On AI, Elite Poker and Beyond
Last week, the report “Superhuman AI for multiplayer poker”, was published by Noam Brown and Tuomas Sandholm. Commenting on their project in a Facebook Artificial Intelligence blog post, Brown said,
“In recent years, new AI methods have been able to beat top humans in poker if there is only one opponent. But developing an AI system capable of defeating elite players in full-scale poker with multiple opponents at the table was widely recognized as the key remaining milestone.
“Pluribus, a new AI bot we developed in collaboration withCarnegie Mellon University, has overcome this challenge and defeated elite human professional players in the most popular and widely played poker format in the world: six-player no-limit Texas Hold'em poker.”
Comments from professional poker players involved in the projectinclude:
Chris Ferguson (six-time World Series of Poker champion): “Pluribus is a very hard opponent to play against. It’s really hard to pin him down on any kind of hand. He’s also very good at making thin value bets on the river. He’s very good at extracting value out of his good hands.”
Jason Les: “It is an absolute monster bluffer. I would say it’s a much more efficient bluffer than most humans. And that’s what makes it so difficult to play against. You're always in a situation with a ton of pressure that the AI is putting on you and you know it’s very likely it could be bluffing here.”
Jimmy Chou: "As humans I think we tend to oversimplify the game for ourselves, making strategies easier to adopt and remember. The bot doesn't take any of these shortcuts and has an immensely complicated/balanced game tree for every decision.”
Michael Gagliano: “It was incredibly fascinating getting to play against the poker bot and seeing some of the strategies it chose.There were several plays that humans simply are not making at all, especially relating to its bet sizing.”
Trevor Savage: “I thought the bot played a very solid, fundamentally sound game. It did a very good job of putting me to tough decisions when I didn’t have a strong hand and getting value when it had the best hand.“
Note: Pluribus was trained in eight days using a 64-core server with less than 512 GB of RAM. The processing cost was less than $150(based on cloud computing rates.)
The significance of Pluribus is that it moves AI's elitecompetitiveness beyond structured games such as chessand Go, into the more dynamic game play of poker, which incorporates strategies such as bluffing.
Given Pluribus’ adaptive nature and low processing costs, its approach will likely reshape how many systems designers pursue application development in areas such autonomous vehicles, cybersecurity, business analytics and more.
As the man/machine gap narrows in game play, addressing the complexities of the real world will require the developmenteven more innovative AI solutions.