| layout |
|---|
default |
I had first contact with AI/ML in my Master, in The Netherlands, which I did under the supervision of my now good friend Marco Wiering. I am grateful to him for sharing his vast knowledge and enthusiasm as well as his positive vibes. See my Master thesis here.
In my PhD, my previous supervisor was Marc Deisenroth. Under him, I have worked on variational inference, which I found very interesting and useful. I learned a lot with Marc, he helped me a great deal, but I felt this was not the way for me to continue.
After that I studied information geometry, which I found extremely interesting, but that turned out to be a much harder and longer task than I initially had believed/hoped. However I did learn a lot in the process.
I finally settled on deep reinforcement learning as the main topic for my thesis, but soon I realized that many of the papers that came out were really just incremental results and very few actually brought real contributions to the field. I think this is generally valid for any rapidly developing field of science, as they are most affected by the illness/frenzy of publishing as many papers as possible. In case you don't know this is a real problem. Luckily, I was not forced to publish (however I had to present some internal documents at Imperial, that would ideally turn into publications). In this process I also kind of developed a resentment towards publishing and felt that I wanted to publish important stuff, not just anything for the sake of publishing.
Among others, this conclusion also lead to the following project, a data-driven survey, which I hope (smart and practical) people will soon realise is the only way to wrap heads around what's going on in some particular field.
So then I decided I am going to do something which I truly believe in. Work to discover traits of general AI, or strong AI. So I started reading a lot of papers and books from neuroscience, psychology, cognitive science to learn about memory, decision making, consciousness, etc. I soon discovered a simple principle that can be applied hierarchically and to all types of memory, be it semantic, episodic, temporal, etc. I call it the neighbouring principle, you can read more about it here. I also wrote a document describing how I see things and what I plan to do, you can check it out [here]. Soon enough some interesting papers appeared, showing me that I am on the right path: deep mind on thinking, big-loop recurrence, nature of thinking.
I also strongly believe that ML should be available to all not only the few, so I am working on a project which exposes ML to the any person, not only computer scientists, similar to Orange.
I enjoy reading, some books I read and am reading now:
| Year | Author | Title |
|---|---|---|
| 2014 | Andrea | Won Outstanding Lead Actor in a miniseries or a movie |
| 2015 | BAFTA | Nominated for Best Leading Actor for Sherlock |
| 2014 | Satellite | Won Best Actor miniseries or television film |