WASHINGTON—It may surprise you to learn that it's more difficult for a computer to play poker, perhaps the world's most popular card game, than chess, the pastime of deep thinkers.
Unlike chess, poker deals with tricky matters such as uncertainty, probability, guesswork and deception—human wiles that a chess-playing robot, such as the one that beat world champion Garry Kasparov in 1997, doesn't need to consider.
As a result, computer experts say, poker is more like real life—with all its subtleties and complications—than chess is, with its fixed rules and vast but finite possibilities.
"Chess might be a better test of raw computer power," said Christian Lebiere, a psychology professor at Carnegie Mellon University in Pittsburgh. But computer poker programs "are indeed more like human problem-solving."
In real life, "we spend a whole lot more time managing uncertainty than doing deep sequential problem-solving" such as chess, said Michael Littman, a computer scientist at Rutgers University in New Brunswick, N.J.
Five programs, with names such as Poki and Poker Probot, will test their problem-solving skills next week in a computer poker competition sponsored by the American Association for Artificial Intelligence in Boston.
"Much as computer chess was an early test of artificial intelligence, computer poker has emerged as an even greater AI challenge," Lebiere said.
During the contest, the rival computers will play thousands of rounds of two-person Texas Hold'Em, the current rage in poker.
Even with only two "players"—two computer-software programs—there are a quintillion (1 with 18 zeros after it) possible combinations of cards and bets for the computers to deal with in each game.
"Poker is a very complex game," said Tuomas Sandholm, a Carnegie Mellon computer scientist. Players have to deal with the luck of the draw as the cards are randomly dealt and must make decisions based on incomplete information.
"You don't know what the other guy is holding," said Sandholm, the inventor of GS2, one of the computer poker systems competing in Boston. "A lot of real-world situations have uncertainty in them, and you have to deal with the uncertainty."
Other problems a poker robot must solve include determining the probability that the next card dealt will be a good one, deciding how much to bet, figuring out what an opponent's bet means and handling bluffing.
"An opponent can purposely bet to mislead its opponent," noted Littman, who'll referee the competition. "The best programs factor this kind of bluffing into their assessments."
No matter which program wins the Boston contest, computers are still no match for a top-level human at the poker table.
Phil Laak, a champion player, crushed Poker Probot at the World Series of Poker last year. During the match, the audience chanted, "Humans! Humans!"
According to Lebiere, poker robots are now about as skillful as computer chess programs were in the 1950s and `60s, when humans could beat them easily.
"At this point humans are still by far the best, but I have little doubt that machines will catch up at some point, perhaps relatively soon," he said.
———
TEXAS HOLD'EM
Here's how the popular poker game Texas Hold'Em works. It has four colorfully named phases:
Pre-flop: Each player is dealt two cards face down. A round of bets, calls and raises follows.
Flop: Three cards are dealt face up so all players can see them. Then comes a second round of bets.
Turn: A fourth card is dealt face up. There's a third betting round.
River: A fifth card is dealt face up. The final round of bets, calls and raises takes place.
Players then show their hands. The winner is the one who can make the strongest five-card hand out of the seven cards dealt.
For more information online, go to http://www.cs.ualberta.ca/(TILDE)games/poker/
———
(c) 2006, McClatchy-Tribune Information Services.
Need to map
Comments