Everything you need to know about playing Libratus, CMU’s poker-winning AI

See what everyone had to say about the Carnegie Mellon bot that beat four poker pros at their own game.

Poker pro Jason Les takes on Libratus at Heads-Up No Limit Texas Hold 'Em.

Poker pro Jason Les takes on Libratus at Heads-Up No Limit Texas Hold 'Em.

MJ slaby / the incline
MJ Slaby

It was official as of 7 p.m. Monday — artificial intelligence Libratus beat four top poker pros at Heads-Up No-Limit Texas Hold’em in a 20-day tournament at Rivers Casino.

Creating a bot that would beat humans in this game has eluded the AI community for decades. In order to win, Libratus had to know how to bluff — and when it was being bluffed.

In the end, Libratus led the pros by $1,766,250 in chips, according to a news release from Carnegie Mellon University, where it was developed. The AI runs through Bridges, a new supercomputer at the Pittsburgh Supercomputing Center.

The historic win sparked coverage from outside of Pittsburgh, which included lines about humanity folding and at least one reference to Kenny Rogers’ “you’ve got to know when to hold ’em, know when to fold ’em.” We looked at as many articles as we could about the AI victory. These were the highlights.

‘Inhumanly good’

Part of what helped Libratus win was the fact that, well, it wasn’t human.

And not being human means it didn’t have distractions that come with life, Toby Walsh, an AI professor at the University of New South Wales, told TechRepublic. He said:

“It has no sentience, no desires, no consciousness. It’s not going to wake up and decide it wants to do anything else than play poker. That’s not in its code. Liberatus is an idiot savant at poker. And to adapt these ideas to anything else is going to take person months or years of effort.”

Bloomberg Technology pointed out that Libratus “never needs a break” — and doesn’t worry about money.

Libratus also tends to make huge, sudden wagers, violating standard betting conventions by throwing its money into the pot in irregular amounts and at odd intervals. Coming from a human player, behavior like this would be irritating, reckless and, over the long run, expensive. But Libratus’s main attribute as a poker player is that it’s inhumanly good.

And no, it doesn’t call CMU Professor Tuomas Sandholm (who developed Libratus with Ph.D. student Noam Brown) “Daddy,” IEEE Spectrum reported. Why? Well, the AI “can’t speak.”

‘At least I’m not losing any money’

To put it simply, playing Libratus at poker for 20 days was exhausting.

The four poker pros — Jimmy Chou, Dong Kim, Jason Les and Daniel McAulay — started with high spirits and ended the first day with “short ribs and glasses of wine,” FiveThirtyEight reported.

But that turned into working late into the night and days with Starbucks cups, water bottles and Chipotle bags piled around the players, according to Bloomberg. Les told Bloomberg that Libratus’s slow speed was frustrating.

“It makes the days longer,” said Les, an earnest, athletic-looking man who seemed eager to take a few minutes off one afternoon last week. “Waiting should not affect me whatsoever, but sometimes you’re just like, ‘OK, is this going to be over yet?’”

Each night, as the players gathered to analyze and talk strategy, Libratus was using Bridges to improve its strategy, staying one step ahead of the humans. (FiveThirtyEight created a chart that shows Libratus keeping its lead.)

And in the end, the players’ effort was appreciated, but not enough.

“We appreciate their hard work, but unfortunately, the computer won,” Craig Clark, the general manager of Rivers said during a Tuesday press conference. He added that it was “fun to watch [the human players] struggle and play.”

On the bright side, the humans didn’t lose any real money. (They played for shares of a $200,000 prize, according to CMU.) “Usually, you have to lose a lot and pay a lot of money for the experience,” Les said at the press conference. “Here, at least I’m not losing any money.”

‘A lot of potential opportunities’

But in the end, this poker victory wasn’t really about poker or gambling. It was about strategic reasoning, Sandholm said during the news conference.

It shows that AI can take an imperfect situation and output a strategy for the situation, he said, adding “I think there are a lot of potential opportunities for the future.”

Sandholm and others previously told The Incline that those opportunities include business, military, cybersecurity, medicine and having an app that will negotiate and purchase a car. Here are a few more we spotted:

Sandholm will share more about Libratus at the Association for the Advancement of Artificial Intelligence meeting, Feb. 4 to 9 in San Francisco and in peer-reviewed scientific journals, per CMU.