Skip to Content
Artificial intelligence

An AI Poker Bot Has Whipped the Pros

It’s another seminal moment for machine learning, and a painful schooling for humans.
January 31, 2017

Humans have been bested by a computer in yet another game once considered too difficult for artificial intelligence to master.

Over the past three weeks, an AI poker bot called Libratus has played thousands of games of heads-up, no-limit Texas hold’em against a cadre of top professional players at Rivers Casino in Pittsburgh. And it beat them all.

Our own Will Knight recently explained why victory for Libratus, which was built by a pair of researchers at Carnegie Mellon University, would be such a big deal:

Poker requires reasoning and intelligence that has proven difficult for machines to imitate. It is fundamentally different from checkers, chess, or Go, because an opponent’s hand remains hidden from view during play. In games of “imperfect information,” it is enormously complicated to figure out the ideal strategy given every possible approach your opponent may be taking.

In heads-up, no-limit Texas hold’em, then, it's virtually impossible, for there is no single correct play. Instead, the AI must use game theory to calculate optimal plays given the uncertainties.

In the end, it wasn’t even close: Libratus made off with $1.8 million in chips, while all four of the pros ended up with a deficit. Artificial intelligence has never beaten top players at a game so lacking in information as no-limit Texas hold’em. Like DeepMind’s Go victory before it, then, the win is a seminal moment for the machine learning community.

But what was it like for the humans to play against? “It’s slightly demoralizing," Jason Les, one of the professionals, told the Guardian. "If you play a human and lose, you can stop, take a break. Here we have to show up to take a beating every day for 11 hours a day. It’s a real different emotional experience when you’re not used to losing that often.”

Daniel McAulay, another professional, explained to Wired that the AI's ability to hold different plays in its memory made it stand apart from human contenders. “It splits its bets into three, four, five different sizes,” he explained. “No human has the ability to do that.” Still, don’t feel too badly for the vanquished humans: despite losing, they're divvying up $200,000 between themselves based on how well they did during the tournament.

For the AI, though, this is just the start. Having proven that it’s possible to beat the professionals at their own game, there is now a very clear next challenge to chew on: multi-player no-limit Texas hold’em. But the game theory used in the current software falls down when there's more than one opponent, and it's not clear what technique to use instead.

Still, given the progress machine learning is currently making, and the fact that other AI poker bots are also being developed, that seemingly impossible challenge may not remain impossible for long.

(Read more: BloombergWired, The Guardian, “Why Poker Is a Big Deal for Artificial Intelligence,” “Poker Is the Latest Game to Fold Against Artificial Intelligence,” “Five Lessons from AlphaGo’s Historic Victory”)

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.