
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
The Key Insights for Doing Well on No Press Diplomacy
A lot of the ideas that we applied to poker, we carried over to diplomacy. If you train a bot completely from scratch with no human data and no press diplomacy, it will do really, really well if there is one bot and one human. But if you stick it in a game with six humans and just the bot, it will get crushed. Because the humans are kind of like speaking this non-verbal language. They have this understanding of what each other is going to do,. Even if it's non-verbal, that the bot just doesn't abide by.
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.