AlphaZero from DeepMind was in trend in the year 2017 with the release of an AI-based community.
Typically an AI-based program makes games like Go, Shego, and Chess, but what AlphaZero does differently is, it uses RL (Reinforcement Learning) with Search(RL+Search), to “learn on its own” by mimicking the world-class players. In the year 2019, Facebook introduced the game named Pluribus Bot which could defeat human experts in six-player no-limit Hold ’em, which was the most popular one.
There is no generalized Artificial Intelligence based algorithm that could help play both Chess and Poker. In technical term the games like Chess are classified as perfect information game, where the player plays according to the movement of the opponent, and also the players are aware of all the possibilities, whereas games like poker are classified as imperfect information game where the players need to balance all the probable outcomes while playing.
The AlphaZero was designed for playing games like chess, but it does not work when imperfect information games are played with it. To overcome this challenge Facebook introduced the general Artificial Intelligence algorithm and named it Recursive Belief-base Learning (ReBeL). ReBel is a modified version of Reinforcement Learning +Search(RL+Search). ReBel is the same as AlphaZero but it comes with some additional capabilities where it estimates the possibilities of the opponent with a particular card, for example, a pair of aces or a pair of spades. While experimenting, ReBel was able to beat a human expert. ReBel proves to work with Liar’s Dice, which is a type of imperfect information games.
ReBel knows the rules of the game, which makes it useful for games like Shego, Go, and Poker. However, the ReBel Algorithm is limited to only two players as of now. Computational prowess is high in ReBel. Games such as Racon has comparatively less computational prowess, the players are not aware of the opponent’s position and depend on sensing. However, despite the shortcomings of the ReBel, it has reached a massive platform. For the future, it is considered as a step towards complicated world applications like cybersecurity and fraud detection.
