Artificial Intelligence

A.I. Poker Program Faces Off Against Seasoned Pros in Epic Contest

A three-week battle that starts today pits man against machine in a $200,000 tournament of Heads-Up No-Limit Texas Hold 'Em.

<p>Carnegie Mellon University</p>

Theorists and scholars can tell you that poker is considered one of the greatest games ever invented, right up there with chess and backgammon. Poker's peculiar combination of skill, luck and psychology makes it a very tough game system for machines to crack.

So it will be interesting to see what goes down this week in Pittsburgh when four of the world's top poker pros square off against an advanced artificial intelligence built from the ground up to win at poker.

The 20-day tournament, dubbed Brains Vs. Artificial Intelligence: Upping the Ante, kicks off today at the Rivers Casino in Pittsburgh, Pa. It's a rematch of sorts, following the inaugural 2015 contest, and will once again pit man against machine in the extreme poker variant knows as Heads-Up No-Limit Texas Hold 'Em.

Let's break that down a bit: Texas Hold 'Em has been the most popular form of poker, by far, since the televised poker craze of a few years back. "No-Limit" indicates that players can make any size bet at any time. And "Heads-Up" means there are only two players at the table - virtual, in this case - going head-to-head for all the chips.

No-Limit Heads-Up poker is a radically different game than standard multiplayer poker, and it puts a high premium on psychological aspects like bluffing and aggressive wagering. That's what makes it so interesting, from a machine intelligence standpoint.

RELATED: Killer Machines and Sex Robots: Unraveling the Ethics of A.I.

The A.I. debuting this week in Pittsburgh is called Libratus, and it's an upgraded version of the system used in the 2015 event, named Claudico. Two years ago, Claudico didn't fare so well. Matched against four poker pros over two weeks and 80,000 hands, it collected fewer chips than all but one of its human opponents.

Developed by Tuomas Sandholm and Carnegie Mellon University's School of Computer Science, Libratus uses machine learning algorithms to build on previous systems used to play games such as chess and, just last year, the ancient Chinese game of Go. Indeed, the CMU labs helped develop two earlier celebrity A.I. systems - Deep Blue and Watson.

Professor Tuomas Sandholm of CMU's School of Computer Science | Carnegie Mellon University

"Poker poses a far more difficult challenge than these games," said Sandholm in press materials announcing the event. "It requires a machine to make extremely complicated decisions based on incomplete information while contending with bluffs, slow play and other ploys."

RELATED: The Freaky Way Robots Are Becoming More Human

The Libratus A.I. is expected to provide a much tougher challenge for human players this time around. The machine learning system is the result of more than 15 million core hours of computation in simulated games, compared with the 2-3 million core hours used for Claudico.

According to the design team, the A.I. is taking a different mathematical approach this time around, too. The system is optimized to attain what game theorists call a Nash equilibrium, named for the late CMU alumnus and Nobel laureate John Forbes Nash Jr. ("A Beautiful Mind"). It gets complicated, but basically a Nash equilibrium is a pair of strategies where neither player can benefit from changing as long as the other player's strategy remains the same.

Besides the upgraded A.I., this year's tournament will introduce some other changes. The four poker pros – Jason Les, Dong Kim, Daniel McAulay and Jimmy Chou – will be vying for shares of a $200,000 real money cash prize purse.

RELATED: College eSports Will Be Bigger Than March Madness

They'll be earning it. This year's tilt is scheduled for 120,000 individual hands dealt over the course of 20 days – a 50 percent bump over last year's 80,000 hands over two weeks. Statistically speaking, this will reduce the influence of blind luck in the competition as the players compete from 11 a.m. to 7 p.m. daily.

Each human competitor, playing via laptop computer, will also be managing two hands simultaneously. Part of each eight-hour day will be spent on the casino floor, and part in an isolated chamber. The tournament designers have also introduced specific mathematical control scenarios to ensure fair competition.

If you want to keep track of the action as it goes down, Rivers Casino has posted an event page with live leaderboards and webcam feeds, or keep an eye on the #BRAINSVSAI Twitter hashtag.

WATCH VIDEO: Face Up Poker