The artificial intelligence research teams from EquiLibre Technologies, Sony AI, Amii, and Midtravel have collaborated with Google's DeepMind project to develop an AI system called "Student of Games" (SoG). This system is capable of defeating humans in various games and learning to play new games. In their paper published in the journal "Science Advances," the team describes the new system and its functionalities.

Over the past half-century, computer scientists and engineers have developed the concepts of machine learning and artificial intelligence, using human-generated data to train computer systems. This technology has been applied in various scenarios, including playing board games or indoor games.

Teaching computers to play board games and then improving their abilities to the point where they can defeat humans has been a milestone in demonstrating the level of AI development. In this new research, the team has taken a step further towards artificial general intelligence.

So far, most computer systems built for playing board games have focused on a single game, such as chess. In the process of building such systems, scientists have developed a type of limited artificial intelligence. In this new endeavor, researchers have created an intelligent system capable of playing various games that require different skills.

In terms of gameplay, there are primarily two types of games: those with perfect or imperfect knowledge. The former refers to games where both players have perfect knowledge of the game, such as knowing the positions of all game pieces. The latter includes games like poker, where only partial information is known to individual players. SoG can play both types of games and has the ability to defeat professional humans.

So far, it has already defeated other AI systems and humans in games such as Go, chess, Scottish draughts, and Texas hold 'em poker. The research team suggests that it may excel in other types of games as well, as it is capable of self-learning to play almost any game.