Rohin Shah
TalkRL: The Reinforcement Learning Podcast
00:00
The Next Paper on the Utility of Learning About Humans for Human Ai Coordination
Ithik: I think i don't plan to do further direct research on this myself. The point of the paper is that when you're building your ai systems, they should be reasoning more. And so i will continue to push for that point, including like a projects at deep mine. But there's still plenty of work than one could do - such as trying to come up with algerdams to directly optimize the maths we wrote down. Ah, but that seems less high leverage to me.
Transcript
Play full episode
