Why it matters: Yet again, artificial intelligence has shown its superiority over us mere mortals when it comes to humanity’s favorite games. On this occasion, an AI has beaten some of the world’s top poker players across 12-days of six-person no-limits Texas hold’em.

We’ve already seen artificial intelligence get the better of humans in Go, Chess, and computer games such as DOTA 2 and Starcraft II. Back in 2017, Libratus, an artificial intelligence program developed at Carnegie Mellon University, beat four professional poker players in 1-on-1 games.

Libratus’ successor, Pluribus, went one better. Designed by researchers from Facebook’s AI lab and Carnegie Mellon, it played over 10,000 hands of poker against five pros—all of whom have won at least $1 million during their poker careers—winning a virtual $48,000. Pluribus also played games against one human and four versions of itself, but the bots could not communicate with each other, preventing any collusion. The program won an average of $5 per hand, taking about 20 seconds to play each one, with hourly winnings of around $1,000.

“It’s the first time AI has achieved superhuman performance in a multiplayer game,” said Tuomas Sandholm, who developed Pluribus alongside Noam Brown.

To deal with the complexity of playing multiple opponents, Pluribus was designed to predict just two or three moves ahead, rather than trying to work out how they would play all the way to the end of the round. It’s said to be an excellent bluffer, and its unpredictability makes the AI difficult to play against.

As with other AIs, Pluribus mastered poker by playing copies of itself, learning the game and slowly improving over eight days and hundreds of thousands of hands. The program was created using a 64-core server packed with less than 512GB of RAM, but it runs on two Haswell processors and uses 128GB during games.

Beyond playing multiple opponents at poker, Pluribus has applications in several areas, including investment banking, self-driving cars, negotiating, and wargames—the US Army Research Office contributed to the project’s funding.