By Deborah Jones
VANCOUVER, Canada, 2007
Humans prevailed after two of the world’s top poker players pitted their wiles and math ability against a computer, in a grueling two-day match scientists called the world’s first man-machine poker championship.
A computer program called Polaris played four games of Texas Hold Em against Phil Laak and Ali Eslami, two gamblers from Los Angeles who are among the top players in the global high-stakes cash poker circuit.
Rows of weary-looking computer scientists, and a few spectators, watched the grueling poker battle in an overheated hotel conference room. When the humans won, the room erupted in cheers.
The humans received some $40 000 for each match they won against the machine – mere peanuts in their world. The cash is modest, noted Eslami, who said he agreed to come to this western Canadian city to compete because he’s interested in artificial intelligence.
“I’m interested in being at the cusp of this wave, it’s like history unfolding. It’s like watching the first launch of the space shuttle … This is the beginning of the next revolution in computers, computers that can become human.
The man-machine poker game was a highlight at the annual global conference of the Association for the Advancement of Artificial Intelligence, attended by more than 1,000 scientists from universities and corporations worldwide.
“The poker is in some ways an entertaining event, but it’s also a very scientific event,” said Michael Bowling, leader of the computer science team that developed Polaris at the University of Alberta in Edmonton.
“The technology is more than about poker – poker is a test for evaluating,” said Bowling, 31. “Building a machine with human intelligence creates an intellectual challenge.”
Winning at poker – with its inherent bluffing, emotions, deliberate deception and elements of chance as well as mathematics – is a milestone for Artificial Intelligence.
Scientists have now developed computers that can beat humans at chess, checkers and backgammon. But if Polaris wins the essentially psychological game of poker – with its inherent bluffing, emotions, deliberate deception and elements of chance as well as mathematics – Bowling said it will be a major milestone for the progress of Artificial Intelligence.
He compared the poker competition to the 1997 chess match between Garry Kimovich Kasparov and an IBM computer Deep Blue. Deep Blue won that match its conquest of the legendary Russian world chess champion made global headlines – and changed how humans regard computers.
Eslami and Laak are technologically savvy as well as professional gamblers. Laak is an engineer by training, and Eslami is a business graduate who worked as a gaming and computer consultant before turning to professional poker.
Eslami, 30, said the point of the Polaris-human event is more than a poker match – it helps further AI technology that humans can use in many ways.
“If they can solve the problem the applications are huge for human kind,” he said. “If computers can understand human emotion they can interact with us more.”
Eslami said possible long-term examples of AI use might be for smart automobile computers that can sense the emotions of drivers and take measures to avoid dangerous situations, a computerised telephone system that can pass a frustrated caller directly to a human operator, and military strategy in which computers would help understand the emotions of opponents.
“But these are large abstractions from what they’re doing here,” said Eslami.
Darse Billings, the lead architect of the Polaris team, said applications of a computer capable of playing poker are in the future. “We don’t really think in terms of applications, we’re doing pure research,” said Billings, 45, a one-time chess master who after completing a masters degree in computer science spent several years as a professional poker player.
It was Billings, who later obtained a computer science PhD from the University of Alberta, who suggested using poker to research artificial intelligence. “It’s a game of hidden information … The challenge is trying to get inside (the opponents) head,” he said.
“Philosophically, there’s nothing that a human can do that a computer can’t do,” said Billings, 45. “The whole world is rife with uncertainty, and we want computers to bridge that gap.”
After the match Eslami said playing against the computer was more exhausting than any previous game in his career. “I’m surprised we won…. it’s already so good it will be tough to beat in future” as scientists make further improvements on Polaris’ programming.
Billings said even though the program lost, it played “brilliantly.” Polaris showed scientists that it is possible for a computer to do well at the essentially psychological game.
“I was expecting a draw,” said computer scientist Michael Littman of Rutgers University in the United States, who served as the official arbiter of the game.
Copyright © 2007 Deborah Jones
Adapted from stories originally published by Agence France-Presse in July, 2007
References and further reading:
University of Alberta Artificial Intelligence
Association for the Advancement of Artificial Intelligence