Time to fold, Humans: Poker-Playing AI Beats Texas Hold'em Pros

 Time to fold, Humans! Poker-Playing AI Beats Texas Hold'em Professionals

Poker is a well-known pastime. It's no surprise that the game is so popular. Players are engaged in a complex dance of reacting to and acting out, which becomes more intense with each increasing bet. Poker's entertaining elements have created a complicated problem for artificial Intelligence (AI). DeepStack, an AI system, was recently defeated by professional poker players in Texas Hold'em. Science published a study today. It is a breakthrough in how AI systems can solve problems.

DeepStack is a poker-instinctive software that was created by University of Alberta researchers. DeepStack employs its poker intuition to help it break down complicated games into manageable pieces. This allows it then to work on the fly. It was able to defeat its human opponents by using this strategy.

Scientists who are developing artificial intelligence use games to evaluate their systems' capabilities and track their progress over the years. Twenty years ago, IBM's Deep Blue, an AI-powered chess player, beat Garry Kasparov (World Chess Champion) to make game-playing AI possible. Google DeepMind’s AlphaGo program beat top human progammers in the game, go, last year. However, there is a fundamental difference in games such as poker and chess. This is because of the amount information players have. Murray Campbell, an IBM computer scientist, says that games of chess or go offer 'perfect information'. "Games like chess or go allow you to see everything right in front you so you can make your decision." "In poker and other imperfect-information games, there's hidden information--private information that only one player knows, and that makes the games much, much harder. Researchers in artificial intelligence have been working on poker since the beginning. AI programs from all around the world have competed against human players in poker tournaments, including in the Annual Computer Poker Competition now in its 10th Year. Heads-up, no-limit Texas hold'em presents a particularly daunting AI challenge: As with all imperfect-information games, it requires a system to make decisions without having key information. There is no limit to the number of players, so there are many possible game scenarios. Poker-playing AIs have tried to predict how to play in each scenario before the game starts. In order to solve complex games, such as heads up, no limit, they have relied upon an abstraction strategy, where different scenarios are combined and treated in the same way. For example, it might not be possible to distinguish between kings or aces in a system. Abstraction can simplify the game, but it leaves holes that opponents could exploit. Are There Really Texas Holdem Poker Systems That Work?