Google Gives Robots Full-Body Intelligence With Gemini Robotics 2

Google DeepMind has unveiled Gemini Robotics 2, a model that opens a new chapter in robotics. The technology lets robots move their whole body in a coordinated way—not only their arms. On-device local operation means they can work without an internet connection and without latency, which strengthens data security and makes instant decision-making easier.The new release is pushing the limits of VLA (Vision-Language-Action) models widely used in robotics. VLA combines what the robot sees from cameras, the commands it receives, and the physical motion it will perform in a single AI system. Challenges remain: for fine skills such as screwing in a light bulb, success rates are still around 36%. Experts explain this with Moravec's Paradox—the idea that physical tasks humans find easy, such as grasping objects and walking, are among the hardest for artificial intelligence.Opinions differ on robots entering homes and handling everyday chores. They are expected to spread quickly in industry and manufacturing, while high cost and safety concerns slow adoption of household robots. Google continues work on safer prototypes that can carry out complex physical tasks and stop automatically when …