Facebook develops poker-playing AI
Researchers from Facebook say they have developed an algorithm that can learn to play poker better than a human.
US.- Facebook researchers have announced the development of a general AI framework called Recursive Belief-based Learning (ReBeL) that they say achieves better-than-human performance in heads-up, no-limit Texas hold’em poker while using less domain knowledge than any prior poker AI.
As part of the trials, researchers said ReBel played against Dong Kim, ranked as one of the best heads-up poker players in the world. ReBeL played faster than two seconds per hand across 7,500 hands and never needed more than five seconds for a decision. It scored 165 (with a standard deviation of 69) thousandths of a big blind per game against humans it played compared with Facebook’s previous poker-playing system, Libratus, which reached 147 thousandths.
According to developers, ReBeL is a step towards developing universal techniques for multi-agent interactions such as auctions, negotiations, cybersecurity and self driving cars.
ReBeL builds on work in which the notion of “game state” is expanded to include agents’ beliefs about what state they might be in based on common knowledge and the policies of other agents.
Researchers say ReBeL trains two AI models — a value network and a policy network — through self-play reinforcement learning, resulting in a simple, flexible algorithm. Researchers claim it is capable of defeating top human players at large-scale, two-player imperfect-information games.
