Tuesday 20 January 2015

Researchers Turn Computer Into Poker Shark

According to the researchers at the University of Alberta, a computer program dubbed as “Cepheus” has been able to solve the game of Poker. Cepheus, has been created by computer poker research group, asserts that it can not only play heads-up limit Texas hold'em poker, but also beat human opponents.

When it comes to artificial intelligence, games like poker and chess have always been used as the test beds by researchers who are looking towards developing new concepts. There have been many success scenarios in the past like IBM's Deep Blue, which had defeated world champion Garry Kasparov in chess, and IBM's Watson, which defeated top-ranking Jeopardy! Champions. When it comes to a game of poker, it gives special challenges to artificial intelligence researchers because the computers or the players do not have the complete knowledge of past events and also they cannot see the cards held by their opponents.

Test of Cepheus: 

According to CPRG, any game can be solved if it is played over a period of time and if it is played with 95% confidence. The computer was trained to test Cepheus with poker over a period of two months. During this time it utilized more than 4,000 CPUs, and each of this CPU can handle 6 billion hands of poker. This was more poker which had been played by my entire human race.

Each of this hand, Cepheus constantly kept on managing its play and the choices kept on refining their choices and to find the best possible solution to each move. The main effort of University of Alberta's effort is to use a computer to play and win an imperfect information game when compared to perfect information games like checkers, chess and tic-tac-toe. The players will have access to all the information pertaining to the game.

While developing Cepheus, the CPRG's goal was to make sure the technology is able to make the computer think in some of the imperfect situations. This technology can be applied to any situations where there is a possibility of imperfect situations. According to Neil Burch, a Ph.D. student at the University of Alberta and a coauthor of the study, it will be interesting to see this research being used in different problems. As far as Cepheus is concerned, it has proved its power with the cards but no one can say about its ability as it will be dependent on the limitation of the computational capabilities.

Computers will learn more as the time passes by as CPU/GPU, storage and RAM will continue to evolve which will allow more complex supercomputing, software programming and problem solving. When it comes to computer, if it goes wrong it will learn to go right and take the correct step in the situation again.

It is important for the computer to learn every possible step based on the database and determine the best possible move. Even though the artificial intelligence has lot awaiting in the future, Cepheus has a lot of contribution towards it.

No comments:

Post a Comment

Note: only a member of this blog may post a comment.