From philosophy to POMDPs to reinforcement learning & hierarchical planning to the publishing model in academia, I really enjoyed this conversation with Leslie.
Sorry Lex, I admit to having misspelled your name when telling people about your channel -- but I won't do it again ;-) Thanks for your efforts and interviews!
Being someone whose work environment is not alongside other programmers where I have to either use lay terminology or keep to myself with descriptions of what I do, it's intriguing and almost refreshing to hear conversations on this podcast constantly referring to everyday life in terms of functions, problem spaces, sets, and the like. really enjoying it.
Begins with a GEB mention, eloquently articulates issues I have been trying to categorise for years, is pro open access, isn't getting distracted by ill-formed big ideas... anything this person has ever said or written is now priority reading for me. Genius.
Leslie is one those amazing guests who do not proffer opinions on supposedly deep questions, but are very lucid in what they do say. I really liked her!
3rd yr undergrad student. Learning ai on his own. I could understand the problem statements for deep learning and computer vision(couldn't solve them , obviously) , but not so much in reinforcement learning. This really helped. Reinforcement learning isn't just a computation problem. It's a merging of the disciplines mentioned under symbolic systems . One can specialize in one or more of thosw fields, but to make things work, all those perspectives are needed. Thanks for this.
I think the output we want from a "perception" system is a cost or a credit of the current state of the world. Just like we do in inverse reinforcement learning or target propagation for the credit assignement problem. Awesome video btw !
Any person who doesn't value competition based on 'personal feelings' loses a lot of credibility. Its better than resenting it for ideological reasons, but its our most general method, and possibly most fundamental method of finding truth. Dismiss it at your peril.