20-Day marathon will showcase over 100,000 hands of Texas hold’em
An artificial intelligence (AI) computer program developed by Carnegie Mellon University (CMU) will play poker against four of the world’s top poker players in a humans vs. machine rematch that begins next week.
Four poker professionals – Jason Les, Dong Kim, Daniel McAulay and Jimmy Chou – will play a collective 120,000 hands of heads-up, no-limit Texas hold’em poker against the CMU computer program named Libratus, which is powered by the Pittsburgh Supercomputing Center’s (PSC) Bridges supercomputer, which runs on Intel Xeon processors.
The competition will be held at Rivers Casino in Pittsburgh over a 20-day period beginning Jan. 11.
“Since the earliest days of AI research, beating top human players has been a powerful measure of progress in the field,” said CMU computer science professor Tuomas Sandholm in a recent CMU press release. “That was achieved with chess in 1997, with Jeopardy! in 2009 and with the board game Go just last year. Poker poses a far more difficult challenge than these games as it requires a machine to make extremely complicated decisions based on incomplete information, while contending with bluffs, slow play and other ploys.”
In 2015, four top poker professionals played 80,000 hands of poker against a CMU computer program called Claudico, which was powered by PSC’s Intel Xeon-based Blacklight supercomputer. They totaled $170 million in bets, and collectively, the human poker players beat the AI program by $732,713.
But CMU scientists said the results were not statistically significant, which means they believe the competition ended in a statistical tie.
For the rematch, the humans will play a collective 120,000 hands of poker against the new Libratus computer program in hopes of better determining whether humans or a computer can establish poker superiority, the press release said.
“I’m very excited to see what this latest AI is like,” said Les, a poker player based in Costa Mesa, Calif. “I thought Claudico was tough to play. Knowing the resources and the ideas that Dr. Sandholm and his team have had available in the 20 months since the first contest, I assume this AI will be even more challenging.”
Sandholm and his team built the Libratus AI program from scratch. They’ve used about 15 million core hours of computation on the Bridges supercomputer to build Libratus, compared with 2 to 3 million core hours used to build Claudico, the release said.
During the first competition, the poker players noticed Claudico was making some obvious bluffs that they exploited. To improve strategic play during the humans vs. machine rematch, Libratus will use the Bridges supercomputer to do live computations for endgame strategies.
“We’re pushing on the supercomputer like crazy,” Sandholm said.