Learning Unsupervised Learning

At Google you were pivoting to RL research and reading Sutton and all of this stuff. Learning RL will happen then. So yeah, as I was exploring the topic, I ended up working on a sub area within RL called unsupervised RL. And there I had this idea where we wanted to like think about what kind of behaviors like agents learn. A lot of that is from a desire to like control your environment or audio. This is related to the environment hypothesis. It ended up seeping into my work at the time and you published this paper called that's which was about like learning behaviors, whichever votes predictable and diverse. You're one of the more robust sort of unsupervised