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
00:00
How Did You Progress From No Press to Full Press?
Bots may naturally need to be robust to much more adversarial situations because humans will like aggressively probe them. That's one of the lessons that I took away from the poker work. By having a controllable dialogue model, we can focus a lot of the computational effort at inference time on coming up with good plans and then conditioning the dialogue generation on those plans.
Transcript
Play full episode
