AI poker-master program developed that can never make a mistake or lose to a human

Scientists from the University of Alberta have developed a computer AI capable of utterly destroying all human competitors in the game of Texas Hold'em. They claim the AI, obnoxiously named “Cepheus,” is so good, that one can spend his/her entire lifetime playing against and still never win. They've essentially developed an algorithm that “can't be beaten with statistical significance within a lifetime of human poker playing,” according to the paper published in Science. 

For those brave(sadistic) souls willing to test their mettle, which is pleasant way of saying “get beat,” the algorithm can be tested here . The website allows users to not only play against Cepheus, but to query the bot and learn its strategy.

The team has been developing the AI poker-ace for the greater part of the last decade, followed by a two months training period. And by “training,” I don't mean coding every single possible combination manually, but teaching the Cepheus the optimal strategy by having it play against itself. 

Using the computing power of 4,000 CPUs, Cepheus considered and played 24 trillion hands of poker per second for the entire two month duration. “That's more poker hands than all of humanity – so in some sense it's not surprising that it has developed the perfect strategy,” exclaimed lead researcher Michael Bowling. Each victory and loss was an opportunity to carefully fine tune its strategy, eventually learning to play, what its developers claim is, “an essentially perfect game.”

Cepheus is not the first artificial intelligence created solely to beat humans at their games. In 1997, IBM's Deep Blue supercomputer defeated Garry Kasparov, the world chess master at the time. Although chess is considered what game theorists classify as a “perfect-information game,” a game in which every player is aware of everything that as previously occurred in the game before making a decision, poker is excluded from this category as players do not know which cards have been dealt to one another. Cepheus has overcome this limitation because it was designed to learn by “regretting” and remembering every decision that's lead to an optimal outcome. 

While Cepheus' concept  may seem more “fun” than practical, the ability to learn and deduce could potentially find into airport security.

Source BBC via Science