A new robot, powered by Google DeepMind’s Gemini Robotics 2, has demonstrated remarkable capabilities by tying a knot, unscrewing a lightbulb, and sealing a bag of grapes, all without prior programming for these tasks. Gemini Robotics 2 enhances humanoid robots with whole-body control, enabling natural movements in various everyday scenarios and allowing for fluid interactions that extend beyond fixed operations to mobile environments.

Google DeepMind: Google DeepMind is Alphabet’s primary AI research laboratory focused on developing advanced artificial intelligence systems for complex problem-solving. In this development, it created the Gemini Robotics 2 model that powers a robot’s ability to perform intricate, unscripted physical tasks such as knot-tying and object manipulation through unified control.
Gemini Robotics 2: Gemini Robotics 2 is an AI model from Google DeepMind designed specifically for robotic control and whole-body intelligence applications. It enables a robot with 22 finger joints to handle diverse tasks like unscrewing lightbulbs and sealing bags using a single integrated system, as shown in recent demonstrations with humanoid platforms.

Integration: Gemini Robotics 2 enhances humanoid robots like Apollo 2 by providing whole-body control for natural movements in everyday scenarios.
Capabilities: The model allows fluid robotic interactions without task-specific programming, extending beyond fixed-station operations to mobile environments.