse or it could be a lizard spot\n",
            "[14:27.620 --> 14:29.180]  or something like that, right?\n",
            "[14:30.780 --> 14:35.060]  And if I say the blank is chasing the blank in the savanna,\n",
            "[14:35.060 --> 14:37.060]  you also have a bunch of plausible options\n",
            "[14:37.060 --> 14:38.460]  for those two words, right?\n",
            "[14:40.180 --> 14:42.860]  Because you have kind of an underlying reality\n",
            "[14:42.860 --> 14:45.380]  that you can refer to sort of fill in those blanks.\n",
            "[14:47.320 --> 14:51.260]  So you cannot say for sure in the savanna,\n",
            "[14:51.260 --> 14:53.700]  if it's a lion or a cheetah or whatever,\n",
            "[14:53.700 --> 14:58.700]  you cannot know if it's a zebra or a do or whatever,\n",
            "[14:58.780 --> 15:00.180]  wildebeest, the same thing.\n",
            "[15:00.180 --> 15:05.180]  But you can represent the uncertainty\n",
            "[15:08.260 --> 15:09.900]  by just a long list of numbers.\n",
            "[15:09.900 --> 15:14.420]  Now, if I do the same thing with video\n",
            "[15:14.420 --> 15:16.940]  and I ask you to predict a video clip,\n",
            "[15:16.940 --> 15:20.020]  it's not a discrete set of potential frames.\n",
            "[15:20.020 --> 15:22.620]  You have to have somewhere representing\n",
            "[15:22.620 --> 15:26.180]  a sort of infinite number of plausible continuations\n",
            "[15:26.180 --> 15:30.100]  of multiple frames in a high dimensional continuous space.\n",
            "[15:30.100 --> 15:32.540]  And we just have no idea how to do this properly.\n",
            "[15:34.340 --> 15:36.300]  Finite high dimensional.\n",
            "[15:38.260 --> 15:39.100]  So like you,\n",
            "[15:39.100 --> 15:42.820]  it's finite high dimensional, yes.\n",
            "[15:44.660 --> 15:45.620]  Just like the words,\n",
            "[15:45.620 --> 15:50.620]  they try to get it down to a small finite set\n",
            "[15:51.580 --> 15:53.580]  of like under a million, something like that.\n",
            "[15:53.580 --> 15:54.420]  Something like that.\n",
            "[15:54.420 --> 15:55.380]  I mean, it's kind of ridiculous\n",
            "[15:55.380 --> 15:58.380]  that we're doing a distribution\n",
            "[15:58.380 --> 16:02.260]  of every single possible word for language and it works.\n",
            "[16:02.260 --> 16:04.660]  It feels like that's a really dumb way to do it.\n",
            "[16:07.020 --> 16:07.860]  Something like that.\n",
            "[16:07.860 --> 16:12.860]  I mean, it seems to be like there should be\n",
            "[16:13.740 --> 16:16.940]  some more compressed representation\n",
            "[16:16.940 --> 16:19.180]  of the distribution of the words.\n",
            "[16:21.020 --> 16:22.220]  You're right about that.\n",
            "[16:24.220 --> 16:25.260]  And so- I agree.\n",
            "[16:25.260 --> 16:26.980]  Do you have any interesting ideas\n",
            "[16:26.980 --> 16:29.940]  about how to represent all of reality in a compressed way\n",
            "[16:29.940 --> 16:32.340]  such that you can form a distribution over it?\n",
            "[16:33.740 --> 16:34.980]  That's one of the big questions.\n",
            "[16:34.980 --> 16:36.140]  How do you do that?\n",
            "[16:36.140 --> 16:37.940]  I mean, what's kind of,\n",
            "[16:37.940 --> 16:42.060]  another thing that really is stupid about,\n",
            "[16:42.060 --> 16:43.020]  I shouldn't say stupid,\n",
            "[16:43.020 --> 16:45.460]  but like simplistic about current approaches\n",
            "[16:45.460 --> 16:49.260]  to self-supervised running in NLP in text\n",
            "[16:49.260 --> 16:51.860]  is that not only do you represent\n",
            "[16:51.860 --> 16:53.740]  a giant distribution over words,\n",
            "[16:53.740 --> 16:55.580]  but for multiple words that are missing,\n",
            "[16:55.580 --> 16:57.620]  those distributions are essentially independent\n",
            "[16:57.620 --> 16:58.460]  of each other.\n",
            "[17:00.100 --> 17:02.980]  And you don't pay too much of a price for this.\n",
            "[17:02.980 --> 17:06.700]  So you can't, so the system,\n",
            "[17:06.700 --> 17:09.660]  in the sentence that I gave earlier,\n",
            "[17:09.660 --> 17:13.580]  if it gives a certain probability for lion and cheetah,\n",
            "[17:13.580 --> 17:17.700]  and then a certain probability for gazelle,\n",
            "[17:17.700 --> 17:20.220]  wildebeest and zebra,\n",
            "[17:20.220 --> 17:24.700]  those two probabilities are independent of each other.\n",
            "[17:26.700 --> 17:29.180]  And it's not the case that those things are independent.\n",
            "[17:29.180 --> 17:32.580]  Lions actually attack like bigger animals than cheetahs.\n",
            "[17:32.580 --> 17:35.700]  So, you know, there's a huge independent hypothesis\n",
            "[17:35.700 --> 17:38.940]  in this process, which is not actually true.\n",
            "[17:38.940 --> 17:41.020]  The reason for this is that we don't know\n",
            "[17:41.020 --> 17:44.140]  how to represent properly distributions\n",
            "[17:44.140 --> 17:47.380]  over combinatorial sequences of symbols,\n",
            "[17:47.380 --> 17:48.500]  essentially when they're,\n",
            "[17:48.500 --> 17:50.140]  because the number grows exponentially\n",
            "[17:50.140 --> 17:52.460]  with the length of the symbols.\n",
            "[17:52.460 --> 17:53.900]  And so we have to use tricks for this,\n",
            "[17:53.900 --> 17:57.540]  but those techniques can, you know, get around,\n",
            "[17:57.540 --> 17:58.940]  like don't even deal with it.\n",
            "[17:58.940 --> 18:01.620]  So the big question is like,\n",
            "[18:01.620 --> 18:04.540]  would there be some sort of abstract\n",
            "[18:04.540 --> 18:08.140]  latent representation of text that would say that,\n",
            "[18:08.140 --> 18:11.820]  you know, when I switch lion for gazelle,\n",
            "[18:11.820 --> 18:20.180]  or lion for cheetah, I also have to switch zebra for gazelle?\n",
            "[18:20.180 --> 18:23.420]  Yeah, so this independence assumption,\n",
            "[18:23.420 --> 18:25.820]  let me throw some criticism at you that I often hear\n",
            "[18:25.820 --> 18:29.020]  and see how you respond.\n",
            "[18:29.020 --> 18:32.100]  So this kind of filling in the blanks is just statistics.\n",
            "[18:32.100 --> 18:34.980]  You're not learning anything\n",
            "[18:34.980 --> 18:37.700]  like the deep underlying concepts.\n",
            "[18:37.700 --> 18:41.740]  You're just mimicking stuff from the past.\n",
            "[18:41.740 --> 18:43.620]  You're not learning anything new\n",
            "[18:43.620 --> 18:48.060]  such that you can use it to generalize about the world.\n",
            "[18:48.060 --> 18:50.180]  Or, okay, let me just say the crude version,\n",
            "[18:50.180 --> 18:52.300]  which is just statistics.\n",
            "[18:52.300 --> 18:54.420]  It's not intelligence.\n",
            "[18:54.420 --> 18:55.700]  What do you have to say to that?\n",
            "[18:55.700 --> 18:56.980]  What do you usually say to that\n",
            "[18:56.980 --> 19:00.340]  if you kind of hear this kind of thing?\n",
            "[19:00.340 --> 19:01.660]  I don't get into those discussions\n",
            "[19:01.660 --> 19:04.460]  because they are kind of pointless.\n",
            "[19:04.460 --> 19:06.460]  So first of all, it's quite possible\n",
            "[19:06.460 --> 19:08.180]  that intelligence is just statistics.\n",
            "[19:08.180 --> 19:10.420]  It's just statistics of a particular kind.\n",
            "[19:10.420 --> 19:11.340]  Yes.\n",
            "[19:11.340 --> 19:15.500]  But this is the philosophical question.\n",
            "[19:15.500 --> 19:20.340]  Is it possible that intelligence is just statistics?\n",
            "[19:22.220 --> 19:23.060]  Yeah.\n",
            "[19:24.300 --> 19:26.220]  But what kind of statistics?\n",
            "[19:26.220 --> 19:28.900]  So if you're asking the question,\n",
            "[19:30.820 --> 19:34.300]  are the models of the world that we learn,\n",
            "[19:34.300 --> 19:35.940]  do they have some notion of causality?\n",
            "[19:35.940 --> 19:37.020]  Yes.\n",
            "[19:37.020 --> 19:39.860]  So if the criticism comes from people who say,\n",
            "[19:39.860 --> 19:42.860]  current machine learning system don't care about causality,\n",
            "[19:42.860 --> 19:44.060]  which by the way is wrong,\n",
            "[19:45.780 --> 19:47.020]  I agree with them.\n",
            "[19:47.020 --> 19:48.820]  You should, your model of the world\n",
            "[19:48.820 --> 19:52.540]  should have your actions as one of the inputs.\n",
            "[19:52.540 --> 19:54.860]  And that will drive you to learn causal models of the world\n",
            "[19:54.860 --> 19:58.500]  where you know what intervention in the world\n",
            "[19:58.500 --> 20:00.140]  will cause what result.\n",
            "[20:00.140 --> 20:02.860]  Or you can do this by observation of other agents\n",
            "[20:02.860 --> 20:05.940]  acting in the world and observing the effect,\n",
            "[20:05.940 --> 20:07.660]  other humans, for example.\n",
            "[20:07.660 --> 20:11.860]  So I think at some level of description,\n",
            "[20:11.860 --> 20:13.660]  intelligence is just statistics.\n",
            "[20:15.620 --> 20:16.460]  Yes.\n",
            "[20:26.380 --> 20:29.940]  But that doesn't mean you don't have models\n",
            "[20:29.940 --> 20:34.820]  that have deep mechanistic explanation for what goes on.\n",
            "[20:34.820 --> 20:36.500]  The question is how do you learn them?\n",
            "[20:36.500 --> 20:37.420]  That's the question.\n",
            "[20:37.420 --> 20:38.260]  I'm interested in.\n",
            "[20:39.180 --> 20:44.180]  Because a lot of people who actually voice their criticism\n",
            "[20:47.820 --> 20:49.460]  say that those mechanistic model\n",
            "[20:49.460 --> 20:51.100]  have to come from someplace else.\n",
            "[20:51.100 --> 20:52.500]  They have to come from human designers,\n",
            "[20:52.500 --> 20:54.660]  they have to come from I don't know what.\n",
            "[20:54.660 --> 20:56.340]  And obviously we learn them.\n",
            "[21:00.820 --> 21:03.340]  Or if we don't learn them as an individual,\n",
            "[21:03.340 --> 21:06.460]  nature learn them for us using evolution.\n",
            "[21:06.460 --> 21:08.700]  So regardless of what you think,\n",
            "[21:08.700 --> 21:10.900]  those processes have been learned somehow.\n",
            "[21:12.780 --> 21:15.460]  So if you look at the human brain,\n",
            "[21:15.460 --> 21:17.180]  just like when we humans introspect\n",
            "[21:17.180 --> 21:18.860]  about how the brain works,\n",
            "[21:18.860 --> 21:22.780]  it seems like when we think about what is intelligence,\n",
            "[21:22.780 --> 21:24.980]  we think about the high level stuff\n",
            "[21:24.980 --> 21:26.460]  like the models we've constructed,\n",
            "[21:26.460 --> 21:28.100]  concepts like cognitive science,\n",
            "[21:28.100 --> 21:31.220]  like concepts of memory and reasoning module,\n",
            "[21:31.220 --> 21:34.180]  almost like these high level modules.\n",
            "[21:34.180 --> 21:37.060]  Does this serve as a good analogy?\n",
            "[21:39.620 --> 21:44.260]  Are we ignoring the dark matter,\n",
            "[21:44.260 --> 21:47.100]  the basic low level mechanisms,\n",
            "[21:47.100 --> 21:49.340]  just like we ignore the way the operating system works,\n",
            "[21:49.340 --> 21:53.180]  we're just using the high level software.\n",
            "[21:53.180 --> 21:56.300]  We're ignoring that at the low level,\n",
            "[21:56.300 --> 21:59.500]  the neural network might be doing something like statistics.\n",
            "[22:01.500 --> 22:02.580]  Sorry to use this word.\n",
            "[22:02.580 --> 22:07.580]  It probably.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Reading the transcription file."
      ],
      "metadata": {
        "id": "HNauX5fXqODD"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install -U webvtt-py"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aEPsEuS2k4yf",
        "outputId": "6e4b76b7-2113-4370-d080-d6c8c7baf003"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting webvtt-py\n",
            "  Downloading webvtt_py-0.4.6-py3-none-any.whl (16 kB)\n",
            "Requirement already satisfied: docopt in /usr/local/lib/python3.7/dist-packages (from webvtt-py) (0.6.2)\n",
            "Installing collected packages: webvtt-py\n",
            "Successfully installed webvtt-py-0.4.6\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import webvtt\n",
        "\n",
        "captions = [[(int)(millisec(caption.start)), (int)(millisec(caption.end)),  caption.text] for caption in webvtt.read('dz.wav.vtt')]\n",
        "print(*captions[:8], sep='\\n')"
      ],
      "metadata": {
        "id": "uNh_GUpNkyxT",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7614fc86-b575-42e5-93a1-1629789108cb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0, 4720, 'The following is a conversation with Yann LeCun,']\n",
            "[4720, 6560, 'his second time on the podcast.']\n",
            "[6560, 11160, 'He is the chief AI scientist at Meta, formerly Facebook,']\n",
            "[11160, 15040, 'professor at NYU, touring award winner,']\n",
            "[15040, 17600, 'one of the seminal figures in the history']\n",
            "[17600, 20460, 'of machine learning and artificial intelligence,']\n",
            "[20460, 23940, 'and someone who is brilliant and opinionated']\n",
            "[23940, 25400, 'in the best kind of way,']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Matching the Transcriptions and the Diarizations"
      ],
      "metadata": {
        "id": "u_UyWQMXpB3N"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Matching each trainscrition line to some diarizations, and generating the HTML file. To get the correct timing, we should take care of the parts in original audio that were in no diarization segment."
      ],
      "metadata": {
        "id": "KndDYy_xMpMq"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vKdx9Hwg630K"
      },
      "outputs": [],
      "source": [
        "preS = '<!DOCTYPE html>\\n<html lang=\"en\">\\n  <head>\\n    <meta charset=\"UTF-8\">\\n    <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\\n    <meta http-equiv=\"X-UA-Compatible\" content=\"ie=edge\">\\n    <title>Lexicap</title>\\n    <style>\\n        body {\\n            font-family: sans-serif;\\n            font-size: 18px;\\n            color: #111;\\n            padding: 0 0 1em 0;\\n        }\\n        .l {\\n          color: #050;\\n        }\\n        .s {\\n            display: inline-block;\\n        }\\n        .e {\\n            display: inline-block;\\n        }\\n        .t {\\n            display: inline-block;\\n        }\\n        #player {\\n\\t\\tposition: sticky;\\n\\t\\ttop: 20px;\\n\\t\\tfloat: right;\\n\\t}\\n    </style>\\n  </head>\\n  <body>\\n    <h2>Yann LeCun: Dark Matter of Intelligence and Self-Supervised Learning | Lex Fridman Podcast #258</h2>\\n  <div  id=\"player\"></div>\\n    <script>\\n      var tag = document.createElement(\\'script\\');\\n      tag.src = \"https://www.youtube.com/iframe_api\";\\n      var firstScriptTag = document.getElementsByTagName(\\'script\\')[0];\\n      firstScriptTag.parentNode.insertBefore(tag, firstScriptTag);\\n      var player;\\n      function onYouTubeIframeAPIReady() {\\n        player = new YT.Player(\\'player\\', {\\n          height: \\'210\\',\\n          width: \\'340\\',\\n          videoId: \\'SGzMElJ11Cc\\',\\n        });\\n      }\\n      function setCurrentTime(timepoint) {\\n        player.seekTo(timepoint);\\n   player.playVideo();\\n   }\\n    </script><br>\\n'\n",
        "postS = '\\t</body>\\n</html>'"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from datetime import timedelta\n",
        "\n",
        "html = list(preS)\n",
        "\n",
        "for i in range(len(segments)):\n",
        "  idx = 0\n",
        "  for idx in range(len(captions)):\n",
        "    if captions[idx][0] >= (segments[i] - spacermilli):\n",
        "      break;\n",
        "  \n",
        "  while (idx < (len(captions))) and ((i == len(segments) - 1) or (captions[idx][1] < segments[i+1])):\n",
        "    c = captions[idx]  \n",
        "    \n",
        "    start = dzList[i][0] + (c[0] -segments[i])\n",
        "\n",
        "    if start < 0: \n",
        "      start = 0\n",
        "    idx += 1\n",
        "\n",
        "    start = start / 1000.0\n",
        "    startStr = '{0:02d}:{1:02d}:{2:02.2f}'.format((int)(start // 3600), \n",
        "                                            (int)(start % 3600 // 60), \n",
        "                                            start % 60)\n",
        "    \n",
        "    html.append('\\t\\t\\t<div class=\"c\">\\n')\n",
        "    html.append(f'\\t\\t\\t\\t<a class=\"l\" href=\"#{startStr}\" id=\"{startStr}\">link</a> |\\n')\n",
        "    html.append(f'\\t\\t\\t\\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime({int(start)})>{startStr}</a></div>\\n')\n",
        "    html.append(f'\\t\\t\\t\\t<div class=\"t\">{\"[Lex]\" if dzList[i][2] else \"[Yann]\"} {c[2]}</div>\\n')\n",
        "    html.append('\\t\\t\\t</div>\\n\\n')\n",
        "\n",
        "html.append(postS)\n",
        "s = \"\".join(html)\n",
        "\n",
        "with open(\"lexicap.html\", \"w\") as text_file:\n",
        "    text_file.write(s)\n",
        "print(s)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vqO6Nd6YfZYa",
        "outputId": "fce0a6b9-c28f-43f9-afd6-cb55998cd965"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "<!DOCTYPE html>\n",
            "<html lang=\"en\">\n",
            "  <head>\n",
            "    <meta charset=\"UTF-8\">\n",
            "    <meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n",
            "    <meta http-equiv=\"X-UA-Compatible\" content=\"ie=edge\">\n",
            "    <title>Lexicap</title>\n",
            "    <style>\n",
            "        body {\n",
            "            font-family: sans-serif;\n",
            "            font-size: 18px;\n",
            "            color: #111;\n",
            "            padding: 0 0 1em 0;\n",
            "        }\n",
            "        .l {\n",
            "          color: #050;\n",
            "        }\n",
            "        .s {\n",
            "            display: inline-block;\n",
            "        }\n",
            "        .e {\n",
            "            display: inline-block;\n",
            "        }\n",
            "        .t {\n",
            "            display: inline-block;\n",
            "        }\n",
            "        #player {\n",
            "\t\tposition: sticky;\n",
            "\t\ttop: 20px;\n",
            "\t\tfloat: right;\n",
            "\t}\n",
            "    </style>\n",
            "  </head>\n",
            "  <body>\n",
            "    <h2>Yann LeCun: Dark Matter of Intelligence and Self-Supervised Learning | Lex Fridman Podcast #258</h2>\n",
            "  <div  id=\"player\"></div>\n",
            "    <script>\n",
            "      var tag = document.createElement('script');\n",
            "      tag.src = \"https://www.youtube.com/iframe_api\";\n",
            "      var firstScriptTag = document.getElementsByTagName('script')[0];\n",
            "      firstScriptTag.parentNode.insertBefore(tag, firstScriptTag);\n",
            "      var player;\n",
            "      function onYouTubeIframeAPIReady() {\n",
            "        player = new YT.Player('player', {\n",
            "          height: '210',\n",
            "          width: '340',\n",
            "          videoId: 'SGzMElJ11Cc',\n",
            "        });\n",
            "      }\n",
            "      function setCurrentTime(timepoint) {\n",
            "        player.seekTo(timepoint);\n",
            "   player.playVideo();\n",
            "   }\n",
            "    </script><br>\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:0.00\" id=\"00:00:0.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(0)>00:00:0.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] The following is a conversation with Yann LeCun,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:2.75\" id=\"00:00:2.75\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(2)>00:00:2.75</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] his second time on the podcast.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:4.59\" id=\"00:00:4.59\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(4)>00:00:4.59</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] He is the chief AI scientist at Meta, formerly Facebook,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:9.19\" id=\"00:00:9.19\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(9)>00:00:9.19</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] professor at NYU, touring award winner,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:13.07\" id=\"00:00:13.07\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(13)>00:00:13.07</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] one of the seminal figures in the history</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:15.63\" id=\"00:00:15.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(15)>00:00:15.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] of machine learning and artificial intelligence,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:18.49\" id=\"00:00:18.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(18)>00:00:18.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] and someone who is brilliant and opinionated</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:21.97\" id=\"00:00:21.97\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(21)>00:00:21.97</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] in the best kind of way,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:23.43\" id=\"00:00:23.43\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(23)>00:00:23.43</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] and so is always fun to talk to.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:25.99\" id=\"00:00:25.99\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(25)>00:00:25.99</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] This is the Lex Friedman podcast.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:27.99\" id=\"00:00:27.99\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(27)>00:00:27.99</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] To support it, please check out our sponsors</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:29.95\" id=\"00:00:29.95\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(29)>00:00:29.95</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] in the description.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:31.23\" id=\"00:00:31.23\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(31)>00:00:31.23</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] And now, here's my conversation with Yann LeCun.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:36.15\" id=\"00:00:36.15\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(36)>00:00:36.15</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] You co-wrote the article,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:37.49\" id=\"00:00:37.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(37)>00:00:37.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] \"'Self-Supervised Learning,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:38.85\" id=\"00:00:38.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(38)>00:00:38.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] the Dark Matter of Intelligence.\"</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:40.89\" id=\"00:00:40.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(40)>00:00:40.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Great title, by the way, with Ishan Mizra.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:43.71\" id=\"00:00:43.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(43)>00:00:43.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So let me ask, what is self-supervised learning,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:46.61\" id=\"00:00:46.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(46)>00:00:46.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] and why is it the dark matter of intelligence?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:49.90\" id=\"00:00:49.90\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(49)>00:00:49.90</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I'll start by the dark matter part.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:51.70\" id=\"00:00:51.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(51)>00:00:51.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] There is obviously a kind of learning</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:55.70\" id=\"00:00:55.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(55)>00:00:55.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that humans and animals are doing</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:00:59.90\" id=\"00:00:59.90\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(59)>00:00:59.90</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that we currently are not reproducing properly</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:2.86\" id=\"00:01:2.86\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(62)>00:01:2.86</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] with machines or with AI, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:4.70\" id=\"00:01:4.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(64)>00:01:4.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So the most popular approaches</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:6.30\" id=\"00:01:6.30\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(66)>00:01:6.30</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to machine learning today are,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:8.50\" id=\"00:01:8.50\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(68)>00:01:8.50</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] or paradigms, I should say,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:9.70\" id=\"00:01:9.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(69)>00:01:9.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] are supervised learning and reinforcement learning.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:12.74\" id=\"00:01:12.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(72)>00:01:12.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And they are extremely inefficient.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:15.16\" id=\"00:01:15.16\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(75)>00:01:15.16</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Supervised learning requires many samples</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:17.66\" id=\"00:01:17.66\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(77)>00:01:17.66</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] for learning anything,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:19.78\" id=\"00:01:19.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(79)>00:01:19.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and reinforcement learning requires</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:21.82\" id=\"00:01:21.82\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(81)>00:01:21.82</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] a ridiculously large number of trial and errors</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:25.24\" id=\"00:01:25.24\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(85)>00:01:25.24</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] for a system to learn anything.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:29.32\" id=\"00:01:29.32\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(89)>00:01:29.32</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And that's why we don't have self-driving cars.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:32.95\" id=\"00:01:32.95\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(92)>00:01:32.95</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] That was a big leap from one to the other.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:34.79\" id=\"00:01:34.79\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(94)>00:01:34.79</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Okay, so to solve difficult problems,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:38.77\" id=\"00:01:38.77\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(98)>00:01:38.77</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] you have to have a lot of human annotation</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:42.35\" id=\"00:01:42.35\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(102)>00:01:42.35</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] for supervised learning to work.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:44.07\" id=\"00:01:44.07\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(104)>00:01:44.07</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] And to solve those difficult problems</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:45.53\" id=\"00:01:45.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(105)>00:01:45.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] with reinforcement learning,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:46.69\" id=\"00:01:46.69\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(106)>00:01:46.69</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] you have to have some way to maybe simulate that problem</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:50.23\" id=\"00:01:50.23\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(110)>00:01:50.23</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] such that you can do that large-scale kind of learning</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:52.71\" id=\"00:01:52.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(112)>00:01:52.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] that reinforcement learning requires.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:54.51\" id=\"00:01:54.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(114)>00:01:54.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Right, so how is it that most teenagers</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:01:58.01\" id=\"00:01:58.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(118)>00:01:58.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] can learn to drive a car in about 20 hours of practice?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:2.29\" id=\"00:02:2.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(122)>00:02:2.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Whereas even with millions of hours of simulated practice,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:7.49\" id=\"00:02:7.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(127)>00:02:7.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] a self-driving car can't actually learn</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:9.21\" id=\"00:02:9.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(129)>00:02:9.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to drive itself properly.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:10.71\" id=\"00:02:10.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(130)>00:02:10.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And so, obviously, we're missing something, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:12.63\" id=\"00:02:12.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(132)>00:02:12.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And it's quite obvious for a lot of people</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:14.35\" id=\"00:02:14.35\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(134)>00:02:14.35</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that the immediate response you get from many people is,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:18.51\" id=\"00:02:18.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(138)>00:02:18.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] well, humans use their background knowledge</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:21.51\" id=\"00:02:21.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(141)>00:02:21.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to learn faster, and they're right.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:24.47\" id=\"00:02:24.47\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(144)>00:02:24.47</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Now, how was that background knowledge acquired?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:26.95\" id=\"00:02:26.95\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(146)>00:02:26.95</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And that's the big question.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:28.78\" id=\"00:02:28.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(148)>00:02:28.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So now you have to ask,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:31.14\" id=\"00:02:31.14\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(151)>00:02:31.14</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] how do babies in the first few months of life</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:33.78\" id=\"00:02:33.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(153)>00:02:33.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] learn how the world works?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:35.90\" id=\"00:02:35.90\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(155)>00:02:35.90</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Mostly by observation,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:36.90\" id=\"00:02:36.90\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(156)>00:02:36.90</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because they can hardly act in the world.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:39.34\" id=\"00:02:39.34\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(159)>00:02:39.34</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And they learn an enormous amount</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:41.02\" id=\"00:02:41.02\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(161)>00:02:41.02</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] of background knowledge about the world</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:42.30\" id=\"00:02:42.30\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(162)>00:02:42.30</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that may be the basis of what we call common sense.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:46.50\" id=\"00:02:46.50\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(166)>00:02:46.50</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] This type of learning is not learning a task,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:49.74\" id=\"00:02:49.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(169)>00:02:49.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] it's not being reinforced for anything,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:52.02\" id=\"00:02:52.02\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(172)>00:02:52.02</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] it's just observing the world and figuring out how it works.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:56.73\" id=\"00:02:56.73\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(176)>00:02:56.73</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Building world models, learning world models.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:02:59.69\" id=\"00:02:59.69\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(179)>00:02:59.69</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] How do we do this?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:0.53\" id=\"00:03:0.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(180)>00:03:0.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And how do we reproduce this in machines?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:3.01\" id=\"00:03:3.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(183)>00:03:3.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So self-supervised learning is a very, very important thing</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:6.01\" id=\"00:03:6.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(186)>00:03:6.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is one instance or one attempt</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:10.29\" id=\"00:03:10.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(190)>00:03:10.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] at trying to reproduce this kind of learning.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:13.07\" id=\"00:03:13.07\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(193)>00:03:13.07</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Okay, so you're looking at just observation,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:16.35\" id=\"00:03:16.35\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(196)>00:03:16.35</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] so not even the interacting part of a child.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:18.67\" id=\"00:03:18.67\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(198)>00:03:18.67</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] It's just sitting there watching mom and dad walk around,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:21.55\" id=\"00:03:21.55\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(201)>00:03:21.55</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] pick up stuff, all of that.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:23.43\" id=\"00:03:23.43\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(203)>00:03:23.43</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] That's what we mean by background knowledge.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:25.49\" id=\"00:03:25.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(205)>00:03:25.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Perhaps not even watching mom and dad,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:29.96\" id=\"00:03:29.96\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(209)>00:03:29.96</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Just having eyes open or having eyes closed</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:31.88\" id=\"00:03:31.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(211)>00:03:31.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or the very act of opening and closing eyes</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:34.44\" id=\"00:03:34.44\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(214)>00:03:34.44</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] that the world appears and disappears,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:36.24\" id=\"00:03:36.24\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(216)>00:03:36.24</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] all that basic information.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:39.10\" id=\"00:03:39.10\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(219)>00:03:39.10</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] And you're saying in order to learn to drive,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:43.10\" id=\"00:03:43.10\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(223)>00:03:43.10</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] like the reason humans are able to learn to drive quickly,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:45.82\" id=\"00:03:45.82\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(225)>00:03:45.82</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] some faster than others,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:47.34\" id=\"00:03:47.34\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(227)>00:03:47.34</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] is because of the background knowledge</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:51.71\" id=\"00:03:51.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(231)>00:03:51.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] in the many years leading up to it,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:53.59\" id=\"00:03:53.59\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(233)>00:03:53.59</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] the physics of basic objects, all that.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:55.72\" id=\"00:03:55.72\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(235)>00:03:55.72</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] That's right.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:56.56\" id=\"00:03:56.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(236)>00:03:56.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I mean, the basic physics of objects,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:03:57.40\" id=\"00:03:57.40\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(237)>00:03:57.40</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you don't even need to know how a car works</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:0.84\" id=\"00:04:0.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(240)>00:04:0.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because that you can learn fairly quickly.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:2.47\" id=\"00:04:2.47\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(242)>00:04:2.47</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I mean, the example I use very often</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:3.79\" id=\"00:04:3.79\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(243)>00:04:3.79</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is you're driving next to a cliff.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:6.62\" id=\"00:04:6.62\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(246)>00:04:6.62</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And you know in advance because of your understanding</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:10.54\" id=\"00:04:10.54\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(250)>00:04:10.54</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] of intuitive physics that if you turn the wheel</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:13.14\" id=\"00:04:13.14\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(253)>00:04:13.14</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to the right, the car will veer to the right,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:15.02\" id=\"00:04:15.02\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(255)>00:04:15.02</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] will run off the cliff, fall off the cliff,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:17.54\" id=\"00:04:17.54\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(257)>00:04:17.54</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and nothing good will come out of this, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:20.38\" id=\"00:04:20.38\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(260)>00:04:20.38</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] But if you are a sort of tabularized</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:23.74\" id=\"00:04:23.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(263)>00:04:23.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] reinforcement learning system</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:25.06\" id=\"00:04:25.06\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(265)>00:04:25.06</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that doesn't have a model of the world,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:28.14\" id=\"00:04:28.14\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(268)>00:04:28.14</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you have to repeat falling off this cliff</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:30.46\" id=\"00:04:30.46\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(270)>00:04:30.46</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] thousands of times before you figure out it's a bad idea.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:32.78\" id=\"00:04:32.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(272)>00:04:32.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And then a few more thousand times</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:34.54\" id=\"00:04:34.54\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(274)>00:04:34.54</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] before you figure out how to not do it.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:37.06\" id=\"00:04:37.06\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(277)>00:04:37.06</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And then a few more million times before you figure out</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:39.18\" id=\"00:04:39.18\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(279)>00:04:39.18</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] how to not do it in every situation you ever encounter.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:42.52\" id=\"00:04:42.52\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(282)>00:04:42.52</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So self-supervised learning still has to have</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:45.80\" id=\"00:04:45.80\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(285)>00:04:45.80</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] some source of truth being told to it by somebody.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:50.56\" id=\"00:04:50.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(290)>00:04:50.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So you have to figure out a way without human assistance</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:54.56\" id=\"00:04:54.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(294)>00:04:54.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or without significant amount of human assistance</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:56.60\" id=\"00:04:56.60\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(296)>00:04:56.60</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] to get that truth from the world.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:04:59.08\" id=\"00:04:59.08\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(299)>00:04:59.08</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So the mystery there is how much signal is there,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:3.96\" id=\"00:05:3.96\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(303)>00:05:3.96</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] how much truth is there that the world gives you,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:6.28\" id=\"00:05:6.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(306)>00:05:6.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] whether it's the human world, like you watch YouTube</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:9.12\" id=\"00:05:9.12\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(309)>00:05:9.12</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or something like that, or it's the more natural world.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:12.97\" id=\"00:05:12.97\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(312)>00:05:12.97</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So how much signal is there?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:14.90\" id=\"00:05:14.90\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(314)>00:05:14.90</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So here's the trick, there is way more signal</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:18.55\" id=\"00:05:18.55\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(318)>00:05:18.55</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] in sort of a self-supervised setting</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:20.59\" id=\"00:05:20.59\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(320)>00:05:20.59</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] than there is in either a supervised</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:22.51\" id=\"00:05:22.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(322)>00:05:22.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] or reinforcement setting.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:24.51\" id=\"00:05:24.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(324)>00:05:24.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And this is going to my analogy of the cake.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:28.39\" id=\"00:05:28.39\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(328)>00:05:28.39</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] The luck cake, as someone has called it,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:32.35\" id=\"00:05:32.35\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(332)>00:05:32.35</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] where when you try to figure out how much information</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:36.03\" id=\"00:05:36.03\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(336)>00:05:36.03</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you ask the machine to predict and how much feedback</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:38.71\" id=\"00:05:38.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(338)>00:05:38.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you give the machine at every trial.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:41.00\" id=\"00:05:41.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(341)>00:05:41.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] In reinforcement learning,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:41.84\" id=\"00:05:41.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(341)>00:05:41.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you give the machine a single scalar.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:43.28\" id=\"00:05:43.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(343)>00:05:43.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] You tell the machine you did good, you did bad.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:45.36\" id=\"00:05:45.36\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(345)>00:05:45.36</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And you only tell this to the machine once in a while.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:49.60\" id=\"00:05:49.60\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(349)>00:05:49.60</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] When I say you, it could be the universe</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:51.40\" id=\"00:05:51.40\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(351)>00:05:51.40</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] telling the machine, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:54.08\" id=\"00:05:54.08\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(354)>00:05:54.08</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] But it's just one scalar.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:55.80\" id=\"00:05:55.80\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(355)>00:05:55.80</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And so as a consequence,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:57.12\" id=\"00:05:57.12\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(357)>00:05:57.12</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you cannot possibly learn something very complicated</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:05:59.56\" id=\"00:05:59.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(359)>00:05:59.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] without many, many, many trials</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:1.08\" id=\"00:06:1.08\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(361)>00:06:1.08</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] where you get many, many feedbacks of this type.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:4.72\" id=\"00:06:4.72\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(364)>00:06:4.72</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Supervised learning, you give a few bits to the machine</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:8.84\" id=\"00:06:8.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(368)>00:06:8.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] at every sample.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:11.22\" id=\"00:06:11.22\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(371)>00:06:11.22</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Let's say you're training a system</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:13.70\" id=\"00:06:13.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(373)>00:06:13.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] on recognizing images on ImageNet,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:16.30\" id=\"00:06:16.30\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(376)>00:06:16.30</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] there is 1,000 categories,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:17.66\" id=\"00:06:17.66\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(377)>00:06:17.66</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that's a little less than 10 bits of information per sample.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:22.15\" id=\"00:06:22.15\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(382)>00:06:22.15</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] But self-supervised learning, here is the setting.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:23.81\" id=\"00:06:23.81\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(383)>00:06:23.81</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] You ideally, we don't know how to do this yet,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:26.33\" id=\"00:06:26.33\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(386)>00:06:26.33</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but ideally you would show a machine a segment of a video</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:31.63\" id=\"00:06:31.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(391)>00:06:31.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and then stop the video and ask the machine to predict</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:34.17\" id=\"00:06:34.17\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(394)>00:06:34.17</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] what's going to happen next.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:37.56\" id=\"00:06:37.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(397)>00:06:37.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And so you let the machine predict</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:38.68\" id=\"00:06:38.68\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(398)>00:06:38.68</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and then you let time go by</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:41.36\" id=\"00:06:41.36\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(401)>00:06:41.36</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and show the machine what actually happened</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:44.32\" id=\"00:06:44.32\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(404)>00:06:44.32</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and hope the machine will learn to do a better job</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:47.90\" id=\"00:06:47.90\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(407)>00:06:47.90</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] at predicting next time around.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:49.36\" id=\"00:06:49.36\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(409)>00:06:49.36</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] There's a huge amount of information you give the machine</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:51.56\" id=\"00:06:51.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(411)>00:06:51.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because it's an entire video clip.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:53.78\" id=\"00:06:53.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(413)>00:06:53.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] It's of the future after the video clip</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:06:57.92\" id=\"00:06:57.92\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(417)>00:06:57.92</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you fed it in the first place.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:0.24\" id=\"00:07:0.24\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(420)>00:07:0.24</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So both for language and for vision,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:2.84\" id=\"00:07:2.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(422)>00:07:2.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] there's a subtle, seemingly trivial construction,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:6.92\" id=\"00:07:6.92\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(426)>00:07:6.92</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] but maybe that's representative</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:8.52\" id=\"00:07:8.52\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(428)>00:07:8.52</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] of what is required to create intelligence,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:10.62\" id=\"00:07:10.62\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(430)>00:07:10.62</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] which is filling the gap.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:13.75\" id=\"00:07:13.75\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(433)>00:07:13.75</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So. Filling the gaps.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:13.78\" id=\"00:07:13.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(433)>00:07:13.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Filling the gaps.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:12.70\" id=\"00:07:12.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(432)>00:07:12.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] It sounds dumb, but can you,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:18.34\" id=\"00:07:18.34\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(438)>00:07:18.34</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] it is possible that you could solve</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:21.18\" id=\"00:07:21.18\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(441)>00:07:21.18</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] all of intelligence in this way.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:23.02\" id=\"00:07:23.02\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(443)>00:07:23.02</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Just for both language,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:25.28\" id=\"00:07:25.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(445)>00:07:25.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] just give a sentence and continue it</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:28.80\" id=\"00:07:28.80\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(448)>00:07:28.80</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or give a sentence and there's a gap in it.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:32.08\" id=\"00:07:32.08\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(452)>00:07:32.08</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Some words blanked out and you fill in what words go there.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:35.70\" id=\"00:07:35.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(455)>00:07:35.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] For vision, you give a sequence of images</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:39.18\" id=\"00:07:39.18\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(459)>00:07:39.18</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] and predict what's going to happen next</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:40.94\" id=\"00:07:40.94\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(460)>00:07:40.94</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or you fill in what happened in between.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:43.82\" id=\"00:07:43.82\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(463)>00:07:43.82</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Do you think it's possible that formulation alone</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:48.61\" id=\"00:07:48.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(468)>00:07:48.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] as a signal for self-supervised learning</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:50.96\" id=\"00:07:50.96\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(470)>00:07:50.96</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] can solve intelligence for vision and language?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:53.60\" id=\"00:07:53.60\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(473)>00:07:53.60</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I think that's our best shot at the moment.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:56.28\" id=\"00:07:56.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(476)>00:07:56.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So whether this will take us all the way</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:07:59.12\" id=\"00:07:59.12\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(479)>00:07:59.12</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to human level intelligence or something</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:1.76\" id=\"00:08:1.76\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(481)>00:08:1.76</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] or just cat level intelligence is not clear,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:4.84\" id=\"00:08:4.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(484)>00:08:4.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but among all the possible approaches</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:7.32\" id=\"00:08:7.32\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(487)>00:08:7.32</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that people have proposed, I think it's our best shot.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:9.50\" id=\"00:08:9.50\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(489)>00:08:9.50</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So I think this idea of an intelligence system</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:14.64\" id=\"00:08:14.64\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(494)>00:08:14.64</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] filling in the blanks, either predicting the future,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:18.86\" id=\"00:08:18.86\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(498)>00:08:18.86</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] inferring the past, filling in missing information.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:23.76\" id=\"00:08:23.76\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(503)>00:08:23.76</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I'm currently filling the blank of what is behind your head</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:26.66\" id=\"00:08:26.66\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(506)>00:08:26.66</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and what your head looks like from the back</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:30.60\" id=\"00:08:30.60\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(510)>00:08:30.60</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because I have basic knowledge about how humans are made.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:33.76\" id=\"00:08:33.76\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(513)>00:08:33.76</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And I don't know if you're gonna,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:35.52\" id=\"00:08:35.52\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(515)>00:08:35.52</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] what word you're gonna say,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:36.36\" id=\"00:08:36.36\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(516)>00:08:36.36</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] which point you're gonna speak,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:37.28\" id=\"00:08:37.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(517)>00:08:37.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] whether you're gonna move your head this way or that way,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:38.96\" id=\"00:08:38.96\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(518)>00:08:38.96</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] which way you're gonna look,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:40.26\" id=\"00:08:40.26\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(520)>00:08:40.26</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but I know you're not gonna just dematerialize</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:42.08\" id=\"00:08:42.08\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(522)>00:08:42.08</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and reappear three meters down the hall</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:46.26\" id=\"00:08:46.26\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(526)>00:08:46.26</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because I know what's possible and what's impossible</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:50.92\" id=\"00:08:50.92\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(530)>00:08:50.92</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So you have a model of what's possible,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:52.64\" id=\"00:08:52.64\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(532)>00:08:52.64</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] what's impossible and then you'd be very surprised</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:54.52\" id=\"00:08:54.52\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(534)>00:08:54.52</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] if it happens and then you'll have to reconstruct your model.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:57.85\" id=\"00:08:57.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(537)>00:08:57.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Right, so that's the model of the world.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:08:59.61\" id=\"00:08:59.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(539)>00:08:59.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] It's what tells you, you know, what fills in the blanks.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:2.29\" id=\"00:09:2.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(542)>00:09:2.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So given your partial information</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:4.49\" id=\"00:09:4.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(544)>00:09:4.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] about the state of the world, given by your perception,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:8.09\" id=\"00:09:8.09\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(548)>00:09:8.09</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] your model of the world fills in the missing information</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:11.37\" id=\"00:09:11.37\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(551)>00:09:11.37</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and that includes predicting the future,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:13.77\" id=\"00:09:13.77\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(553)>00:09:13.77</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] re-predicting the past, you know,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:16.21\" id=\"00:09:16.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(556)>00:09:16.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] filling in things you don't immediately perceive.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:18.39\" id=\"00:09:18.39\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(558)>00:09:18.39</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] And that doesn't have to be purely generic vision</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:22.27\" id=\"00:09:22.27\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(562)>00:09:22.27</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or visual information or generic language.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:24.35\" id=\"00:09:24.35\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(564)>00:09:24.35</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] You can go to specifics like predicting</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:28.91\" id=\"00:09:28.91\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(568)>00:09:28.91</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] what control decision you make</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:30.29\" id=\"00:09:30.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(570)>00:09:30.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] when you're driving in a lane.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:31.63\" id=\"00:09:31.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(571)>00:09:31.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] You have a sequence of images from a vehicle</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:35.63\" id=\"00:09:35.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(575)>00:09:35.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] and then you have information if you record it on video</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:39.67\" id=\"00:09:39.67\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(579)>00:09:39.67</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] where the car ended up going so you can go back in time</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:43.67\" id=\"00:09:43.67\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(583)>00:09:43.67</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] and predict where the car went</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:45.53\" id=\"00:09:45.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(585)>00:09:45.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] based on the visual information.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:46.69\" id=\"00:09:46.69\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(586)>00:09:46.69</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] That's very specific, domain specific.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:49.43\" id=\"00:09:49.43\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(589)>00:09:49.43</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Right, but the question is whether we can come up</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:51.47\" id=\"00:09:51.47\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(591)>00:09:51.47</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] with sort of a generic method for, you know,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:56.39\" id=\"00:09:56.39\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(596)>00:09:56.39</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] training machines to do this kind of prediction</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:58.53\" id=\"00:09:58.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(598)>00:09:58.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] or filling in the blanks.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:09:59.85\" id=\"00:09:59.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(599)>00:09:59.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So right now, this type of approach</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:3.27\" id=\"00:10:3.27\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(603)>00:10:3.27</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] has been unbelievably successful</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:5.59\" id=\"00:10:5.59\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(605)>00:10:5.59</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] in the context of natural language processing.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:8.19\" id=\"00:10:8.19\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(608)>00:10:8.19</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Every modern natural language processing</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:9.71\" id=\"00:10:9.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(609)>00:10:9.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is pre-trained in self-supervised manner</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:12.30\" id=\"00:10:12.30\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(612)>00:10:12.30</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to fill in the blanks.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:13.74\" id=\"00:10:13.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(613)>00:10:13.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] You show it a sequence of words, you remove 10% of them</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:16.42\" id=\"00:10:16.42\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(616)>00:10:16.42</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and then you train some gigantic neural net</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:17.94\" id=\"00:10:17.94\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(617)>00:10:17.94</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to predict the words that are missing.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:20.34\" id=\"00:10:20.34\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(620)>00:10:20.34</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And once you've pre-trained that network,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:22.74\" id=\"00:10:22.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(622)>00:10:22.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you can use the internal representation learned by it</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:26.62\" id=\"00:10:26.62\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(626)>00:10:26.62</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] as input to, you know, something that you train,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:29.78\" id=\"00:10:29.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(629)>00:10:29.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] supervised or whatever.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:32.22\" id=\"00:10:32.22\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(632)>00:10:32.22</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] That's been incredibly successful.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:33.40\" id=\"00:10:33.40\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(633)>00:10:33.40</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Not so successful in images, although it's making progress</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:37.62\" id=\"00:10:37.62\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(637)>00:10:37.62</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and it's based on sort of manual data augmentation.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:42.46\" id=\"00:10:42.46\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(642)>00:10:42.46</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] We can go into this later,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:43.58\" id=\"00:10:43.58\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(643)>00:10:43.58</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but what has not been successful yet</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:45.70\" id=\"00:10:45.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(645)>00:10:45.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is training from video.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:47.22\" id=\"00:10:47.22\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(647)>00:10:47.22</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So getting a machine to learn,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:48.54\" id=\"00:10:48.54\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(648)>00:10:48.54</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to represent the visual world, for example,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:51.58\" id=\"00:10:51.58\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(651)>00:10:51.58</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] by just watching video.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:52.84\" id=\"00:10:52.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(652)>00:10:52.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Nobody has really succeeded in doing this.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:54.84\" id=\"00:10:54.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(654)>00:10:54.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Okay, well, let's kind of give a high-level overview.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:10:57.60\" id=\"00:10:57.60\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(657)>00:10:57.60</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] What's the difference in kind and in difficulty</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:2.43\" id=\"00:11:2.43\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(662)>00:11:2.43</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] between vision and language?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:4.00\" id=\"00:11:4.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(663)>00:11:4.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So you said people haven't been able</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:6.27\" id=\"00:11:6.27\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(666)>00:11:6.27</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] to really kind of crack the problem of vision open</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:10.51\" id=\"00:11:10.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(670)>00:11:10.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] in terms of self-supervised learning,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:12.00\" id=\"00:11:12.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(671)>00:11:12.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] but that may not be necessarily</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:13.84\" id=\"00:11:13.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(673)>00:11:13.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] because it's fundamentally more difficult.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:15.89\" id=\"00:11:15.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(675)>00:11:15.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Maybe like when we're talking about achieving,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:18.75\" id=\"00:11:18.75\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(678)>00:11:18.75</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] like passing the Turing test in the full spirit</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:22.36\" id=\"00:11:22.36\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(682)>00:11:22.36</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] of the Turing test in language might be harder than vision.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:24.96\" id=\"00:11:24.96\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(684)>00:11:24.96</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] That's not obvious.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:26.43\" id=\"00:11:26.43\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(686)>00:11:26.43</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So in your view, which is harder,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:29.46\" id=\"00:11:29.46\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(689)>00:11:29.46</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] or perhaps are they just the same problem?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:33.22\" id=\"00:11:33.22\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(693)>00:11:33.22</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] The farther we get to solving each,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:34.88\" id=\"00:11:34.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(694)>00:11:34.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] the more we realize it's all the same thing.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:36.74\" id=\"00:11:36.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(696)>00:11:36.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] It's all the same cake.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:40.22\" id=\"00:11:40.22\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(700)>00:11:40.22</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that make them look essentially like the same cake,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:43.62\" id=\"00:11:43.62\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(703)>00:11:43.62</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but currently they're not.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:44.82\" id=\"00:11:44.82\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(704)>00:11:44.82</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And the main issue with learning world models</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:48.52\" id=\"00:11:48.52\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(708)>00:11:48.52</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] or learning predictive models is that the prediction</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:53.18\" id=\"00:11:53.18\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(713)>00:11:53.18</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is never a single thing</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:55.94\" id=\"00:11:55.94\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(715)>00:11:55.94</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because the world is not entirely predictable.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:59.26\" id=\"00:11:59.26\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(719)>00:11:59.26</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] It may be deterministic or stochastic.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:0.74\" id=\"00:12:0.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(720)>00:12:0.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] We can get into the philosophical discussion about it,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:2.98\" id=\"00:12:2.98\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(722)>00:12:2.98</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but even if it's deterministic,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:5.30\" id=\"00:12:5.30\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(725)>00:12:5.30</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] it's not entirely predictable.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:7.46\" id=\"00:12:7.46\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(727)>00:12:7.46</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And so if I play a short video clip</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:11.78\" id=\"00:12:11.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(731)>00:12:11.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and then I ask you to predict what's going to happen next,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:14.18\" id=\"00:12:14.18\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(734)>00:12:14.18</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] there's many, many plausible continuations</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:16.38\" id=\"00:12:16.38\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(736)>00:12:16.38</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] for that video clip.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:18.38\" id=\"00:12:18.38\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(738)>00:12:18.38</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And the number of continuation grows</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:20.58\" id=\"00:12:20.58\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(740)>00:12:20.58</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] with the interval of time that you're asking the system</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:23.94\" id=\"00:12:23.94\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(743)>00:12:23.94</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to make a prediction for.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:26.50\" id=\"00:12:26.50\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(746)>00:12:26.50</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And so one big question with self-supervised learning</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:29.92\" id=\"00:12:29.92\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(749)>00:12:29.92</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is how you represent this uncertainty,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:32.38\" id=\"00:12:32.38\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(752)>00:12:32.38</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] how you represent multiple discrete outcomes,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:35.26\" id=\"00:12:35.26\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(755)>00:12:35.26</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] how you represent a sort of continuum</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:37.14\" id=\"00:12:37.14\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(757)>00:12:37.14</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] of possible outcomes, et cetera.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:40.46\" id=\"00:12:40.46\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(760)>00:12:40.46</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And if you are sort of a classical machine learning person,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:45.26\" id=\"00:12:45.26\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(765)>00:12:45.26</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you say, oh, you just represent a distribution, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:11:59.08\" id=\"00:11:59.08\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(719)>00:11:59.08</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Yeah.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:49.17\" id=\"00:12:49.17\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(769)>00:12:49.17</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And that we know how to do when we're predicting words,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:52.61\" id=\"00:12:52.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(772)>00:12:52.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] missing words in the text,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:53.75\" id=\"00:12:53.75\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(773)>00:12:53.75</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because you can have a neural net give a score</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:56.89\" id=\"00:12:56.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(776)>00:12:56.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] for every word in the dictionary.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:12:58.65\" id=\"00:12:58.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(778)>00:12:58.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] It's a big list of numbers,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:0.85\" id=\"00:13:0.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(780)>00:13:0.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] maybe 100,000 or so.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:2.53\" id=\"00:13:2.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(782)>00:13:2.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And you can turn them into a probability distribution</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:5.29\" id=\"00:13:5.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(785)>00:13:5.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that tells you when I say a sentence,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:9.89\" id=\"00:13:9.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(789)>00:13:9.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] the cat is chasing the blank in the kitchen.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:13.19\" id=\"00:13:13.19\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(793)>00:13:13.19</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] There are only a few words that make sense there.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:15.89\" id=\"00:13:15.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(795)>00:13:15.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] It could be a mouse or it could be a lizard spot</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:18.43\" id=\"00:13:18.43\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(798)>00:13:18.43</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] or something like that, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:21.59\" id=\"00:13:21.59\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(801)>00:13:21.59</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And if I say the blank is chasing the blank in the savanna,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:25.87\" id=\"00:13:25.87\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(805)>00:13:25.87</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you also have a bunch of plausible options</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:27.87\" id=\"00:13:27.87\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(807)>00:13:27.87</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] for those two words, right?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:30.99\" id=\"00:13:30.99\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(810)>00:13:30.99</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Because you have kind of an underlying reality</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:33.67\" id=\"00:13:33.67\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(813)>00:13:33.67</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that you can refer to sort of fill in those blanks.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:38.13\" id=\"00:13:38.13\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(818)>00:13:38.13</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So you cannot say for sure in the savanna,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:42.07\" id=\"00:13:42.07\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(822)>00:13:42.07</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] if it's a lion or a cheetah or whatever,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:44.51\" id=\"00:13:44.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(824)>00:13:44.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you cannot know if it's a zebra or a do or whatever,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:49.59\" id=\"00:13:49.59\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(829)>00:13:49.59</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] wildebeest, the same thing.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:13:56.86\" id=\"00:13:56.86\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(836)>00:13:56.86</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] by just a long list of numbers.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:1.81\" id=\"00:14:1.81\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(841)>00:14:1.81</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and I ask you to predict a video clip,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:4.33\" id=\"00:14:4.33\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(844)>00:14:4.33</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] it's not a discrete set of potential frames.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:7.41\" id=\"00:14:7.41\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(847)>00:14:7.41</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] You have to have somewhere representing</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:10.01\" id=\"00:14:10.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(850)>00:14:10.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] a sort of infinite number of plausible continuations</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:13.57\" id=\"00:14:13.57\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(853)>00:14:13.57</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] of multiple frames in a high dimensional continuous space.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:17.49\" id=\"00:14:17.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(857)>00:14:17.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And we just have no idea how to do this properly.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:20.53\" id=\"00:14:20.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(860)>00:14:20.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Finite high dimensional.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:22.88\" id=\"00:14:22.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(862)>00:14:22.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So like you,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:25.28\" id=\"00:14:25.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(865)>00:14:25.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Just like the words,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:26.24\" id=\"00:14:26.24\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(866)>00:14:26.24</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] they try to get it down to a small finite set</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:32.20\" id=\"00:14:32.20\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(872)>00:14:32.20</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] of like under a million, something like that.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:34.20\" id=\"00:14:34.20\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(874)>00:14:34.20</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Something like that.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:35.04\" id=\"00:14:35.04\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(875)>00:14:35.04</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] I mean, it's kind of ridiculous</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:36.00\" id=\"00:14:36.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(876)>00:14:36.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] that we're doing a distribution</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:39.00\" id=\"00:14:39.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(879)>00:14:39.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] of every single possible word for language and it works.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:42.88\" id=\"00:14:42.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(882)>00:14:42.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] It feels like that's a really dumb way to do it.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:34.22\" id=\"00:14:34.22\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(874)>00:14:34.22</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Something like that.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:43.83\" id=\"00:14:43.83\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(883)>00:14:43.83</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] I mean, it seems to be like there should be</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:49.71\" id=\"00:14:49.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(889)>00:14:49.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] some more compressed representation</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:52.91\" id=\"00:14:52.91\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(892)>00:14:52.91</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] of the distribution of the words.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:54.92\" id=\"00:14:54.92\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(894)>00:14:54.92</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] You're right about that.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:56.12\" id=\"00:14:56.12\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(896)>00:14:56.12</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] And so- I agree.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:57.16\" id=\"00:14:57.16\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(897)>00:14:57.16</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Do you have any interesting ideas</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:14:58.88\" id=\"00:14:58.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(898)>00:14:58.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] about how to represent all of reality in a compressed way</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:1.84\" id=\"00:15:1.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(901)>00:15:1.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] such that you can form a distribution over it?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:3.79\" id=\"00:15:3.79\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(903)>00:15:3.79</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] That's one of the big questions.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:5.03\" id=\"00:15:5.03\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(905)>00:15:5.03</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] How do you do that?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:6.19\" id=\"00:15:6.19\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(906)>00:15:6.19</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I mean, what's kind of,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:7.99\" id=\"00:15:7.99\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(907)>00:15:7.99</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] another thing that really is stupid about,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:12.11\" id=\"00:15:12.11\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(912)>00:15:12.11</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I shouldn't say stupid,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:13.07\" id=\"00:15:13.07\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(913)>00:15:13.07</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but like simplistic about current approaches</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:15.51\" id=\"00:15:15.51\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(915)>00:15:15.51</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] to self-supervised running in NLP in text</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:19.31\" id=\"00:15:19.31\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(919)>00:15:19.31</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] is that not only do you represent</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:21.91\" id=\"00:15:21.91\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(921)>00:15:21.91</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] a giant distribution over words,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:23.79\" id=\"00:15:23.79\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(923)>00:15:23.79</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but for multiple words that are missing,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:25.63\" id=\"00:15:25.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(925)>00:15:25.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] those distributions are essentially independent</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:27.67\" id=\"00:15:27.67\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(927)>00:15:27.67</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] of each other.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:30.15\" id=\"00:15:30.15\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(930)>00:15:30.15</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And you don't pay too much of a price for this.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:33.03\" id=\"00:15:33.03\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(933)>00:15:33.03</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So you can't, so the system,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:36.75\" id=\"00:15:36.75\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(936)>00:15:36.75</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] in the sentence that I gave earlier,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:39.71\" id=\"00:15:39.71\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(939)>00:15:39.71</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] if it gives a certain probability for lion and cheetah,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:43.63\" id=\"00:15:43.63\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(943)>00:15:43.63</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] and then a certain probability for gazelle,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:47.75\" id=\"00:15:47.75\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(947)>00:15:47.75</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] wildebeest and zebra,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:50.27\" id=\"00:15:50.27\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(950)>00:15:50.27</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] those two probabilities are independent of each other.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:55.41\" id=\"00:15:55.41\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(955)>00:15:55.41</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And it's not the case that those things are independent.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:15:57.89\" id=\"00:15:57.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(957)>00:15:57.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Lions actually attack like bigger animals than cheetahs.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:1.29\" id=\"00:16:1.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(961)>00:16:1.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So, you know, there's a huge independent hypothesis</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:4.41\" id=\"00:16:4.41\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(964)>00:16:4.41</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] in this process, which is not actually true.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:7.65\" id=\"00:16:7.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(967)>00:16:7.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] The reason for this is that we don't know</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:9.73\" id=\"00:16:9.73\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(969)>00:16:9.73</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] how to represent properly distributions</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:12.85\" id=\"00:16:12.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(972)>00:16:12.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] over combinatorial sequences of symbols,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:16.09\" id=\"00:16:16.09\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(976)>00:16:16.09</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] essentially when they're,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:17.21\" id=\"00:16:17.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(977)>00:16:17.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because the number grows exponentially</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:18.85\" id=\"00:16:18.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(978)>00:16:18.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] with the length of the symbols.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:21.17\" id=\"00:16:21.17\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(981)>00:16:21.17</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And so we have to use tricks for this,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:22.61\" id=\"00:16:22.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(982)>00:16:22.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] but those techniques can, you know, get around,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:26.25\" id=\"00:16:26.25\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(986)>00:16:26.25</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] like don't even deal with it.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:27.65\" id=\"00:16:27.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(987)>00:16:27.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So the big question is like,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:30.33\" id=\"00:16:30.33\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(990)>00:16:30.33</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] would there be some sort of abstract</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:33.25\" id=\"00:16:33.25\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(993)>00:16:33.25</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] latent representation of text that would say that,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:36.85\" id=\"00:16:36.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(996)>00:16:36.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] you know, when I switch lion for gazelle,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:45.48\" id=\"00:16:45.48\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1005)>00:16:45.48</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Yeah, so this independence assumption,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:48.72\" id=\"00:16:48.72\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1008)>00:16:48.72</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] let me throw some criticism at you that I often hear</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:52.93\" id=\"00:16:52.93\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1012)>00:16:52.93</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So this kind of filling in the blanks is just statistics.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:56.01\" id=\"00:16:56.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1016)>00:16:56.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] You're not learning anything</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:16:58.89\" id=\"00:16:58.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1018)>00:16:58.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] like the deep underlying concepts.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:1.61\" id=\"00:17:1.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1021)>00:17:1.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] You're just mimicking stuff from the past.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:5.65\" id=\"00:17:5.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1025)>00:17:5.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] You're not learning anything new</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:7.53\" id=\"00:17:7.53\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1027)>00:17:7.53</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] such that you can use it to generalize about the world.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:11.97\" id=\"00:17:11.97\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1031)>00:17:11.97</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Or, okay, let me just say the crude version,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:14.09\" id=\"00:17:14.09\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1034)>00:17:14.09</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] which is just statistics.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:16.21\" id=\"00:17:16.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1036)>00:17:16.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] It's not intelligence.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:18.33\" id=\"00:17:18.33\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1038)>00:17:18.33</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] What do you have to say to that?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:19.61\" id=\"00:17:19.61\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1039)>00:17:19.61</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] What do you usually say to that</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:22.62\" id=\"00:17:22.62\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1042)>00:17:22.62</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I don't get into those discussions</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:23.94\" id=\"00:17:23.94\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1043)>00:17:23.94</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] because they are kind of pointless.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:26.74\" id=\"00:17:26.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1046)>00:17:26.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So first of all, it's quite possible</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:28.74\" id=\"00:17:28.74\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1048)>00:17:28.74</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that intelligence is just statistics.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:30.46\" id=\"00:17:30.46\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1050)>00:17:30.46</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] It's just statistics of a particular kind.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:32.70\" id=\"00:17:32.70\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1052)>00:17:32.70</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Yes.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:35.29\" id=\"00:17:35.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1055)>00:17:35.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Is it possible that intelligence is just statistics?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:40.28\" id=\"00:17:40.28\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1060)>00:17:40.28</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Yeah.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:41.57\" id=\"00:17:41.57\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1061)>00:17:41.57</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] But what kind of statistics?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:43.49\" id=\"00:17:43.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1063)>00:17:43.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So if you're asking the question,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:47.17\" id=\"00:17:47.17\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1067)>00:17:47.17</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] are the models of the world that we learn,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:50.65\" id=\"00:17:50.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1070)>00:17:50.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] do they have some notion of causality?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:52.29\" id=\"00:17:52.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1072)>00:17:52.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Yes.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:53.37\" id=\"00:17:53.37\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1073)>00:17:53.37</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So if the criticism comes from people who say,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:56.21\" id=\"00:17:56.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1076)>00:17:56.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] current machine learning system don't care about causality,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:59.21\" id=\"00:17:59.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1079)>00:17:59.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] which by the way is wrong,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:2.13\" id=\"00:18:2.13\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1082)>00:18:2.13</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I agree with them.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:3.37\" id=\"00:18:3.37\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1083)>00:18:3.37</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] You should, your model of the world</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:5.17\" id=\"00:18:5.17\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1085)>00:18:5.17</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] should have your actions as one of the inputs.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:8.89\" id=\"00:18:8.89\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1088)>00:18:8.89</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And that will drive you to learn causal models of the world</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:11.21\" id=\"00:18:11.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1091)>00:18:11.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] where you know what intervention in the world</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:14.85\" id=\"00:18:14.85\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1094)>00:18:14.85</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] will cause what result.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:16.49\" id=\"00:18:16.49\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1096)>00:18:16.49</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Or you can do this by observation of other agents</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:19.21\" id=\"00:18:19.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1099)>00:18:19.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] acting in the world and observing the effect,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:22.29\" id=\"00:18:22.29\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1102)>00:18:22.29</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] other humans, for example.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:24.01\" id=\"00:18:24.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1104)>00:18:24.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So I think at some level of description,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:28.21\" id=\"00:18:28.21\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1108)>00:18:28.21</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] intelligence is just statistics.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:17:52.16\" id=\"00:17:52.16\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1072)>00:17:52.16</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Yes.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:31.44\" id=\"00:18:31.44\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1111)>00:18:31.44</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] But that doesn't mean you don't have models</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:35.00\" id=\"00:18:35.00\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1115)>00:18:35.00</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] that have deep mechanistic explanation for what goes on.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:39.88\" id=\"00:18:39.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1119)>00:18:39.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] The question is how do you learn them?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:41.56\" id=\"00:18:41.56\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1121)>00:18:41.56</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] That's the question.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:42.48\" id=\"00:18:42.48\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1122)>00:18:42.48</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] I'm interested in.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:49.20\" id=\"00:18:49.20\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1129)>00:18:49.20</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] say that those mechanistic model</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:50.84\" id=\"00:18:50.84\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1130)>00:18:50.84</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] have to come from someplace else.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:52.48\" id=\"00:18:52.48\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1132)>00:18:52.48</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] They have to come from human designers,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:53.88\" id=\"00:18:53.88\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1133)>00:18:53.88</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] they have to come from I don't know what.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:56.04\" id=\"00:18:56.04\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1136)>00:18:56.04</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] And obviously we learn them.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:18:59.13\" id=\"00:18:59.13\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1139)>00:18:59.13</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] Or if we don't learn them as an individual,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:1.65\" id=\"00:19:1.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1141)>00:19:1.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] nature learn them for us using evolution.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:4.77\" id=\"00:19:4.77\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1144)>00:19:4.77</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] So regardless of what you think,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:7.01\" id=\"00:19:7.01\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1147)>00:19:7.01</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Yann] those processes have been learned somehow.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:10.10\" id=\"00:19:10.10\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1150)>00:19:10.10</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] So if you look at the human brain,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:12.78\" id=\"00:19:12.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1152)>00:19:12.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] just like when we humans introspect</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:14.50\" id=\"00:19:14.50\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1154)>00:19:14.50</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] about how the brain works,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:16.18\" id=\"00:19:16.18\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1156)>00:19:16.18</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] it seems like when we think about what is intelligence,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:20.10\" id=\"00:19:20.10\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1160)>00:19:20.10</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] we think about the high level stuff</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:22.30\" id=\"00:19:22.30\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1162)>00:19:22.30</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] like the models we've constructed,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:23.78\" id=\"00:19:23.78\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1163)>00:19:23.78</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] concepts like cognitive science,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:25.42\" id=\"00:19:25.42\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1165)>00:19:25.42</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] like concepts of memory and reasoning module,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:28.54\" id=\"00:19:28.54\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1168)>00:19:28.54</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] almost like these high level modules.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:31.50\" id=\"00:19:31.50\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1171)>00:19:31.50</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Does this serve as a good analogy?</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:36.09\" id=\"00:19:36.09\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1176)>00:19:36.09</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Are we ignoring the dark matter,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:40.73\" id=\"00:19:40.73\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1180)>00:19:40.73</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] the basic low level mechanisms,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:43.57\" id=\"00:19:43.57\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1183)>00:19:43.57</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] just like we ignore the way the operating system works,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:45.81\" id=\"00:19:45.81\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1185)>00:19:45.81</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] we're just using the high level software.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:49.65\" id=\"00:19:49.65\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1189)>00:19:49.65</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] We're ignoring that at the low level,</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:52.77\" id=\"00:19:52.77\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1192)>00:19:52.77</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] the neural network might be doing something like statistics.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:57.97\" id=\"00:19:57.97\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1197)>00:19:57.97</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] Sorry to use this word.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t\t\t<div class=\"c\">\n",
            "\t\t\t\t<a class=\"l\" href=\"#00:19:59.05\" id=\"00:19:59.05\">link</a> |\n",
            "\t\t\t\t<div class=\"s\"><a href=\"javascript:void(0);\" onclick=setCurrentTime(1199)>00:19:59.05</a></div>\n",
            "\t\t\t\t<div class=\"t\">[Lex] It probably.</div>\n",
            "\t\t\t</div>\n",
            "\n",
            "\t</body>\n",
            "</html>\n"
          ]
        }
      ]
    }
  ],
  "metadata": {
    "colab": {
      "collapsed_sections": [],
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "546a16efa53e462f88890ebe3977a3bc": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_1f7c6499be4846e1bc674761f3b875f2",
              "IPY_MODEL_268c730678bb4af6baf7bb4f185ddb1e",
              "IPY_MODEL_350678a2f548480b9d6acceda5d78949"
            ],
            "layout": "IPY_MODEL_8f9cafa27efc43c4bb3f566f87df523c"
          }
        },
        "1f7c6499be4846e1bc674761f3b875f2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_e4802c94ccaf404180f3760e3b859a3a",
            "placeholder": "​",
            "style": "IPY_MODEL_71d0b4b532e649829bc813d0f09b2c3b",
            "value": "Downloading: 100%"
          }
        },
        "268c730678bb4af6baf7bb4f185ddb1e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_5fb9f7acb4c7491a986f6d674f467028",
            "max": 598,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_40efa9c5b79e40cab21dcdcf98533caa",
            "value": 598
          }
        },
        "350678a2f548480b9d6acceda5d78949": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_766c1a1b8c1248e79dfa03ec89dbc4dd",
            "placeholder": "​",
            "style": "IPY_MODEL_15e62f4a65104bdea233a96c4dbb6db8",
            "value": " 598/598 [00:00&lt;00:00, 16.1kB/s]"
          }
        },
        "8f9cafa27efc43c4bb3f566f87df523c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "e4802c94ccaf404180f3760e3b859a3a": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "71d0b4b532e649829bc813d0f09b2c3b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "5fb9f7acb4c7491a986f6d674f467028": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "40efa9c5b79e40cab21dcdcf98533caa": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "766c1a1b8c1248e79dfa03ec89dbc4dd": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "15e62f4a65104bdea233a96c4dbb6db8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "327bebaf45eb46b289615c8deb29a69d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_a6919e4b1e2c4510958bf343ef2524a8",
              "IPY_MODEL_309cc6bf3b1e41b1a558fbdb3745c301",
              "IPY_MODEL_ec474373198046d299e09b4823a2766d"
            ],
            "layout": "IPY_MODEL_88186bbd5a1c4425b254d608ca85cd8a"
          }
        },
        "a6919e4b1e2c4510958bf343ef2524a8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_458d4d31cd87402c97c4038553c06574",
            "placeholder": "​",
            "style": "IPY_MODEL_d6d8814501a34ef391c565ba4748e365",
            "value": "Downloading: 100%"
          }
        },
        "309cc6bf3b1e41b1a558fbdb3745c301": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_a9a5de2de75643048ba821d73a00d5c7",
            "max": 17719103,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_51422049fa5846bb9e009b530f0a7526",
            "value": 17719103
          }
        },
        "ec474373198046d299e09b4823a2766d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_ae3785b55e8c4b7e963dc357fb255f05",
            "placeholder": "​",
            "style": "IPY_MODEL_7e8227f2dac34d83a3c670e4d4e0a249",
            "value": " 17.7M/17.7M [00:01&lt;00:00, 33.8MB/s]"
          }
        },
        "88186bbd5a1c4425b254d608ca85cd8a": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "458d4d31cd87402c97c4038553c06574": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "d6d8814501a34ef391c565ba4748e365": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "a9a5de2de75643048ba821d73a00d5c7": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "51422049fa5846bb9e009b530f0a7526": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "ae3785b55e8c4b7e963dc357fb255f05": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "7e8227f2dac34d83a3c670e4d4e0a249": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "fe9e7728c09946c5908d7502c3f77432": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_b67a926cc4904c2f84efd509098e103c",
              "IPY_MODEL_fd98d75dc2424c2faecfb4b5f6a94aa2",
              "IPY_MODEL_187e2d909bfc4a41b2502b964a03fcad"
            ],
            "layout": "IPY_MODEL_bf18adc120174a8bb6cd6f09d853911f"
          }
        },
        "b67a926cc4904c2f84efd509098e103c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_0c4de4008c2e45e5a31d21cdb954f0f2",
            "placeholder": "​",
            "style": "IPY_MODEL_28ef8f3ac2e44445ac9e40d140e513d3",
            "value": "Downloading: 100%"
          }
        },
        "fd98d75dc2424c2faecfb4b5f6a94aa2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_d137a72df0a849a8913fd62924c1848a",
            "max": 318,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_105cf78a882a4cc3bbdaaa2b80217110",
            "value": 318
          }
        },
        "187e2d909bfc4a41b2502b964a03fcad": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_b2791832f51f4794acfdbded875bc9b5",
            "placeholder": "​",
            "style": "IPY_MODEL_28c4d707b0764f51ba2d06eeed5dad96",
            "value": " 318/318 [00:00&lt;00:00, 12.0kB/s]"
          }
        },
        "bf18adc120174a8bb6cd6f09d853911f": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "0c4de4008c2e45e5a31d21cdb954f0f2": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "28ef8f3ac2e44445ac9e40d140e513d3": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "d137a72df0a849a8913fd62924c1848a": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "105cf78a882a4cc3bbdaaa2b80217110": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "b2791832f51f4794acfdbded875bc9b5": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "28c4d707b0764f51ba2d06eeed5dad96": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "0b616fa3a46a4502b14f93270fcb1ca7": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_fd8fd5f045414ab199d3f37c9397ad54",
              "IPY_MODEL_9652595d01cc478c952a368a1f86940d",
              "IPY_MODEL_3a648dda02c9485e8e975e7a1ad2c27e"
            ],
            "layout": "IPY_MODEL_eb2e878c6f0d413b9e82b93b6057e66f"
          }
        },
        "fd8fd5f045414ab199d3f37c9397ad54": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_a6df649cf752475c9540bf1ba2fc7199",
            "placeholder": "​",
            "style": "IPY_MODEL_e491634079f44e65a6edb5d799a2b74e",
            "value": "Downloading: 100%"
          }
        },
        "9652595d01cc478c952a368a1f86940d": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_c9389e064ac54435a3fd356b13b86f4e",
            "max": 1920,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_098d718a941542b99b772068e442e9fa",
            "value": 1920
          }
        },
        "3a648dda02c9485e8e975e7a1ad2c27e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_4a1e581661b84dc6b9c52c1c6efa006b",
            "placeholder": "​",
            "style": "IPY_MODEL_3d1b76d21f6349ccbeefd795974321f7",
            "value": " 1.92k/1.92k [00:00&lt;00:00, 60.7kB/s]"
          }
        },
        "eb2e878c6f0d413b9e82b93b6057e66f": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a6df649cf752475c9540bf1ba2fc7199": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "e491634079f44e65a6edb5d799a2b74e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "c9389e064ac54435a3fd356b13b86f4e": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "098d718a941542b99b772068e442e9fa": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "4a1e581661b84dc6b9c52c1c6efa006b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "3d1b76d21f6349ccbeefd795974321f7": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "bbacac6c5aff4686a31130560e9cf86b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_ade8c72f0eb64de58c4a1351c0b5cdbd",
              "IPY_MODEL_ea54cd59edc64f95958f62a9b6e182c9",
              "IPY_MODEL_5a1b7cd142e94f22a9abc9e64b11d633"
            ],
            "layout": "IPY_MODEL_9403474d6e774f4dbb36c843620ba666"
          }
        },
        "ade8c72f0eb64de58c4a1351c0b5cdbd": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_68eb916bf9494da6a1740e404da1c602",
            "placeholder": "​",
            "style": "IPY_MODEL_1c72713c82fa4ab89cb8fbe84089828f",
            "value": "Downloading: 100%"
          }
        },
        "ea54cd59edc64f95958f62a9b6e182c9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_a730cdaad4e24736b872723240a094dd",
            "max": 83316686,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_a0b10ced5ddc47b5867d864bcc786e0e",
            "value": 83316686
          }
        },
        "5a1b7cd142e94f22a9abc9e64b11d633": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_7c74e70870304e83826f2a322341a000",
            "placeholder": "​",
            "style": "IPY_MODEL_5a98ade616cd441da36ed51b85f4b61c",
            "value": " 83.3M/83.3M [00:01&lt;00:00, 54.3MB/s]"
          }
        },
        "9403474d6e774f4dbb36c843620ba666": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "68eb916bf9494da6a1740e404da1c602": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "1c72713c82fa4ab89cb8fbe84089828f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "a730cdaad4e24736b872723240a094dd": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a0b10ced5ddc47b5867d864bcc786e0e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "7c74e70870304e83826f2a322341a000": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "5a98ade616cd441da36ed51b85f4b61c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "36ecee3e086a4a34bd6f6fe435b20a6e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_f638d0f51ac548e8a36a0b8858082e29",
              "IPY_MODEL_41359dbbceea458b9137b8cb8fb2e7e1",
              "IPY_MODEL_a9b8f75e82274a4e8cfa1026fe8408de"
            ],
            "layout": "IPY_MODEL_b71b33109b124e30b0004634467e64dc"
          }
        },
        "f638d0f51ac548e8a36a0b8858082e29": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_32792ea893a140dab321565df126e461",
            "placeholder": "​",
            "style": "IPY_MODEL_e9b04004efb944c6ad7d53763d426626",
            "value": "Downloading: 100%"
          }
        },
        "41359dbbceea458b9137b8cb8fb2e7e1": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_5d5fa4a478844c798c91a274361cda2b",
            "max": 1921,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_323bd202a5a540b4b8805b88138468ad",
            "value": 1921
          }
        },
        "a9b8f75e82274a4e8cfa1026fe8408de": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_0831e94760fa421384e2b33c97532d09",
            "placeholder": "​",
            "style": "IPY_MODEL_b7073c903e1e4701b68516dc8a867fc1",
            "value": " 1.92k/1.92k [00:00&lt;00:00, 63.8kB/s]"
          }
        },
        "b71b33109b124e30b0004634467e64dc": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "32792ea893a140dab321565df126e461": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "e9b04004efb944c6ad7d53763d426626": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "5d5fa4a478844c798c91a274361cda2b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "323bd202a5a540b4b8805b88138468ad": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "0831e94760fa421384e2b33c97532d09": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "b7073c903e1e4701b68516dc8a867fc1": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "28e3b80e86634a32aaef5b67345c217c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_3a1c92eec125401ca3184d713b27f4a5",
              "IPY_MODEL_500472e00812437390d5caf9bc0ee67b",
              "IPY_MODEL_41aa0f5eaddd4907802992b380e85ded"
            ],
            "layout": "IPY_MODEL_030169a7a2594b5a8646f327a0566dfe"
          }
        },
        "3a1c92eec125401ca3184d713b27f4a5": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_55d89d284453466c8013f2a884d55426",
            "placeholder": "​",
            "style": "IPY_MODEL_47506b0858534c6ebc53e8e500ef56b6",
            "value": "Downloading: 100%"
          }
        },
        "500472e00812437390d5caf9bc0ee67b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_feaf363bb8c548fcae1cd13fd1074581",
            "max": 5534328,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_fc80e8542e34422b86874eee0b20b3f8",
            "value": 5534328
          }
        },
        "41aa0f5eaddd4907802992b380e85ded": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_b8cdfacd7bee4c398fc513ed2a3311e3",
            "placeholder": "​",
            "style": "IPY_MODEL_85c433e956794588b166c9ed6fa68c20",
            "value": " 5.53M/5.53M [00:00&lt;00:00, 14.8MB/s]"
          }
        },
        "030169a7a2594b5a8646f327a0566dfe": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "55d89d284453466c8013f2a884d55426": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "47506b0858534c6ebc53e8e500ef56b6": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "feaf363bb8c548fcae1cd13fd1074581": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "fc80e8542e34422b86874eee0b20b3f8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "b8cdfacd7bee4c398fc513ed2a3311e3": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "85c433e956794588b166c9ed6fa68c20": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "36c91974947c4bf7aef8585de061a5a7": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_a7d1f042f6124c51aa69ac6985e7c83c",
              "IPY_MODEL_3fbcb6d197ab4c44930a2e15e439459b",
              "IPY_MODEL_6b7194d233cd47428f58bf987f9712fe"
            ],
            "layout": "IPY_MODEL_135ef8a71e4041ad80129cc5dbc51354"
          }
        },
        "a7d1f042f6124c51aa69ac6985e7c83c": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_5d46ef19e68c4974a5001a48121047f3",
            "placeholder": "​",
            "style": "IPY_MODEL_65b31b2e94b04dd39298f004f9493e29",
            "value": "Downloading: 100%"
          }
        },
        "3fbcb6d197ab4c44930a2e15e439459b": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_e73e393e01f4487d86dfbe5e401b4c09",
            "max": 128619,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_a93270536a094ed694b825ea38aa36a2",
            "value": 128619
          }
        },
        "6b7194d233cd47428f58bf987f9712fe": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_dbbb60965fb24cd9b2dc1ea510e7e480",
            "placeholder": "​",
            "style": "IPY_MODEL_d06a71fe937b4ee1a44764f020bf3d57",
            "value": " 129k/129k [00:00&lt;00:00, 234kB/s]"
          }
        },
        "135ef8a71e4041ad80129cc5dbc51354": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "5d46ef19e68c4974a5001a48121047f3": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "65b31b2e94b04dd39298f004f9493e29": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "e73e393e01f4487d86dfbe5e401b4c09": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "a93270536a094ed694b825ea38aa36a2": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "dbbb60965fb24cd9b2dc1ea510e7e480": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "d06a71fe937b4ee1a44764f020bf3d57": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        }
      }
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}