Google's Gemini AI Controls Entire Humanoid Robot Bodies

Google DeepMind has announced that its Gemini Robotics AI model's latest version can "control entire humanoid robots." Previously, the model focused on controlling a humanoid robot's upper body, but Gemini Robotics 2 now supports "whole-body motions" from feet to fingertips, the company stated on Thursday.

The new model will enable humanoid robots to perform a wider range of actions, allowing them to walk, crouch, stretch, and manipulate objects. Videos shared by Google demonstrate how Apptronik's Apollo 2 robot can bend over to pick up a watering can, as well as find and take specific items off a shelf.

Gemini Robotics 2 provides humanoid robots with more versatile capabilities than before by combining visual perception and motor control. The model uses data robots receive from their environment and converts it into direct robot commands, enabling surprisingly flexible and precise actions.

According to Google DeepMind, this advancement is a significant step toward more autonomous and versatile robots, which could be utilized in sectors like industry or logistics. While not all models can yet perform all actions, the development trajectory is clear towards more precise and comprehensive robot control.