Home Research AI beats humans at Texas Hold’em poker

AI beats humans at Texas Hold’em poker


DeepStack – an artificial intelligence system – has managed to beat humans at the game of Texas Hold’em poker thereby registering another win against the human mind.

An AI system has already garnered a win against human chess players and while in that particular game, players have equal access to information, Texas Hold’em poker brings to the tablet a completely different level of complexity for artificial intelligence systems for each of the players has asymmetrical information about the state of the game.

From the perspective of an AI system, it has to go through a whopping 10160 decision points to arrive at a move, but scientists managed to develop the DeepStack system such that instead of calculating possible moves to the very end, it limits the depth and the type of moves allowed. The team behind DeepStack managed to reduce the possible from 10160 decision points to just 107 decision points. According to the team the reduction in the decision point meant that DeepStack’s computer chip was able to handle all the decision points in just 5 seconds.

DeepStack outperformed, with statistical significance, 10 of 11 professional poker players after 3,000 games each of heads-up no-limit poker (the system also outperformed the 11th player, but that result is not statistically significant).

Poker apart, such artificial intelligence system will enable us to solve real-world problems involving information asymmetry, such as defending strategic resources and making important medical treatment recommendations, quickly and efficiently – something that human brain takes greater amount of time to deliver.