A few days ago, I noted on this blog that a poker match was taking place between two of the world's finest human players and Polaris, the most advanced poker-playing computer program yet devised. The results are now in, and I'm delighted to say that—albeit by a slim margin—humanity remains ahead of silicon in matters of bluffing. Out of four sessions, one was a draw, one was a machine win, and two were human wins. Many anthropocentric congratulations to Phil Laak and Ali Eslami.
Perhaps most intriguingly, Polaris will soon become available for anyone to play against on Poker Academy Online - a more than formidable challenge for even seasoned pros. But what are the long-term ramifications of computers developing a talent for a game often thought to be beyond artificial mastery due to its layered uncertainties? Plenty of people make good money through online poker (although plenty more lose it) and, while the secrets of Polaris are not yet in the public domain, its very existence and the solid body of research behind it suggest that extremely able laptop-based bluffing engines can't be far away.
Imagine. The idea of playing online chess or checkers for money is ludicrous in part because anyone could cheat by consulting a computer program for advice. As AI advances in leaps and bounds, might the massively lucrative online poker industry be crippled by the next generation of the technologies that brought it into existence?
Polaris out-bluffed
July 26, 2007
