Generally Intelligent cover image

Episode 21: Chelsea Finn, Stanford, on the biggest bottlenecks in robotics and reinforcement learning

Generally Intelligent

CHAPTER

Are There Alternatives to Reinforcement Learning?

Some people say we can't use reinforcement learning forever to get to much more generally capable agents or generally, capable robots. I like predictive world models a lot just because they're interpretable. You can look at the predictions of one of these models and see what it thinks is going to happen. If it just completely blurs out the object, you know it's probably not going to work. So we don't necessarily need to predict everything in the scene and doing so is nearly impossible at long time horizon.

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.
App store bannerPlay store banner