2 years ago
“Poker has been one of the hardest games for AI to crack,” declares Andrew Ng, chief scientist at Baidu. “There is no single optimal move, but instead an AI player has to randomize its actions so as to make opponents uncertain when it is bluffing.” Yet that is just what Libratus (created by Professor Tuomas Sandholm and graduate student Noam Brown from the computer science department at Carnegie Mellon University) have done. It more than held its own against human poker pros at the Brains vs. Artificial Intelligence Showdown at the Rivers Casino in Pittsburgh. In fact, it bested four professional poker players during the 20-day tournament.
Will Knight has delved into the significance of these advances for MIT Technology Review. Knight writes:
“Last year, Sandholm led the development of a previous poker-playing program, called Claudico, which was soundly beaten in a match against several professional poker players. He explains that Libratus uses several new advances to achieve such a high level of play. This includes a new equilibrium approximation technique, Sandholm says, as well as several new methods for analyzing possible outcomes as cards are revealed at later stages of a game.”
The result: a program capable of recognizing a great collection of cards, but also bluffing its way to victory.
To read more, click here. Below, watch poker pro Jason Les seem a little freaked out as he explains just how effectively AI can play a hand.