Google DeepMind wants Gemini to power many different robots
The company’s push into physical AI is putting Gemini inside humanoids—and testing whether its intelligence can survive in the real world
Google DeepMind is developing an AI brain called Gemini that is designed to be portable across various robot bodies. The company's ambition is to create a form of "physical AI" that can inhabit multiple robot bodies, much like how chatbots can simulate sassiness. DeepMind's latest attempt, Gemini Robotics 2, can control whole-body movement in different robot models, from two-armed research platforms to humanoid robots like Apptronik's Apollo 2.
This technology allows the robots to perform tasks such as fetching snacks, changing lightbulbs, and tying knots. DeepMind's approach to achieving this is based on leveraging reinforcement learning, where the AI system is exposed to data and rewarded for the correct actions, enabling it to learn and generalize its skills across different contexts and robot bodies.
The goal is to bring the same level of adaptability and intelligence found in large language models to robotics, enabling robots to handle a wider range of tasks and interact with the physical world more effectively.
Written by urgent.news from Scientific American's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
