Episode 04: Joel Lehman, OpenAI, on evolution, open-endedness, and reinforcement learning
Generally Intelligent
00:00
The Value of Play in Deep Learning
I love the idea of that corn slite flipping fromavers to co operative, and then thinking about what that generates. I think it also leads just to the value of contraryan thinking in general. In a field like deep learning, how do you make thing? How do you do impactful work? It's really difficult. You can make models that predict how well other models will do whand they train. Or you can imagine training and over fitting to just a really small number of data exampor hire’s pobably a really large space of possibility, sir. But maybe there isn't. Andso i want ed to get your thoughts on how do you actually explore
Transcript
Play full episode
