1994: The computer program Chinook takes the title of Man-Machine World Champion in checkers.
1997: IBM's Deep Blue defeats chess master Gary Kasparov.
2011: IBM's Watson takes first place in the game show Jeopardy.
Computers are increasingly beating humans that their own games. That's because computers can store and calculate almost every aspect of the game they're playing and come up with a winning strategy faster than their human opponents. But there are still some games where humans have the edge, like poker. Or do they?
Computers are good at games where the rules are known, but the common wisdom was that they couldn't match humans in games that include luck or bluffing. It turns out that the common wisdom was wrong, at least in certain cases.
Let me introduce you to Cepheus, the poker-playing artificial intelligence.
What makes Cepheus so special is that its programmers managed to "solve" the game of poker. In other words, it knows how to play so it never loses a game.
Of course, that doesn't mean it automatically wins. A skilled human opponent can force a draw, but more often than not the human makes a mistake. It's like a game of Tic-Tac-Toe where one player always makes the right moves.
Cepheus deals with bluffing by calculating the odds of which cards its opponent has. And it knocks out luck by playing more hands. In a long enough game, an unlucky hand here and there gets averaged out.
Now, to be fair to humans, Cepheus only plays Limit Texas Hold'em, which is one of the least intensive versions of poker. And it only plays games with two players.
Even then, to keep track of the game requires 10 terabytes of information. That's 1,000 times more information than a complete online encyclopedia.
In other words, it'll be a while before a computer can match tournament-level players in a No-Limits Texas Hold'em tournament. However, before Cepheus people, thought it was unlikely to happen; now it looks a lot more possible.