Today and tomorrow (July 23 and 24), the world's First Man-Machine Poker Championship is pitting two poker masters against a computer program, called Polaris, at the Association for the Advancement of Artificial Intelligence (AAAI) conference (Vancouver B.C.). Polaris was programmed by the University of Alberta and is competing against Phil "The Unabomber" Laak, a mechanical engineer and winner of the World Poker Tour, and Ali Eslami, a gaming consultant turned professional poker player.
If the humans can beat Polaris, they stand to win a $50,000 pot, but it won't be a cake walk, noted Professor Jonathan Schaeffer, the leader of the Polaris programming team. According to Schaeffer, Polaris encapsulates not only the kind of "perfect knowledge" that IBM's Deep Blue used in 1997 to beat Garry Kasparov at chess, but also includes more advanced artificial intelligence that enables Polaris to make guesses and bluff.
"Chess is a game of perfect knowledge--there is nothing hidden from the players, but in poker you can't see your opponent's hand," said Schaeffer. "Poker is a much harder challenge from an artificial intelligence perspective."
Poker is more challenging, from the computer programmers point of view, because it also contains the element of chance--especially the type of poker being played at the First Man-Machine Poker Championship: a two-player version of Limit Texas Hold'em, with Polaris playing simultaneously against both Laak and Eslami in separate rooms (two copies of Polaris are being used with no communications allowed between them, or between the humans).
The match is being supervised by the Poker Academy and can be followed live on the Web. In all, the two teams will play 2000 hands of poker over the two days, with the cards being "dealt" electronically from randomly generated card decks. Four duplicate sessions of 500 hands each will played, with $5,000 awarded to the humans if they win by more than 25 bets; $2,500 for each session that is won or lost by fewer than 25 bets.