AI can beat top poker professionals in a six-player game. Supercomputers and artificial intelligence (AI) programs have helped crack codes, identify a lost ancestor of humans, and beat chess grandmasters.
Games have often served as a way to test, challenge, or measure progress for artificial intelligence, but most of the games used in research involve two players.
In a historic first, researchers from Carnegie Mellon University have developed an AI capable of beating six players in a game of Texas hold’em.
A study detailing the new milestone was published in the journal Science.
Texas hold’em is one of the most popular poker games played today, but with so many players and so many different potential actions in the game, the researchers had their work cut out for them.
To help reduce the number of different decisions the AI could make, the researchers created a program called Pluribus that first learned the ins and outs of the game by playing against copies of itself.
This self-play strategy made up the blueprint or core of the program, and Pluribus learned to excel at the game without any prior data input from humans.
The researchers used an approach called information abstraction and action abstraction to help reduce the number of actions Pluribus could take, and after mastering the game against itself, Pluribus was pitted against five top professional poker players.
In 10,000 hands of poker, Pluribus outperformed both its copies and the human players, which is a considerable milestone in AI technology.
“Pluribus’s success shows that despite the lack of known strong theoretical guarantees on performance in multiplayer games, there are large-scale, complex multiplayer imperfect-information settings in which a carefully constructed self-play-with-search algorithm can produce superhuman strategies,” the researchers conclude in their study.
Image Credit: Shutterstock/Syda Productions