
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
Counterfactual Regret Minimization Lets AI Succeed in Imperfect Information Games Like Poker
There is a special search algorithm that deals with the fact that you have converted the problem into this other equivalent problem in a particular way. So you're exploiting some knowledge of the game and the conversion process a little bit still. So it doesn't necessarily work for it. It works in principle for all games of this type, but in practice, it would not necessarily run./nThe search algorithm that we end up using is the same search algorithm that we actually used in all the poker bots before. And so it's actually this algorithm called counterfactual regret minimization, which is a very general kind of algorithm. And it's also been very successful in other imperfect information games as well.