In the ultimate poker match between man and machine, man won-at least for this round. In a two-week competition that just ended, four top ten players of Heads-up No-limit Texas Hold'em took on Claudico, an artificial intelligence program created by a Carnegie Mellon University team and won more chips than they lost.
However, although three of the pros had higher winnings than Claudico, the collective $732,713 human lead was not statistically significant. In short, these results aren't scientifically reliable, and it would be imprudent to extrapolate from them. The technical human victory is a statistical tie.
Naturally, the Claudico team is pleased with their program's progress.
"We knew Claudico was the strongest computer poker program in the world, but we had no idea before this competition how it would fare against four Top 10 poker players," said Claudico development director Tuomas Sandholm, a professor of computer science at CMU. "It would have been no shame for Claudico to lose to a set of such talented pros, so even pulling off a statistical tie with them is a tremendous achievement."
Sandholm explains that poker has become a major test of artificial intelligence because it is based on incomplete information. Players actively try to mislead each other and don't know what cards their opponents hold.
"Beating humans isn't really our goal; it's just a milestone along the way," Sandholm says. "What we want to do is create an artificial intelligence that can help humans negotiate or make decisions in situations where they can't know all of the facts."
Claudico played 20,000 two-player hands against each Bjorn Li, Doug Polk, and Dong Kim, all top-rated pros. The wagering was virtual; the pros will receive performance-based appearance fees from a donated prize purse of $100,000, funded in part by Microsoft.
"We know theoretically that artificial intelligence is going to overtake us one day," Li says, unbothered by the statistical significance issue. "But at the end of the day, the most important thing is that the humans remain on top for now."
Claudico uses algorithms; the equivalent of human poker skill is not part of the programming. The algorithms are based only on the rules of poker. This kind of algorithm has a wealth of potential applications; these can be used to create business, cybersecurity, and healthcare strategies.
Claudico's ancestor, Tartanian7, achieved a statistically significant win in the Heads-Up No-limit Texas Hold'em category against each human opponent last July. The pros who played Claudico were able to observe how the older version played.
Polk feels that Claudico is a good player, but not as good as a pro.
"There are spots where it plays well and others where I just don't understand it. Betting $19,000 to win a $700 pot just isn't something that a person would do," he observed. The CMU computer science team expressed that Claudico's moves sometimes confused them, too. The tournament provided far more data than they can possibly analyze.
Caludico's big advantage? He never gets rattled.
That, and he has a team working to improve his playing; Sandholm believes they will eventually be able to beat human players, noting that they now have 80,000 additional hands' worth of data for training purposes.
Sam Ganzfried, one of Claudico's designers, said: "I think we can have a new program that can beat the humans in a year."