● Covered by 4 sources · 4 reportsMedium impact4 neutral
Google DeepMind's Gemini Robotics 2 Enables Whole-Body Control for Humanoid Robots
Key points
- Gemini Robotics 2 controls humanoid robots from feet to fingertips.
- Enables tasks like walking, crouching, manipulating objects, and cleaning.
- Supports improved dexterity for complex five-fingered hands.
- Includes new sub-models for embodied reasoning and on-device operation.
- Gemini Robotics ER 2 is publicly available for developers.
Enhanced Robotic Control
Google DeepMind has introduced Gemini Robotics 2, an updated version of its AI model designed to control humanoid robots. The previous model focused on upper-body control, but Gemini Robotics 2 now supports "whole-body motions," extending from a robot's feet to its fingertips. This allows robots to perform a broader array of actions, including walking, crouching, stretching, and manipulating objects.
Increased Dexterity and Complex Tasks
The new model improves dexterity, enabling control over more complex, five-fingered hands. This capability allows robots to execute delicate tasks such as putting tape into a boombox, screwing in a lightbulb, tying a garbage bag, or closing a storage bag. Videos demonstrate robots picking up watering cans and finding specific items on shelves, indicating progress towards more intricate real-world applications.
New Sub-Models and Developer Access
Gemini Robotics 2 integrates a trio of new sub-models to enhance its capabilities. These include a vision-language-action (VLA) model that converts visual and language inputs into motor control, and Gemini Robotics On-Device 2, a lightweight version that runs locally without internet connectivity. The embodied reasoning (ER) model, Gemini Robotics ER 2, allows robots to understand their surroundings and communicate, devising plans for multi-step tasks. Gemini Robotics ER 2 is now publicly available for developers through the Gemini Live API.
Towards Physical AGI
Google DeepMind views this update as a significant step towards creating a generalist robot, sometimes referred to as "physical AGI." The goal is for robots to perform any task a human could, moving beyond pre-programmed, narrow actions. While movement speed still needs advancement, the whole-body coordination enabled by Gemini Robotics 2 is considered crucial for completing complex, real-world tasks.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
How outlets covered it
Google DeepMind introduced Gemini Robotics 2, an intelligence layer with three new models designed to improve physical AI adaptability. This update provides robots with more dexterous, full-body control, enabling them to perform complex multi-step tasks.
Google DeepMind launched Gemini Robotics 2.0, enhancing robot capabilities for complex tasks, continuous environmental analysis, and collaboration through new sub-models. This update aims to advance generalist robots, allowing them to perform diverse actions and control humanoid robots with greater dexterity.
Google DeepMind announced Gemini Robotics 2, an updated AI model that enables humanoid robots to perform tasks requiring "intelligent whole-body control," such as cleaning and manipulating objects. This development advances robotic capabilities beyond arm-specific actions and represents a step towards physical Artificial General Intelligence (AGI).
Google DeepMind released Gemini Robotics 2, an updated AI model that enables humanoid robots to control their entire bodies, from feet to fingertips, and perform more complex dexterous tasks. This advancement allows robots to execute a wider range of actions and improves their ability to coordinate for real-world tasks.