Artificial Intelligence has learned to successfully play poker

As the famous saying goes, "when gambling, casino always wins." In the case of online casinos, casinos often do not win, but special bots "trained" for certain tactics and not knowing fatigue, but as you know, progress does not stand still, and two groups of scientists independently decided to raise the bar higher. They are training artificial intelligence systems for poker, which are already making significant progress in this.

The development is based on scientists from the University of Alberta, Charles University and the Czech Technical University, who created a system called DeepStack AI. Their AI learns to play one of the most common types of poker - Texas Hold'em. During the tests, DeepStack AI played 44,852 hits against 33 professional players, members of the International Poker Federation, from across the world. During all games, the AI ​​showed a success rate of 492 mbb / g (milli-big-blinds per game). It is worth saying that quite experienced players consider the indicator in the region of 100 mbb / g already high enough.

The second similar system was created on the basis of the Carnegie-Mellon University and was named Libratus. Libatrus is being prepared right now to play in one of the largest casinos in Pittsburgh. AI from the scientists at the Carnegie University - Mellon will play for 20 days, and in the course of this he will play about 120,000 parties. If Libratus shows the same high results as DeepStack AI, it's worth forgetting about visiting online casinos even for the most inveterate players. Although scientists are not going to earn extra money with the help of their inventions, sooner or later such technologies will become available and unfair to the players.

The article is based on materials https://hi-news.ru/entertainment/iskusstvennyj-intellekt-nauchilsya-uspeshno-igrat-v-poker.html.