Artificial Intelligence takes an important step forward: Poker-Playing AI Beats Professionals
A study published recently in Science describes an Artificial Intelligence (AI) system called DeepStack that defeated professional human players in the popular card game poker. The feat represents a leap forward in developing artificial intelligence that can learn with incomplete information.
For decades scientists developing artificial intelligence have used games to test the capabilities of their systems and benchmark their progress. However games such as Chess and Go are ‘perfect information’ games where all the required information is upfront and the AI is supposed to make a decision using the complete information. There are on the other hand so-called “imperfect-information”games like poker where there’s hidden information that only one player knows, and that makes it much more difficult to create AI that can make winning strategies based on the imperfect-information.
The new AI system DeepStack, developed by researchers at the University of Alberta, relies on the use of artificial neural networks that researchers trained ahead of time to develop poker intuition. During play, DeepStack uses its understanding of the game and playing strategies to break down a complicated game into smaller, more manageable pieces that it can then work through rapidly. Using this strategy allowed it to defeat its human opponents. In the view of experts DeepStack is an important step in AI toward tackling complicated, real-world problems.
The DeepStack artificial intelligence system outperformed humans at the ultimate, no-limit version of Poker, a complex game involving 10^160 possible moves. Credit: Carla Schaffer AAAS
Link to source article: https://www.scientificamerican.com/article/time-to-fold-humans-poker-playing-ai-beats-pros-at-texas-hold-rsquo-em/