Norway-Flag-iconUnited-Kingdom-flag-iconsweden-flag-icon

Two weeks ago, a match between Libratus, a poker playing computer, and several humane opponents kicked off in Pennsylvania. Going by the recent results of a similar match against Claudico, most people expected to see a clear-cut victory by human players. However, with almost two-thirds of the match now behind us, this doesn’t seem to be the case.

Libratus

Strong Display by Libratus

The 2015 match against Claudico saw four players take on the computer over the course of 80,000 hands. When the last card was dealt, humans walked away with a victory, although scientists claimed that it was a tie – statistically speaking. While nuances could be argued, it was clear Claudico wasn’t ready to stand up to human players.

With Libratus, things have been different from the get go. The artificial intelligence has gotten off to a nice start, taking a $50,000 lead after 27,000 hands. Over the course of next 40,000 hands, this advantage increased significantly, to slightly above $700k mark. Using the poker lingo, humans seem to be getting completely crushed.

Tough Challenge for Humans

Going into the match, Jason Les, Jimmy Chou, Dong Kim, and Daniel McAulay probably didn’t expect things to be this tough. It is clear now that Tuomas Sandholm, the man behind the Libratus project, has carefully considered the information gathered from the previous match and introduced some serious improvements.

Humans are facing a challenging task, with some 50,000 hands to go. They are stuck more than $700,000 combined, and it seems like they don’t have a real solution for the Libratus enigma. As things stand right now, it seems like the AI will take a decisive victory.

Broader Implications

According to Sandholm, the reason why he decided to use No Limit Texas Hold’em as the testing ground for Libratus is the complexity of the game. The end-game isn’t beating humans at poker, of course, but developing an artificial intelligence capable of tackling real-life situations with incomplete and imperfect information. Such an AI could be used in many important areas, like military strategy or cyber security.

However, online poker players aren’t too happy with this development. They are afraid that developing a perfect poker playing bot could seriously damage and even ruin the game. While the scientists may not be too interested in poker as such, there are certainly those who could see this as a profitable opportunity.

Share.

Comments are closed.