Does a computer bluff well? Can you read a computer’s reaction to his/her/its cards?
These are a few of the questions that will be facing two poker professionals on Monday in Vancouver, British Columbia as they face a poker-playing computer in a showdown for $50,000.
Professional players Phil Laak, who recently won NBC’s Poker After Dark, and Ali Eslami, an elite ranked poker player each have the opportunity to win $25,000 if they win all four games against the super computer during the Association for the Advancement of Artificial Intelligence conference being held at Vancouver’s Hyatt Regency Hotel.
The computer, named “Polaris” by its creators from the University of Alberta, can simulate actual poker players, allowing it to aggressively bluff or remain cautious and under-bet. Jonathan Schaeffer, one of the lead researchers on the Polaris project, says that the goal of the computer is to be “unpredictable” for human bettors.
"With poker, the key thing is unpredictability," said Jonathan Schaeffer, one of the lead artificial intelligence researchers on the Polaris project. "A computer is very good at predicting odds . . . it can do that very quickly."
Although past projects have proven successful against human opponents in games such as checkers and chess, researchers wonder whether the sheer luck involved in poker may be the determining factor.
"I'm not at all sure that the computer is going to win this thing, precisely because it is so difficult to learn about the style of your opponent," said Oliver Schulte, an SFU cognitive scientist who once worked with the Polaris creators.
