Gemini Robotics is Google DeepMind’s vision‑language‑action approach that brings the Gemini model into the physical world. The system pairs perception with planning and tool use so robots can follow natural‑language commands, perform multi‑step tasks, and adapt to new environments with fewer demonstrations. Recent updates focus on safer operation around people, better manipulation, and running more capability directly on device. While it’s still early, the direction suggests robots that generalize across platforms and improve continuously from experience.
Sources: Google DeepMind official videos and materials.