Google DeepMind has unveiled Gemini Robotics 2, a new artificial intelligence model designed to control the full-body movements of humanoid robots from head to toe. Representing a major leap over previous generations in both gross motor skills and fine motor skills—including the use of five-fingered hands—the new system enables robots to perform complex physical tasks in dynamic environments.
New Features and Capabilities with Gemini Robotics 2
With the new model, humanoid robots can now perform basic movements such as walking, crouching, and stretching, alongside high-dexterity, precision tasks like sealing ziplock bags, tying up trash bags, or unscrewing lightbulbs. Initial footage shared by the company showed Apptronik’s Apollo 2 robot bending down to pick up a watering can and selecting and retrieving items from shelves. While company officials noted that movement speed is still under development, they emphasized that this capability level marks a critical threshold for real-world tasks.
The Gemini Robotics ER 2 model, which forms the system's cognitive and perceptual core, enables robots to analyze their surroundings, process verbal instructions, and manage long-horizon tasks. This update allows robots to better understand the start and end points of a task while facilitating collaborative operations among multiple, different types of robots.
Safety-Focused Improvements
Google DeepMind describes Gemini Robotics 2 as its safest robotics model to date, integrating advanced human detection features. When a human is detected nearby, the system triggers safety protocols that immediately switch the robot into a safe posture mode.
Industry Implications and Future Outlook
Google's transition to full-body control in AI-powered robotics opens a new era for hardware manufacturers operating in logistics, manufacturing, and domestic services. The advancement of fine motor skills and the standardization of multi-robot coordination have the potential to directly impact the costs and efficiency of industrial automation processes.
Frequently Asked Questions
Which robot models has Gemini Robotics 2 been tested with, and is it compatible with?
Initial visuals and tests shared by Google DeepMind featured Apptronik's Apollo 2 robot. The model is generally designed to control the full-body and hand movements of advanced humanoid hardware.
What practical advantage does multi-robot coordination offer?
Introduced with Gemini Robotics ER 2, this feature allows multiple robots of different types to simultaneously plan and execute a task together, optimizing the division of labor, particularly in large warehousing and manufacturing facilities.
*This report is based on data published by Webtekno — Artificial Intelligence.
💬 Comments
No comments yet. Be the first!
You must be logged in to comment.
🔑 Log In