
Episode 27: Noam Brown, FAIR, on achieving human-level performance in poker and Diplomacy, and the power of spending compute at inference time
Generally Intelligent
Is There a Nash Equilibrium?
The idea that basically bot equilibrium weight, either like multiple equilibrium and this equilibrium weight end up in some not very useful space where the robot is speaking gibberish to each other. That's one of the issues. So just because you found one equilibrium, it doesn't really help you unless the other players are playing the same equilibrium. But there's another problem which is that the humans might not be playing an equilibrium at all. We see this in games like diplomacy. Humans play very suboptimally. And you can see this also for things like imagine a self-driving car. This is a good example because you could train a self- driving car from scratch just by simulating it with
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.