
Episode 22: Archit Sharma, Stanford, on unsupervised and autonomous reinforcement learning
Generally Intelligent
Human-Eyed Robots?
I used to think like human-eyed robots are so annoying because they have to stand up. Actually, it's really hard. So what if you made tour robots? But then the demonstration data for this is very complicated because now you have like a lot of human nicks. Also, does it sit in your kitchen? I don't know. It's just a lot of problems. One problem is like, okay, we need to like learn in a continual setting,. There are like human interventions and part of your level. What should our learning objective look like? And something impossible thing is like we should try maximizing information. Where do their reward comes from? Is another like super, super
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.