Google The DeepMind team recently released SIMA Model,Through training and learning of game skills, its overall operating style is made closer to that of humans.
SIMA stands for "Scalable, Instructable, Multiworld Agent". It is currently only in the research stage and will be able to play games with you in the future.
SIMA is not positioned as an AI NPC in the game, but to be your teammate and cooperate with your operations tacitly. SIMA combines natural language and image recognition technology to understand the 3D game world, hoping to become a tacit teammate of players.
Google worked with eight game developers including Hello Games, Embracer, Tuxedo Labs, and Coffee Stain to train and test SIMA.
Google has trained SIMA in No Man's Sky, Teardown, Valheim and Goat Simulator 3 to teach AI agents the basics of playing games. Google said in a blog post that SIMA doesn't need a custom API to play games or access source code.
SIMA currently has about 600 basic skills, such as turning left, climbing ladders, opening menus and using maps, etc. It is not yet able to implement complex commands such as "finding resources and building camps", but it is hoped that related skills will be further expanded in the future.
Attach the paper reference