Decision-Making Robot Defeats Four Human Poker Pros, Now Looking For A Real Job

Image courtesy of Adam Fagen

We usually deploy robots and artificial intelligence in jobs that humans find repetitive or unfulfilling, like factory tasks, big box store customer service, telemarketing, and mall security. But now there’s a robot that wants to take away a job that seems enjoyable: pro poker player.

The researchers who created a poker-playing machine that recently bested four human pros weren’t trying to create a crude version of Bender from Futurama or a self-sustaining funding stream for their lab. It was a test of an artificial intelligence program developed at Carnegie Mellon University called Liberatus.

Playing poker means processing huge amounts of information, and also means that the computer must take into account the possibility that its human opponents are bluffing. That’s called making decisions based on imperfect information, and is something that humans have to do as well. In a move that will inevitably doom us all when Skynet takes over, the program learned to bluff, too.

As with IBM’s Watson system, which was a champion Jeopardy! player but now works hard finding the best treatment options for cancer patients, and helping tourists in multiple languages, the researchers hope that Liberatus will have similar knowledge and insight into human nature so it can help with medical decisions, business negotiations, military strategy, and cybersecurity.

The bot and four humans played 120,000 hands of poker over 20 days, with Liberatus coming out ahead by $1,766,250 in chips. The four opponents shared a $200,000 pot of actual money.

(via Ars Technica)