Google DeepMind is using Gemini to train agents inside Goat Simulator 3
Google DeepMind's new SIMA 2 agent can navigate and solve problems across multiple 3D virtual worlds, representing a step toward more general-purpose AI agents and better robots. Associate Professor Julian Togelius called it an interesting result, noting that previous attempts at training systems to play multiple games haven't succeeded. "Playing in real time from visual input only is 'hard mode,'" he said. While open-minded about SIMA 2's potential for robotics, Togelius noted that "the real world is both harder and easier than video games."
