Google unveils AI model to help robots mimic human movement

Google DeepMind, a subsidiary of Alphabet, has unveiled a new artificial intelligence model designed for robots that enables humanoid machines to coordinate their movements using their entire bodies. This marks the latest effort to expand AI applications, allowing robots to think, plan multi-step tasks, and adapt to human environments.

According to available information, the new system, called Gemini for Robots 2, allows robots to walk, bend, and handle objects while reasoning during tasks. The company explained that, unlike the previous version which primarily controlled the robot's upper body, the new model can direct the entire humanoid robot from head to toe.

In this context, Carolina Parada, Vice President of Robotics at Google DeepMind, said: "Our goal is to bring AI into the physical world, then build a layer of intelligence that can be used by all robots," according to Bloomberg.

Meanwhile, Kanishka Rao, Director of Robotics at the company, stated that "achieving a high level of motor dexterity remains a distant goal," noting that "robot movements are still slow and deliberate, as they need to pause to consider decisions that humans make automatically."

This announcement is a continuation of the planned launch of Gemini for Robots in 2025, a version of the company's main AI model that could translate linguistic and visual information into robotic commands. That launch revived Google's robotics ambitions, which date back more than a decade, after the company acquired several robotics startups in the early 2010s before later scaling back much of these activities, including shutting down its Everyday Robots unit in 2023.