Google unveils Sima the new AI that completes video games instead of players

Video game enthusiasts know well that when you get stuck in a level and progression seems impossible, you rely on the so-called “solvers”: expert players who share strategies and tricks to overcome obstacles, revealing the keys to finishing the levels. However, we may soon see a significant change thanks to artificial intelligence, specifically created by Google’s artificial intelligence company DeepMind.
SIMA (Scalable, Instructable, Multiworld Agent) is currently being tested, an artificial intelligence agent trained specifically to complete the challenges normally encountered in video games, processing natural language, understanding 3D worlds and recognizing scenarios and images.
To achieve this, DeepMind collaborated with eight game developers, who provided valuable data on how humans interact with games, including Hello Games, Embracer, Tuxedo Labs, Coffee Stain, and others. An approach that allowed SIMA to learn effectively and distinguish complex situations to refine their skills and be able to play any video game, not only linear ones, but also titles based on freely explorable open worlds.
SIMA, as stated, does not need a customized API to play or access to the source code of the games, but simply “plays” having around 600 skills, from simple movements to the ability to open the menu to use the various features available and thus exploit all the potential offered by the video game.
To ensure that SIMA could interact with the most disparate situations, acting in a free and creative manner, the developers fed the AI ​​even complex titles such as, for example, the open-universe “No Man’s Sky”, in which play being a space explorer in a vast universe, complex shooters like “Teardown”, exploration and survival games like “Valheim” and even the bizarre open-world “Goat Simulator 3”.
In practice, SIMA represents a further step forward in the use of artificial intelligence, opening up new possibilities for players and the video game industry. According to Frederic Besse, chief engineer of Google DeepMind, «Ai agents will be able to take control of computers to perform complex tasks, becoming an integral part of the gaming experience, personalizing challenges and adapting to player preferences».