
Episode 22: Archit Sharma, Stanford, on unsupervised and autonomous reinforcement learning
Generally Intelligent
00:00
Is There a Good Time to Do Something?
If something you have to do it at the difficulty, we should probably get really good at it. But if something you expect to like only encounter once in your life, you're probably not careful about doing it. I think it's a great heuristic actually. Kind of like maybe what you want is some kind of all like switching between these different algorithms as you make predictions about how often you're going to encounter this thing. Yeah. That's what I'm going to do. With an opal guy, Jeff, Jeff,Jeff. Like, yeah. Or opt. or run or something. It's interesting because you actually expect that a generally intelligent agent would have like somewhat each of
Transcript
Play full episode