Poker pros will play heads-up NL Hold’em against a new artificial intelligence system developed by the Carnegie Mellon University to find out if the odds favour machine or human. The game will begin at Rivers Casino in Pittsburgh on January 11 under the banner “Brains vs. Artificial Intelligence: Upping the Ante.”
Four poker pros—Dong Kim, Jason Les, Jimmy Chou, and Daniel McAulay—will play 120,000 hands of heads up NL Hold’em over 20 days for their share of $200,000.
Libratus, the computer they will play against, was created after a similar contest of last year when the previous computer developed by The Carnegie Mellon University, called Claudico, got a chip stack much smaller than three of its human opponents. The previous contest was played over 80,000 hands, which turned out to be too few to determine who the winner was—machine or human.
This time, the professional poker players will take part in duplicate matches in pairs. The computer and the pair of players will receive the same cards in isolated rooms as well as casino floors. Statistical significance can be achieved this time because the contest will be played over several days over a two-table format, in which two hands will be simultaneously played. Every day, the game will start at 1:00 a.m. and will conclude at 7:00 p.m. in the poker room of Rivers Casino. The public will be allowed to view the match.
Tuomas Sandholm, the computer science professor at Carnegie Mellon University, said: “Since the earliest days of AI research, beating top human players has been a powerful measure of progress in the field.” The professor said that humanity achieved this goal with chess in the year 1997 and with Jeopardy in the year 2009 and with Go, a board game, the previous year. According to Sandholm, poker is more challenging because it needs a machine capable of making complex decisions on the basis of incomplete information while simultaneously dealing with slow play, bluffs, and other strategies.
Sandholm developed the bot with Noam Brown, a research student. Called Libratus, it was created with an algorithm capable of computing strategies for games involving incomplete information and of using the Bridge Supercomputer from the Pittsburgh Supercomputing Center.
Libratus is a Latin word, which means powerful and balanced. It is called so not just because it incorporates new technology, but also in honor of John Forbes Nash Jr., an alumnus of Carnegie Mellon.