Google DeepMind Unveils Gemini Robotics 2 for Whole-Body Control
Google DeepMind has launched Gemini Robotics 2, an artificial intelligence platform enabling humanoid robots to execute complex, whole-body movements and finer hand dexterity. The update expands beyond previous upper-body limitations, allowing machines to walk, crouch, and handle tasks like unscrewing lightbulbs and tying trash bags. Company officials state the advancement is a crucial step toward tackling intricate, real-world duties.
Key points
- Google DeepMind announced the release of Gemini Robotics 2 on Thursday.
- The new AI model upgrades previous upper-body restrictions to support whole-body motions including walking, crouching, and stretching.
- Demonstration videos show Apptronik's Apollo 2 robot picking up watering cans, screwing in lightbulbs, and sealing bags.
- Google DeepMind also updated its Gemini Robotics ER vision-language model to improve multi-step task handling over extended periods.
- Company officials highlighted that the technology relies on fully autonomous operation rather than human teleoperation.
Google DeepMind has introduced Gemini Robotics 2, an upgraded artificial intelligence platform designed to provide humanoid robots with comprehensive whole-body control. According to an announcement made on Thursday, the new software expands significantly upon the original system, which was largely confined to managing upper-body functions and robotic arms.
The latest iteration enables robots to execute synchronized movements from their feet to their fingertips. Promotional footage released by Google DeepMind showcases Apptronik's Apollo 2 robot performing complex physical actions, such as bending over to retrieve a watering can, removing items from a shelf, inserting a tape into a boombox, and tying up garbage bags. Furthermore, the platform introduces enhanced dexterity for five-fingered hands, allowing machines to tackle intricate tasks like unscrewing lightbulbs and sealing storage bags.
Alongside the motor control updates, Google DeepMind has revised the Gemini Robotics ER (embodied reasoning) vision-language model. This component aids robots in analyzing their environment, processing human instructions, and executing multi-step tasks over longer durations with a clearer understanding of task boundaries. Company representatives emphasize that the footage depicts fully autonomous operations, positioning the development as a foundational milestone toward enabling robots to perform complex, real-world duties.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.