Four of the top poker players in the world have been sitting in a Pittsburgh casino every day for the last 20 days between 11 a.m. and about 10 p.m. playing against a robot called Libratus. Libratus has won more than $1.5 million worth of chips from the humans with only a few hours of the Brains vs. AI competition left, notes The Guardian.
AI robot was seen as an underdog in poker
It would take a miracle for the human players, Jason Les, Daniel McCauley, Jimmy Chou, and Dong Kim, all specialists in the two-player unlimited bid form of poker (dubbed “Texas Hold’em”), to make a comeback. It may be a defeat for humanity but a huge milestone for artificial intelligence. The players are not playing with actual money but instead for a lump sum prize of $200,000.
Machines have become smart enough to defeat humans at other games like chess and Go, but poker is more difficult because it is a game with imperfect information, notes The Guardian. A player can see the complete board with chess and Go, but with poker, players do not get to see each other’s hands. Moreover, in order to win, the AI must correctly interpret misleading information and bluffs.
Dov Gertzulin's DG Capital is having a strong year. According to a copy of the hedge fund's letter to investors of its DG Value Partners Class C strategy, the fund is up 36.4% of the year to the end of June, after a performance of 12.8% in the second quarter. The Class C strategy is Read More
Carnegie Mellon University Professor of Computer Science Tuomas Sandholm built Libratus with his Ph.D. student Noam Brown. Sandholm said this challenge was so big and complicated that it has been elusive to AI researchers until now. Sandholm added that he was not at all confident that Libratus would defeat the poker pros, as international betting sites put the robot as the 4-1 underdog.
Taking the revenge for Claudico
Brown said that they did not tell Libratus how to play poker; they just gave it the rules of poker and said “learn on your own.” The robot began playing randomly, and over the course of trillions of hands, it was able to refine its approach and arrive at a winning strategy, notes The Guardian.
The competitors were no match for Libratus, which improved on the previous poker-playing AI called Claudico. In the same tournament in 2015, Claudico lost against four poker pros.
Jason Les, who played against Claudico two years ago, said, “Libratus turned out to be way better than we imagined. It’s slightly demoralizing. If you play a human and lose, you can stop, take a break. Here we have to show up to take a beating every day for 11 hours a day. It’s a real different emotional experience when you’re not used to losing that often.”