Can this Al win a $200,000 poker battle?

An AI system named Libratus will go up against top poker players during matches later this month in Pittsburgh.

Artificial intelligence has already become a part of our everyday lives through AI-assisted services like Siri. But Al has its own hobbies too, including the mind-whirling game of poker.

Scholars at Carnegie Mellon University have developed an AI system named Libratus that will wage a poker battle (with a $200,000 pot) against four of the best human pros out there. In a tournament called Brains Vs. Artificial Intelligence: Upping the Ante, Libratus and the humans will play matches of Heads-Up No-Limit Texas Holdem beginning on Jan. 11 at the Rivers Casino in Pittsburgh, Pennsylvania, the university announced in a press release.

A total of 120,000 hands will be played during the 20-day tournament. The professional players Libratus will compete with include Jason Les, Dong Kim, Daniel McAulay, and Jimmy Chou.

It wouldn’t be the first time bots have played some big games against humans. Indeed such match-ups go back decades. Last March, a Google-produced bot named AlphaGo easily beat 18-time world Go champion Lee Se-dol.

Image: Google via getty Images

But the thing is that poker isn’t like many of these other games.

As explained by the MIT Technology Review, the game of poker may not be as easily dominated by computerized players as others on accounts of its unique rules.

“Unlike board games such as Go or chess, poker is a game of ‘imperfect information,’ and for this reason it has proved even more resistant to computerization than Go,” the MIT Technology Review‘s Will Knight once wrote.

In games like chess, “you know exactly what has happened in the game so far,” explains Carnegie Mellon computer science researcher Tuomas Sandholm, who developed Libratus alongside Ph.D. student Noam Brown. Such an advantage doesn’t exist in poker, where there’s crucial unknown information like your opponents’ hands.

And the game of Heads-Up No-Limit Texas Holdem has been especially difficult for AI to take on since it deals with a huge scale of probabilities making it a “much bigger game” than Limit Texas Hold’em, Sandholm said.

In the past, game-playing AI has used strategies like programmed human knowledge or machine learning. But more recently, the approximation of equilibrium strategies or rational play has been more effectively used by AI to compete and sometimes beat human players, Sandholm told Mashable.

“It’s very different from how humans play this game.”

Libratus uses the Nash Equilibrium, named after the famed mathematics scholar John Nash, the subject of A Beautiful Mind.

“Named for the late Carnegie Mellon alumnus and Nobel laureate John Forbes Nash Jr., a Nash equilibrium is a pair of strategies (one per player) where neither player can benefit from changing strategy as long as the other players strategy remains the same,” a statement from CMU says. “One of Libratus new technologies is a faster equilibrium-finding method. It identifies some paths for playing a hand as not promising. Over time, the algorithm starts to ignore those bad paths.”

Sandholm and others also created Claudico, an AI system that ultimately lost to human players at poker matches last year. But Sandholm said Libratus has been improved, with more core hours put into programming the system, two new algorithms intended to better play the game and more computing resources and hardware.

It has been programmed to know the rules of poker, but Libratus doesn’t rely on any other information. For example, it has never referenced material from a poker book or poker expert or any other sources, Sandholm told Mashable. This means it will sometimes play the game in a totally new way, making moves that may seem unfavorable to the best of traditional human players.

“It’s very different from how humans play this game,” Sandholm said of the Libratus approach.

“It plays like a martian,” he said. “It’s deriving its own strategy from just the rules of the game.”

Read more: http://mashable.com/2017/01/06/ai-poker/