Carnegie Mellon computer scientists achieved a major new computational milestone with their algorithm DeepStack recently. Now the full analysis in the peer-reviewed journal Science is available to the public.
Artificial intelligence has made previous breakthroughs in other games like Chess where each player has a full understanding of all the game specifics at every moment of the content. However, in a game like poker where key information is missing, bluffing is possible and players can make illogical moves computers have continued to struggle… until now.
Poker is the quintessential game of imperfect information, and with DeepStack, combining recursive reasoning in innovative ways, the algorithm was able to handle “information asymmetry, decomposition to focus computation on the relevant decision, and a form of intuition that is automatically learned from self-play using deep learning” according to its creators.
In a study involving 44,000 hands of poker, DeepStack defeated professional poker players in heads-up no-limit Texas hold’em card tournament play. Obviously poker is just the first of many applications for this kind of computational advancement. As we all live in a world that provides us only a limited viewpoint with an imperfect amount of information available prior to virtually any decision we make in business or personal life, the ability of artificial intelligence to assist humans in navigating difficult choices may soon be a reality.
Here at NationalNet, utilizing new AI to protect client data security, predict data throughput patterns and anticipate bottlenecks that can be further optimized before they occur are potential ways to deploy DeepStack styled technologies, and we are keeping a close eye on their continued development as part of our own vigilant approach to making sure our clients are always holding a winning hand.