
Episode 18: Oleh Rybkin, UPenn, on exploration and planning with world models
Generally Intelligent
00:00
Did a Lot of Things Work?
When i started my pege g i waso kind of, i worked on things that i didn't even know whether they're important or not. Towards the end of my peg g, i am trying to kind of focus myself on something that i personally think is important. I tried to use empowerment, which sometimes its called variational intrinsic control. And it's just a clever way of learning skills that's not go condition, but learns a sort of a later representation of skills. But right now it doesn't really feel like its working yet. The standard thing in reinforcement learning thog is when you have rewards, still should be extremely useful. So i've been trying for some time to
Transcript
Play full episode