multi-tabling online poker

By: Nick Gambino

In the newest chapter of “This is How the World Ends” an AI bot, Pluribus, participated in a 12-day poker game against five human pros and won.

The game was a knock-down drag-out that saw them playing over 10,000 hands in no-limit Texas hold ’em. At the end, the non-human “walked away” with $48,000, making an average of $1,000 an hour. The players, each of whom had made over $1 million in their time as professional players, had been picked from a pool of willing participants.

Now, this isn’t the first time an AI bot has won a serious poker game against a human, but this is the first time they’ve won against multiple humans in one game. If you’ve ever played Texas hold ’em then you know the more players at the table, the more complex the game gets. Hence, why the game took 12 days.

“It’s the first time AI has achieved superhuman performance in a multiplayer game,” Tuomas Sandholm, one of the AI’s developers at Carnegie Mellon University said.

The reason this is a major accomplishment in the field is that the level of complexity is beyond anything they’ve been able to get AI to conquer before. Texas hold ’em against multiple players introduces some wild variables and also includes the added element of unknowns.

“When you go to a game like poker there is hidden information involved, where you have access to information that your opponents don’t see, and that greatly complicates things,” said Sandholm’s developing partner Noam Brown, a Facebook AI research scientist.

In an earlier test, Darren Elias, a four-time World Poker Tour champion, played Texas hold ’em against five AI bots who were not able to collaborate with each other. Can you guess who won? Hint: it wasn’t the one cloaked in flesh who breathes real oxygen. In the thousands of hands Elias played, the multiple Pluribuses learned what worked and what didn’t until it improved so vastly it could be considered at the skill level of a world-class professional poker player.

So when you’re eventually melting down carbon in grueling factory conditions to feed your AI overlords, you can thank the brainiacs over at Facebook AI and Carnegie Mellon.