Prescription Drug Caused Compulsive Gambling
Cards and casino chips are displayed during the Global Gaming Expo Asia at the Venetian Macao-Resort-Hotel in Macau REUTERS

Meet Cepheus, the virtually unbeatable poker playing machine. Researchers at the University of Alberta have created a near-invincible poker-playing computer program they call Cepheus. This program is so good at poker that the scientists say, even if a person played 60 million hands of poker for 70 years, 12 hours a day, the person would still not be statistically as good at poker as Cepheus is.

Like any player, Cepheus learns through experience, except that its only opponent is itself. At first it plays a random game, assigning equal probability to each action, then it starts learning from its mistakes and starts questioning its own moves. It asks itself whether it could have done better at this or that instance. Scientists can measure how much Cepheus regrets the actions it didn't take, and the computer program uses this regret to play better, similar to human players.

Cepheus is the first player to finally solve Two-player limit Texas hold'em poker, and the scientists have created a Web site where people can test themselves against Cepheus. Even after a lifetime of playing poker against the computer program, though, a human would not know that Cepheus isn't a perfect player, because it is very near-perfect.

Cepheus has been created for fun, but the scientists who made it are aware that the machine could be used in real world situations that have opponents with their own motives, and where one party doesn't know what the other party will do, such as in real world negotiation, cybersecurity and medical applications. In medicine, the disease is the opponent and a game similar to poker is played.

This is a landmark development but there is a harder problem - no-limit hold'em poker where the sizes of bets and raises are variable, and it seems that it might be impossible for a machine to perfect this poker game.

To contact the writer, email: sonali.raj@gmail.com