Facebook Announces AI Bot That Can Beat Professional Poker Players

Facebook, the American online social media and social networking service company based in Menlo Park, announced that an AI-powered bot created by researchers at Carnegie Mellon University(CMU) and at Facebook, can beat human poker professionals in a six-player game.

A game of Poker.

A game of Poker.

Across 2 weeks and several thousand games, the AI system named Pluribus played with 12 poker professionals. Pluribus won an average of $5 per hand with hourly winnings of around $1,000 which is a decisive margin of victory. The other unique thing to note was the cost of designing and running, Pluribus cost a meager $150 compared to the millions spent by Google’s DeepMind, which have relied on supercomputers consisting of more the 5000 processors. Facebook trained the AI-infused algorithm to play hands against itself trillions of times in a process known as reinforced learning. This allows Pluribus to improve upon the blueprint strategy by searching for a better strategy in real-time for the situations it finds itself in during the game.

Pluribus achieved superhuman performance at multiplayer poker, which is a recognized milestone in artificial intelligence and in-game theory that has been open for decades,
— Tuomas Sandholm, Professor at CMU

The professor added: “Thus far, superhuman AI milestones in strategic reasoning have been limited to the two-party competition. The ability to beat five other players in such a complicated game opens up new opportunities to use AI to solve a wide variety of real-world problems.”

Pluribus was created in just 8 days using a 64 core server infused with less than 512GB of RAM and no GPU, making it a solid bargain compared to the millions of dollars spent by other companies. Pluribus was developed by Noam Brown of Facebook AI Research and Tuomas Sandholm of Carnegie Mellon University. Pluribus’ unique algorithms are now being analyzed for other real-world applications. Many application involves Texas Hold ‘em poker — which includes several players, hidden knowledge and several scenarios to consider. However, all is not well for the future as these kinds of AI-infused bots could destroy the high-stakes online poker as we know it. Research suggests that many people will play online poker when they know that this type of software is out there and people could use it for financial gain.

Indraneel Dulange