Facebook Unveils Poker Ace AI

Facebook Unveils Poker Ace AI
Pluribus needed just 20 hours to learn to play Poker on a level with the world's best human players.

(CCM) — Scientists at Facebook have developed a "superhuman" artificial intelligence (AI) system that can beat the world's best poker players at their own game, the BBC is reporting.

The AI, known as Pluribus, was programmed with the rules of poker, and then made to play trillions of games against itself to teach itself how to play to win. Pluribus needed just 20 hours to learn to play to the level of a poker professional, according to the report.

A number of top human poker players were invited to play against Pluribus and were impressed with its skills. "Pluribus is a very hard opponent to play against," Chris Ferguson, a World Series of Poker champion, said in the report. "It's really hard to pin him down on any kind of hand," he added.

One thing that is unusual about Pluribus is that it needs just $150-worth of cloud computing resources to run, according to the report. That contrasts with AI systems developed by Google's Deepmind research initiative which rely on supercomputers with more than 5,000 specialist processors which reportedly cost millions of dollars.

Facebook says that it has no plans to make the Pluribus program code public in case it has a negative impact on the poker community, although examples of techniques used will be provided to other AI researchers.

Image: © Rokas Tenys - Shutterstock.com