Google DeepMind unveils AI model that controls humanoid robots' entire body

Google DeepMind has announced the launch of a new version of its artificial intelligence model for robotics, capable of coordinating the full-body movements of a humanoid robot. According to The Verge, this update represents a significant leap from the previous generation, which was limited to controlling the upper body of the robot.

From hands to feet

The system, called Gemini Robotics 2, expands the control capabilities of humanoid robots so they can perform "whole-body movements," according to the company's announcement cited by The Verge. In practice, the algorithm no longer manages only the robot's arms and hands, but coordinates its entire locomotor system, including legs and fingers, enabling more natural and complex movements.

In a demonstration published by Google, the Apollo 2 robot, developed by Apptronik, was filmed picking up a baseball glove from a shelf, illustrating the precision and fluidity of movement that the new model can generate. This practical demonstration shows how artificial intelligence can translate complex commands into coordinated physical actions—an area where progress has historically been slower compared to other AI applications.

A step toward more autonomous robots

The shift from partial to full control of a robot's body marks a crucial milestone in developing humanoid robots capable of operating autonomously in real-world environments such as warehouses, factories, or even homes. Robotics companies, including Apptronik, are increasingly partnering with AI developers to integrate advanced algorithms that allow robots to interact more naturally with objects and the space around them.

Although the full technical details of the Gemini Robotics 2 architecture have not been made entirely public, Google DeepMind's announcement points to a clear direction: the development of universal AI systems capable of operating across various robotic platforms, rather than being limited to specific models. This approach could accelerate the adoption of humanoid robots across a range of industries, from logistics to home assistance, as the technology becomes increasingly sophisticated and accessible.

It remains to be seen how quickly these capabilities will be integrated into commercial products and how their performance will compare with that of competing systems developed by companies such as Tesla and Boston Dynamics, both of which are active in the race to build functional humanoid robots.