Artificial Intelligence (AI) made a major advance over the weekend – that indicates their ascendancy might be closer than we think. An AI card shark named Libratus beat four top professional poker players at Texas Hold’ Em.

Winning at cards is a major advance for artificial intelligence, because it means the machine can out think human card sharps. Libratus was able to win by making guesses and hunches, just like a human poker player would.

What’s more disturbing is that Libratus might be a far better card player than any human being – even World Series of Poker aces such as Phil Hellmuth. Its’ creators at Carnegie Mellon University may have finally created a machine that can outwit some of the trickiest humans, The Christian Science Monitor reported.

It was an absolute beat down

“Once you face Libratus, there’s nothing worse any human could ever do to you,” one of poker pros it bested; Jason Les, admitted. “Every human is going to seem like a walk in the park.”

“It was an absolute beat down,” professional player Doug Polk; another of Libratus’s victims, said of the Brains vs. Artificial Intelligence Poker Tournament at the Rivers Casino in Pittsburgh in a Forbes interview. “The human team lost at a rate of approximately 2.5 times what the program lost at before.”

Even Libratus’s creators; Carnegie Mellon engineers Tuomas Sandholm and Noam Brown, were surprised by the size of its win, The Monitor reported. Disturbingly Libratus seems to be a better poker player than the human pros.

The tournament lasted three weeks and involved 120,000 hands of poker. At the end all the human players were in the red and Libratus was ahead by 1.77 million chips. Strangely enough Libratus did not win any money, but the pros each received a $50,000 payment for their participation.

The “biggest difference is that most human poker players do what Libratus does, betting with multiple bet sizes, but humans only have one or two usually,” a third defeated pro, Dong Kim said. “Libratus mixes in a bunch, maybe 10 or 15, and even mixes in situations that I thought didn’t make sense before this competition started. It would be way too complicated for a human to do this sort of thing correctly.”

Did an AI learn how to Play Poker?

Here’s the scariest part; Libratus was not designed or programmed to play poker. It apparently learned the game on its own the hard way – by playing the humans.

“None of the algorithms in Libratus are specific to poker,” Brown admitted. “We did not program it to play poker. We programmed it to learn any imperfect-information game, and fed it the rules of No-Limit Texas Hold’em as a way to evaluate its performance.”

Libratus’s achievement is big news until now AI’s have only been able to win perfect information games such as chess, Go and checkers. In those games each player knows everything about his opponent’s game.

Cards are different, because players only have part of the information. They know what cards they are holding, but they have to guess what their opponents have. This is why almost all card games; including poker and bridge, involve bluffing.

Until now, human pros have easily been able to best AIs at poker by simply bluffing. Back in 2015 Brown and Sandholm’s last AI; Claudico, was easily defeated in a similar tournament by a wide margin.

Therefore it should be possible to employ something like Libratus for many tasks beyond poker. For example it might be able to pick stocks, devise military strategy or run a political campaign.

Libratus Might Want Your Job

“In truth, most real-world scenarios involve hidden information,” Brown said. “In the real world, not all the information is laid out neatly for all sides to see like pieces on a chessboard. There is uncertainty and deception.”

It looks as if truck drivers, accounting clerks and auto workers are not the only ones that have to worry about technological employment. AIs might soon be taking work from professional poker players, investment bankers, political consultants, football coaches, odds makers, lawyers and generals.

You should be very careful next time you play poker online. You might be facing Libratus, and it may take all your money.

https://i0.wp.com/geekcrunch.reviews/wp-content/uploads/2017/02/BrainsVSAIHalfwayPoint_853x480-min.jpg?fit=853%2C480&ssl=1https://i0.wp.com/geekcrunch.reviews/wp-content/uploads/2017/02/BrainsVSAIHalfwayPoint_853x480-min.jpg?resize=150%2C150&ssl=1Daniel JenningsInnovationAI,AI card shark,Artificial Intelligence beats Poker Champs,Carnegie Mellon University,Did an AI learn how to Play Poker?,Dong Kim,Doug Polk;,It was an absolute beat down,Jason Les,Libratus,major advance for artificial intelligence,No-Limit Texas Hold'em,Noam Brown,perfect information games,Tuomas SandholmArtificial Intelligence (AI) made a major advance over the weekend - that indicates their ascendancy might be closer than we think. An AI card shark named Libratus beat four top professional poker players at Texas Hold’ Em. Winning at cards is a major advance for artificial intelligence, because it means...From A Geek to a Geek