As part of the Brains vs Artificial Intelligence competition at a Pittsburgh Casino, a software robot called Libratus has won 1.5 million dollars worth of chips from four of the world’s top poker players. With the four players, Dong Kim, Jason Les, Jimmy Chou and Daniel McCauley, nearing the end of the 20 day competition, it appeared highly unlikely the human players will be able to turn it around and win. The event marks an important milestone for artificial intelligence.
While AI has already started to become able to beat humans at games such as chess and Go, a win at poker is even more significant since the player is provided with imperfect information, unable to see other player’s hands. A win also requires the AI to bluff properly and interpret potentially misleading information.
The Libratus software was created by Carnegie Mellon University professor of computer science, Tuomas Sandholm, with the help of his PhD student, Noam Brown.
Sandholm said of poker, that “this challenge is so huge and complicated that it’s been elusive to AI researchers until now,” saying he had not been confident the software would beat human players. “The international betting sites put us as 4-1 underdog and the humans expected to win,” he said.
Sandholm and Brown had previously created AI poker software called Claudico, which lost against four top human players during the same competition in 2015. The new software boasted more computing power as well as an enhanced algorithm for handling situations with hidden information.
“We didn’t tell Libratus how to play poker. We gave it the rules of poker and said ‘learn on your own’,” according to Brown. He also said watching the software repeatedly beat top human players gave him a “proud parent feeling.”
“When I see the bot bluff the humans, I’m like, ‘I didn’t tell it to do that. I had no idea it was even capable of doing that.’ It’s satisfying to know I created something that can do that,” said Brown.
Libratus’s algorithms are not specific to just poker, with possible applications in other situations with imperfect information, ranging from business negotiations to military or security strategy, or even the planning of medical treatment.
“People have this idea that poker is a very human game and that bots can’t bluff, for example. That’s totally wrong. It’s not about reading your opponent and trying to tell if they are lying, it’s about the cards and probabilities,” said Brown.
“We are seeing a re-evaluation of the types of things machines can excel at – although I can’t see a computer writing a prize-winning novel any time soon.”