DeepStack Emerges Victorious Against Humans

0 Comment

Two AI programs seem to have broken the barrier when it comes to poker and winning against human players.

The advances in AI are part of research studies in different places and one way of testing the advances in such fields is to put robot programs to test against human players in different games of skill and strategies.

While simpler games like go, tic tac toe and others, including chess have already been mastered by AI, poker has always remained a more difficult game to stand up to. There are simpler and more complex poker games. While the simplest games are easy to master and stand up to by such programs against human players, the complex games like Texas hold’em and others often results in failure of the machines, albeit by a narrow margin.

The successful poker bot that has been developed was by the Czech Republic and Canada researchers. The bot has successfully defeated human players in one on one games of the no limit Texas hold’em variant. It is felt that the bot has been able to use something akin to human instincts which is an important variable in these games.

Tuomas Sandholm has been a key figure from CMU who has led the effort of testing bot intelligence against human players. The test involved about 120,000 hands being played over a time span of 20 days. DeepStack is one poker bot that was put to the test. It was developed by a team led by Professor Michael Bowling who represents the University of Alberta. There were other team members from Czech Technical University as well. The results of the event that are ongoing showcase that DeepStack has been successful in playing about 45000 hands of poker against human players and has been able to beat them in each hand.