Google has unveiled the second version of its Gemini Robotics platform, promising significant improvements in robotic dexterity and safety. Robots powered by Gemini AI models can now perform more complex tasks, analyze changing environments in real-time, and interact with other robots.
The ultimate goal of this technology is to achieve “Physical AGI” (Artificial General Intelligence), meaning a robot that can replicate any task a human can do. The new version includes the Gemini Robotics ER 2 model, which has made notable progress in embodied reasoning, allowing developers to have precise control over humanoid robots, even in complex areas such as hands.