Scientists at Microsoft Research (MSR) Asia created an artificial intelligence (AI) system that learned the intricacies of playing mahjong.
Mahjong is considered one of the most difficult games to master in terms of algorithms. it has a random factor.
Super Phoenix (Suphx) became the first AI system to achieve 10th dan (rank) on Tenhou, an online riichi mahjong competition platform with over 300,000 participants from all over the world. Only 180 people have this qualification. This is the maximum level for the algorithm at the moment, only a few recognized professionals play at a higher level.
To create the system, deep reinforcement learning was used – a complex principle of learning algorithms based on a combination of deep learning and reinforcement learning. Suphx has mastered the intricacies of mahjong by playing 5,000 games in 4 months. The system gradually increased the quality and speed of the game. Now the algorithm has its own strategy, which includes intentionally weak moves, allowing to gain a greater advantage in the end.
Unlike chess and go, where AI has already made significant progress, mahjong is a game of incomplete information. Participants do not know all the factors, for example, you cannot see the images on the opponent’s knuckles. The game process is similar to poker – the participants need a good memory, observation, strategic thinking and the ability to bluff.
According to scientists from MSR, the presence of a random factor in mahjong is a unique challenge for AI systems, and the experience gained in teaching algorithms for nonlinear problems can be applied in real life situations associated with a set of unknown factors.
Previously, Microsoft Research and Maluuba trained artificial intelligence to play Ms. Pac-Man, which was released for one of the first Atari 2600 home consoles. The algorithm broke the world record with 999,990 points, while the best human score is 266,360 points.
Description
Mahjong or Majiang is a Chinese game of chance using dice for four players (each player plays for himself). Widely distributed in China, Japan and other countries of East and Southeast Asia. The game is played with bones resembling dominoes, according to the rules it is similar to poker, it requires players to have such qualities as experience, memory and observation. There is also a random factor in the game, the role of which, depending on the game rules used, can be either small or decisive. The goal of the game is to score as many points as possible by collecting the most valuable combination of a given number of dice.
It is necessary to distinguish gambling game of mahjong from mahjong solitaire – the latter is a game for one player, an analogue of card solitaire, played with mahjong dice.
Initially, the rules of Japanese mahjong were very close to those of China (this variety is now known as “Japanese traditional mahjong”). However, since the 1950s, the rules have changed and a version of mahjong has emerged that is now referred to as “Japanese modern”, “reetee mahjong” or “dora reetee mahjong”. Of the features of this type of mahjong, the most famous is the “reetee” rule – the player’s announcement of the obligation to collect a winning combination with a deposit of 1000 points, as well as the use of “dor” – bones randomly selected for each round, the presence of which in a winning hand increases its value. Some players in Russia use the names “riichi mahjong” and “riichi”, derived from reading the English transcription.