It is not an exaggeration to say that poker is half a psychological game, so it is a different game from Go or chess. I think the ReBeL released by Facebook this time is remarkable in this regard. In particular, it is characterized by the use of reinforcement learning and search together, and it seems to show well what areas Facebook is focusing on recently, such as attempts to combine a conversation model with a search engine like RAG.
Interestingly, in order to realize the “bluffing” in poker, they set up “belief” for each player and used it, which seems to be meaningful in many ways when applying AI to the game. Here's a link to the original article on Facebook ReBel's github repository and AITimes.
facebookresearch/rebel
An algorithm that generalizes the paradigm of self-play reinforcement learning and search to imperfect-information games. – Facebookresearch/rebel
