CS Doctoral student tackles multiplayer poker
The average person makes hundreds of decisions every day. Sometimes the choices we make are logical while others are intuitive. Intrigued by human intellect, artificial intelligence researchers have been striving towards creating machines that can also reason about a domain and make independent decisions.
Because of the complexity of the human mind, research in this field is split into a number of branches. One of these branches involves deduction and reasoning – for example, the logical way we would solve a puzzle or game.
Many computer programs have already been developed which can successfully play games at expert levels. Chinook (which plays checkers), for instance, will never lose – it is able to play to a draw if it does not win. Though impressive, games like checkers and chess are deterministic (all future game states are known to both players).
Poker, on the other hand, is more closely related to real world scenarios in which there is hidden information along with the element of chance. The U of A Computer Poker Research Group’s current star is Polaris, a program that defeated some of the best heads-up (two player) limit (fixed bet size) poker players in the world in the Second Man-Machine Poker Competition.
Richard’s research is focused on improving their program’s strategies for multiplayer (starting with 3-player) Texas Hold’em.
After finishing a math degree at Simon Fraser University, his interest in both artificial intelligence and games drew Richard to computing at the U of A.
"While studying math was rewarding in its own ways, being able to put my work to the test against human players and other programs is something I really enjoy about poker research," says Richard. "Maybe I am just too competitive, but facing off against the world’s best players and programs is very exciting. I feel lucky to be able to incorporate these competitions into my academic research."
Article and photos, 2010.