
Episode 16: Yilun Du, MIT, on energy-based models, implicit functions, and modularity
Generally Intelligent
00:00
How Have Your Interests Evolved Over Time?
In the beginning of your career, you were very optimistic about intelligence. You think that simple techniques can be used to train a large amount of skill. But now he says there are some crucial inducted biaces or structures we need to build in for better performance. He thinks convolutiona narrow networks have a very general idea but don't do well with visual tasks.
Transcript
Play full episode