“It’s safe to say we’re at a superhuman level and that’s not going to change.”
An artificial intelligence called Pluribus, created by Facebook’s AI lab and researchers from Carnegie Mellon University, just beat a team of 12 of the world’s best human online poker players.
The AI won an average of five bucks per hand, cashing in roughly a grand every hour.
“It’s safe to say we’re at a superhuman level and that’s not going to change,” Noam Brown, a research scientist at Facebook AI Research and co-author of the paper that was published in the journal Science today, told The Verge.
To get to that extraordinary level of complexity and competency, Pluribus first learned how to play against — and beat — versions of itself. Rather than getting caught up trying to predict its opponent’s moves all the way to the end, the AI focused on only two or three upcoming moves.
“The AI doesn’t see bluffing as deceptive. It just sees the decision that will make it the most money in that particular situation,” Brown told The Verge. “What we show is that an AI can bluff, and it can bluff better than any human.”
READ MORE: Facebook and CMU’s ‘superhuman’ poker AI beats human pros [The Verge]
More on game AIs: Hellishly Hard New Game Is Specifically Designed to Confound AI