
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
How to Play Perfect Information Games Like Backgammon
The bot that played six player poker only cost $150 to train, like an equivalent of $150. And again, it was a very search heavy approach, it used 28 CPUs at inference time. I think that shows that this would have been possible 20 years ago if people knew that this was the approach to take. It's kind of satisfying to have the accomplishment be an algorithmic improvement instead of just like applying large amounts of competition.
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.