Cepheus: One after another, the "castles" of human intelligence are falling, and in the future, casinos would do well to ban access not only to minors, but also to robot players. The best poker player, who never makes mistakes, is a new software program created by Canadian scientists and is based on a sophisticated artificial intelligence algorithm.
The researchers, led by Professor Michael Bowling of the University of Alberta, who published the relevant paper in the journal "Science", according to the BBC, Reuters and "Nature", stated that the poker algorithm follows such a perfect game strategy on the electronic "felt" that it cannot be defeated, even if a human plays his entire life against it (this is poker played with only two people). 
No matter what card the human player has, no matter what strategy he follows, in the end the computer will win, which, among other things, makes quite a few bluffs!
The new AI achievement joins a string of similar successes that have garnered international attention, starting in 1997, when IBM's "Deep Blue" supercomputer first defeated world chess champion Garry Kasparov. More recently, in 2001, IBM's Watson supercomputer beat its human opponents in the quiz game Jeopardy.
Games like chess are considered easier for an AI machine to play because all players have perfect information about what is happening in the game before making a decision. However, in the case of games like poker, the same is not true because players do not know what cards others are holding.
A new poker program called “Cepheus,” based on game theory, can overcome even this problem of uncertainty. Having played countless poker games on its own, it is designed to learn from experience and remember any decisions that do not lead to the best possible outcome.
"Our model spent two months playing poker over and over again. It was throwing 24 trillion poker cards every second for two months, more than all of humanity has ever thrown. So in a sense, it's not surprising that it has now developed the perfect strategy," said Michael Bowling, who in 2007 had done something similar with checkers, creating the ideal software program for this board game.
The researchers believe that algorithms similar to "Kifeas", which solve the problem of uncertainty and incomplete information, could in the future be used in various other cases, such as airport security checks, medicine, business negotiations, investment decisions, etc.
Anyone who wants to can try their luck (in vain) against the algorithm by playing a game of electronic poker at the Canadian university's address poker.srv.ualberta.ca.
