The world’s best poker player is now a computer

For the past 20 days, an artificial intelligence named Libratus has been destroying the world’s top poker players at a no-limit Texas Hold ‘Em tournament in Pittsburgh. The game ended last night, marking another milestone in the history of AI and machine learning.

Unlike AlphaGo, the AI that defeated the world’s top Go player, this machine wasn’t created by the likes of Google. It was invented by a professor-student combo at Carnegie Mellon. Professor Tuomas Sandholm and grad student Noam Brown first designed a version of the AI two years ago, but lost early competitions against other leading poker players.

But not this time.

The four human poker players ended up with a negative chip count, while Libratus finished with $1.7 million. While the 120,000 hands of poker didn’t have financial implications—the poker players will be splitting a $200,000 purse among themselves—it demonstrated the power of modern computing and the potential capabilities of future AI like no other achievement before it.

The significance of this feat comes from the complexities of the card game. Poker poses a number of challenges that go beyond numbers and patterns. The AI needed to learn from incomplete information sets created by bluffs, slow play, and other ploys. It overcame these complications using fundamental machine learning, or a subset of AI that provides computers with the ability to learn without being explicitly programmed.

The poker players, Daniel McAulay, Dong Kim, Jason Les, and Jimmy Chou, picked up on that learning process as it was happening right in front of their eyes. Every day the AI was getting smarter and making fewer mistakes. Going up against Libratus was like fighting a boxer that got bigger and stronger each round.

“It learns from us and the weakness disappears the next day,” poker player Jimmy Chou said during the contest.

Winning a game of Texas Hold ‘Em may not seem like the most game-changing use case, but the skills needed for a card game like poker—long-term strategy, game theory, opponent learning—can be used for making big decisions down the road.

In the future, AI could be used to help make political, economic, and even health-related decisions based on past data sets and the brilliant, fast-learning minds of modern-day computing. 

Phillip Tracy

Phillip Tracy

Phillip Tracy is a former technology staff writer at the Daily Dot. He's an expert on smartphones, social media trends, and gadgets. He previously reported on IoT and telecom for RCR Wireless News and contributed to NewBay Media magazine. He now writes for Laptop magazine.