NPR logo

In Man Vs. Machine Poker Match, Machine Wins

  • Download
  • <iframe src="https://www.npr.org/player/embed/92577686/92577666" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
In Man Vs. Machine Poker Match, Machine Wins

Technology

In Man Vs. Machine Poker Match, Machine Wins

In Man Vs. Machine Poker Match, Machine Wins

  • Download
  • <iframe src="https://www.npr.org/player/embed/92577686/92577666" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

Poker champion Ali Eslami defeated Polaris 1 at poker in Vancouver, Canada, last year. This year, however, Polaris 2 defeated its human challengers. Don MacKinnon/AFP/Getty Images hide caption

toggle caption Don MacKinnon/AFP/Getty Images

A computer defeated six top poker players in Texas Hold 'Em at a convention in Las Vegas.

Though computers have beaten top chess champions — IBM's Deep Blue famously defeated Gary Kasparov more than a decade ago — it's actually harder to program a computer to win at poker. Unlike chess or checkers, where all the information is out to see, poker is a game of bluffing and of reading subtle personality clues.

"The big thing about poker is the uncertainty involved — that you're uncertain about what cards your opponent has, you're uncertain about what future cards are going to come up, that you're uncertain about what style of play your opponent is using — so handling that uncertainty is what makes the AI [artificial intelligence] challenging," says Michael Bowling, an associate professor at the University of Alberta, who helped develop the computer, known as Polaris 2.

Bowling, who heads the university's Computer Poker Research Group, says one of the improvements over Polaris 1 was a more sophisticated learning component. Polaris 2 was designed to predict what human players would do based on their past strategies. The system is also based on game theory, which doesn't try to read opponents but instead figures out the outcomes of different choices.

The players were excited to play against the computer to see how far they could get and what they could learn. Bowling says many of them want a rematch.

We no longer support commenting on NPR.org stories, but you can find us every day on Facebook, Twitter, email, and many other platforms. Learn more or contact us.