[ad_1]
Getty Images For decades, poker has been insurmountable by a computer invasion
Computers have invaded the world of the most popular games in the world, such as Checkers, Scrabble and Go.
Poker has now joined this list of games, although players often rely on deception to defeat their opponents.
On July 12, for the first time in history, BluePbad, a poker software, has defeated several heroes simultaneously in one of the most popular forms of poker: Texas Hold 'Em unlimited. ".
This may be bad news for online poker players, but could it benefit humanity at other levels?
- American earns $ 8.53 million in poker
- Malicious Internet program "weasels" on poker players
What is the difficulty of computer poker?
For decades, poker has been too complex for a machine to learn.
Unlike the schering game, in which all the coins and stages are visible, poker is considered a "lack of information" and with the cards hidden, players only have one acquaintance partial of the current state of the game.
CMU Noam Brown (right) contributed to software development
Poker is best known for its deception, or to try to lure others, telling it that it has better or worse cards.
With these factors, the AI can not develop strategy to win the game.
Crash Legend
But the myth of human domination over poker has been broken.
Yes, and BlueBas, developed by Noam Brown, a researcher at Facebook's Artificial Intelligence Unit, and Thomas Sandolm, a computer science professor at Carnegie Mellon University.
The software beat 13 professional poker players, including four-time world champion Darren Elias.
2012 Getty Images World Poker Champion Greg Merson was among the professional players defeated by the BlueReps program. Merson is pictured in the picture in the baseball game
Using an artificial intelligence software in the real world
But why are researchers working on machines capable of beating humans?
In summary, scientists believe that the artificial intelligence used in string games or poker can be used to perform more complex tasks.
"We use poker as a landmark to solve more important problems related to the development of artificial intelligence in order to deal with hidden information," Brown told the BBC.
"For AI to be used in the real world and interact with human beings and other AI technologies, it must be able to understand that other participants can see the world differently. from what they seem and have access to hidden information.
"Poker is an excellent simulation of this kind of challenge," said Brown.
Getty Images The artificial intelligence used by Deep Mind in the GO game can be used in research in chemistry and engineering.
When Deep Mind of Google developed the Alpha software, this one became in 2015 the first tool to beat a professional Go player. Software developers felt that they could be used for chemistry tasks and # 39; engineering.
Similarly, Brown and Sandolum believe that the BlueReps problem-solving software can be used in future applications to perform tasks at the levels of financial negotiation and self-driving, in all situations involving multiple parties and missing information.
"Most real-world situations lack as much information as games," said Sandolm.
The CMU promoted the game technologies managed by Sandolm (center) with the US Army last year
How to learn software
Far from being the scientists' wish, Brown and Sandolum developed Libratus two years ago, which crushed their human rivals into individual matches.
The experts felt that the strategy gained by the software in these games could be used in the military simulation.
Reports published by Wired magazine revealed the conclusion of a strategic robot company specializing in artificial intelligence, founded by Sandolm, a $ 10 million deal with the US military in August 2018.
"We can learn a lot of software in terms of decision making," says researcher Adam Kocharsky. The best software can learn by comparing the results and asking, "What if I made a different decision?
"By losing to the benefit of a software program, the human player can learn a lot from his way of playing."
Getty Images' BlueBas capabilities can be used to make strategic military decisions
Electronic phishing
But how to deceive the Bluripas software?
Well, both Brown and Sandolm have managed to make this software a master poker software by putting it into billions of games against themselves and revising their strategy.
The decisions that led to the victory are more used in future games.
One of those decisions was the illusion of the opposing player that the playing cards in his hand were weak and then pushed the opponent to bet on the double value, which is exactly the same. the equivalent of deception.
"People think that deception is a human trait that the machine can not do," says Brown.
"People think that deception is a human trait that the machine can not do," Brown said.
"But the fact that we have seen that the machine can deceive and exceed the capabilities of man"
"The software does not see deception or betrayal of trust in the process, it is only a step that generates the biggest gains given the weakness of playing cards in the hand."
According to Carnegie Mellon researchers, artificial intelligence only requires $ 150, the value of cloud computing resources.
Deep Mind, meanwhile, relies on supercomputers offering more than 5,000 specialized treatment programs, at a cost of several million dollars.
[ad_2]
Source link
