Is it possible for a computer to be better in No-Limit Texas Hold’em Poker?
In 1957 two scientists working on artificial intelligence Herbert Simon and Allan Newell forecasted, that in the next 10 years computer will be able to beat the best chess player. In 1997 IBM Deep Blue computer won over world’s chess champion Garija Kasparow (31,5-21,5). That was exactly 30 from the time two experts estimated. Even the smartest people were wrong- writes Lee Davy in his article.
Currently in USA takes place 2-week poker game- on one hand four talented poker players- Doug Polk, Jason Les, Dong Kim and Bjorn Li, and on the other computer called Claudico (in Latin- limp). Game will last 80,000 handouts of No Limit Texas Hold’em. By now the scientists “solved” Pot-Limit Texas Hold’em. On University of Alberta created a program called “Cepheus”- in two months it played over million handouts learning the strategy this way. As we wrote on January this year, the program doesn’t win every handout, but in a longtime play is unbeatable. Cepheus learns by algorithm that considers every decision linked with the handout, and then learns which move was beneficial and which wasn’t.
In his article, Lee Davy admits, that the duel in Claudico upsets him. He says: the last thing we need is the machine able to win at the most complicated game ever invented.
Scientist Tuomas Sandholm Carnegie Mellon University, where Claudico was invented says: I think there is a big chance that we will lose. Doug Polk sees it in a different way. –I don’t have to worry about stuff like “if he thinks that I think etc.”
Sandholm admits, that algorithms used in their program are widely use in sectors with incomplete information, also in internet safety, negotiations or medicine.
The duel between poker players and Claudeco will be emitted live on Twitch and tagged on Twitter with #BrainsVsAI.
Good news- at the moment poker players are winning (+166,566 virtual dollars). 16,500 of 80,000 of handouts was played.