Google DeepMind unveils its first “thinking” robotics AI

Google DeepMind launched its first "thinking" robotics AI system featuring two models that work together to control robots through complex, multi-step tasks. The Gemini Robotics-ER 1.5 model processes requests and visual information to create step-by-step instructions, while Gemini Robotics 1.5 executes these actions by "thinking" through each step before acting. Built on Gemini foundation models and fine-tuned for physical environments, the system can transfer skills between different robot types without custom programming. The action-controlling model remains limited to select testers, while the instruction-generating model is now available to developers through Google AI Studio.