
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
Is There a Future in Robotic Learning?
In the long run, I think it'll be much easier to use data rather than simulation to enable robots to do things. In NLP or computer vision, nobody in NLP thinks about coding up a simulator to simulate how people produce language. So get humans to perform some tasks, maybe a robotic task or maybe another booking flights from the Internet and then basically do a really good job of emulating behavior. That would be very good to leverage if you're starting a company right now. But if you want highly effective AI systems, I think we need to go beyond that. It harkens back to an earlier quote by Rodney Brooks said imitation is doomed to succeed. He said something like
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.