Skip to main content
ABC News
A Computer Just Clobbered Four Pros At Poker

About three weeks ago, I was in a Pittsburgh casino for the beginning of a 20-day man-versus-machine poker battle. Four top human pros were beginning to take on a state-of-the-art artificial intelligence program running on a brand-new supercomputer in a game called heads-up no-limit Texas Hold ’em. The humans’ spirits were high as they played during the day and dissected the bot’s strategy over short ribs and glasses of wine late into the evening.

On Monday evening, however, the match ended and the human pros were in the hole about $1.8 million. For some context, the players (four men and the machine, called Libratus) began each of the 120,000 hands with $20,000 in play money, posting blinds of $50 and $100. Here’s how the days progressed:

roeder-poker-update-1

Tuomas Sandholm, a Carnegie Mellon computer scientist who created the program with his Ph.D. student Noam Brown, was giddy last week on the match’s livestream, at one point cheering for his bot as it turned over a full house versus human pro Jason Les’s flush in a huge pot, and proudly comparing Libratus’s triumph to Deep Blue’s monumental win over Garry Kasparov in chess.

And, indeed, some robot can now etch heads-up no-limit Texas Hold ’em (2017) — alongside checkers (1995), chess (1997), Othello (1997), Scrabble (c. 2006), limit Hold ’em (2008), Jeopardy! (2011) and Go (2016) — into the marble cenotaph of human-dominated intellectual pursuits.

Brown told me that he was keen to tackle other versions of poker with his A.I. algorithms. What happens when a bot like this sits down at a table with many other players, rather than a single foe, for example? Sandholm, on the other hand, is quick to say that this isn’t really about poker at all. “The AI’s algorithms are not for poker: they are game independent,” his daily email updates read. The other “games” the algorithms may be applied to in the future: “negotiation, cybersecurity, military setting, auctions, finance, strategic pricing, as well as steering evolution and biological adaptation.”

Another of the human pros, Jimmy Chou, had had just about enough of Libratus.

https://twitter.com/Chouchoupoker/status/826193184002867200

Last week, between especially frustrating hands and with the match quickly slipping away, Les jokingly suspected that the poker match was “being co-opted by the Carnegie Mellon psych department. ‘What can we put a human through?’”

Read more: “The Machines Are Coming For Poker”

Oliver Roeder was a senior writer for FiveThirtyEight. He holds a Ph.D. in economics from the University of Texas at Austin, where he studied game theory and political competition.

Comments