Move over overpowered AI opponents, Google DeepMind is developing a new kind of gaming partner: SIMA (Scalable Instructable Multiworld Agent). Instead of aiming to crush you, SIMA is designed to learn and play alongside you, adapting to different games and following your instructions.
This research project isn't meant to replace existing game AI, but rather to create a collaborative teammate. SIMA understands 3D environments, recognizes objects, and can even follow your spoken commands. DeepMind researchers, led by Tim Harley, highlight how SIMA focuses on learning and following instructions within a game, rather than just winning at all costs.
To train this versatile AI, Google partnered with eight game developers, including the makers of No Man's Sky and Goat Simulator 3. Impressively, SIMA doesn't need special access to game code, it simply learns by watching and interacting with the game world.
Please send a mail and let us know about your concerns.We will try our best to provide optimized solutions.