
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
Intrinsic Motivation and Reinforcement Learning: Minimizing Surprise
Intrinsic motivation and reinforcement learning can be approached from the perspective of minimizing surprise rather than seeking novelty./nThis approach is inspired by the idea that in an ecological view of intelligence, minimizing surprise can lead to finding a safe and comfortable niche./nMinimizing surprise may require taking coordinated action, such as going on an adventure to prepare for potential surprises in the future./nMinimizing surprise can lead to behavior that resembles curiosity or novelty seeking./nThis approach can be beneficial for situating agents in open world settings where exploration is desired without human supervision, while avoiding distractions.
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.