Poker bot Libratus has shown no mercy to the four human poker pros who challenged it to an almost three-week match, each playing the artificial intelligence (AI) creation of the Computer Science Department at Carnegie-Mellon University in heads-up matches.
Depressingly, it appears that human dominion over the game of no-limit hold’em is at an end. On Tuesday night, at the Rivers Casino in Pittsburgh, Libratus officially whupped humanity at poker for the 20th day straight.
The entire experience was, said heads-up contender Jason Les, utterly exhausting.
“Libratus turned out to be way better than we imagined. It’s slightly demoralizing,” Les told the UK’s 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.”
Les and his team of pros, comprising Dong Kim, Jimmy Chou, and Daniel McAulay, collectively finished the 180,000-hand challenge just under $1.8 million down to Libratus, although it was not real money. Of the humans, Dong Kim fared the best: he was just $85,649 down, while Les fared the worst, with an $880,000 loss.
Poker Bot Not Favored
Libratus was not the favorite going into the match. The computer had been priced at a 4/1 underdog by bookies, who had studied the human “form” and noticed we tend to do well in these contests. After all, in a similar match-up in 2015, Libratus’ predecessor, Claudico, was beaten by a human team that included Les, Kim, and Doug Polk.
While computers have been able to beat Homo sapiens at chess since the 1980s, no limit hold’em, as a game of imperfect information with no “perfect” strategy, proved to be a much tougher prospect for programmers.
Tuomas Sandholm and Noam Brown, who built Libratus, had been tight-lipped about the computer’s methodology, but after its victory, were willing to open up to the press. They also said they had been by no means sure that their creation would beat the humans.
Self-Taught Super Computer
“We didn’t tell Libratus how to play poker,” said Brown. “We gave it the rules of poker and said ‘learn on your own.’ The bot started playing randomly, but over the course of playing trillions of hands, was able to refine its approach and arrive at a winning strategy.
“When I see the bot bluff the humans, I’m like, ‘I didn’t tell it to do that. I had no idea it was even capable of doing that.’ It’s satisfying to know I created something that can do that,” Brown added.
In fact, it’s inaccurate to call Libratus a “pokerbot.” Ultimately, it’s a supercomputer that has been instructed to learn the rules of the game.
Thankfully, Libratus is unlikely to ever be found plugged into PokerStars grabbing our money. Its future applications lie away from the game, as a decision-maker in the fields of business, the military, and science.
But the truth is, we all should be a little bit concerned for our jobs. They may even one day teach Libratus to craft elegant poker articles, which would put our infinite team of monkeys on typewriters out of work.
The price we pay for progress.