
Episode 28: Sergey Levine, UC Berkeley, on the bottlenecks to generalization in reinforcement learning, why simulation is doomed to succeed, and how to pick good research problems
Generally Intelligent
Predictive Modeling
The challenge is that actually prediction is often much harder than generation. It's like once you have to do prediction, you need consistency. And there are like all of these other things to work on. I do think we still see a lot of model based RL that does these kind of frame by frame rollout versus predicting point in the future or something like that. But just because it's clean doesn't mean it's right.
00:00
Transcript
Play full episode
Remember Everything You Learn from Podcasts
Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.