Artificial intelligence Abused the Strongest poker Players
Humanity was once again put To shame, and the robot Won $
Artificial intelligence completely beat four Of the most powerful poker playersGracefully and without giving the Meat machines the slightest chance, He spent the entire twenty Days of the tournament proving That intuition, "lucky hand" and Bluff were just empty words Compared to soulless but well-Chosen supercomputer algorithms. million in chips. The Libratus program, created at Carnegie Mellon University, won the Competition "Brains Vs. The Texas hold'em tournament A type of poker was Played in the casino for A long twenty days, and You could watch its progress On Twitch.
Artificial Intelligence: Upping the Ante"
The result is already known To you, but initially it Was not so obvious: a Previous similar attempt by AI To beat poker players failed miserably. Then, in, the program Claudico Could not compare with people In skill and lost thousand Dollars in chips. Libratus turned out to be Much stronger in all its Characteristics: it didn't just Play by the rules, it Calculated bluffs, took risks at The right moments for it, And predicted people's actions. The players themselves were discouraged And broken, some of them Admitted that at some point It began to seem like An artificial barrier. intelligence simply reads their thoughts. In General, AI beating Champions In intellectual games like chess And go is not such A big news. But poker is a completely Different matter and a different Level of tasks for the machine. It requires more than just Playing, knowing the location of All the pieces and trying To predict the actions of The enemy. In poker, you have to Act in a situation of Uncertainty, when many data are Unknown and fragmented. Players are bluffing, and their Position and capabilities are initially unknown. But artificial intelligence did it, And it was a big Breakthrough for its creators. In addition, Libratus was also The first such system that Was trained not by observing Real games, but by playing Only with itself and constantly Improving its skills. She played more than a Trillion games in preparation before Meeting a human for the First time at a poker table. All these twenty days of The tournament, professional players tried To find weaknesses in the AI and find mistakes, but It did they are extremely rare. In addition, artificial intelligence itself Recognizes them and constantly adjusts The further game depending on The mistakes made by it And its opponents. Another reason why Libratus was Able to win over people Was revealed, in General, unexpectedly. It turned out that there Are moments when people almost Never take risks, for fear Of instantly losing to the nines. At the same time, the AI didn't care about conventions.
He is not subject to The fear of risk and Simply goes to his goal.
One of the creators of The program, Tuomas Sandholm, enthusiastically Says that it is similar To the victory of David Over Goliath, when David was Not afraid to throw his stone. It is planned that this Ability to act in a Situation of unpredictability and minimal Data will manifest itself in Areas where such a stressful Situation is the norm.
For example, during military operations.
It sounds pretty logical: first We taught them how to Beat us at poker like Kids, and now we're Thinking about how to teach Them how to kill as Much as possible more people. Humanity doesn't seem to Have a chance.