Mountain View, California - Google DeepMind has unveiled Gemini Robotics, an advanced AI model designed for autonomous robots.
Gemini Robotics enables robots to operate without cloud reliance, using on-device processing for faster, more reliable systems. The model is built on Gemini 2.0 and incorporates vision-language-action (VLA) capabilities.
Key features include:
On-device processing to eliminate network latency.
Few-shot learning, allowing adaptation with 50-100 demonstrations.
Adaptability to various robots, including ALOHA, Franka FR3, and Apollo.
Google released the Gemini Robotics SDK through a selective program. This move highlights a strategic shift towards protecting its competitive advantage. The global market for industrial robot installations reached $16.5 billion, with "Physical AI" being a key trend.
The model's generative power extends beyond simple commands, enabling robots to perform new tasks. This positions Google in the competitive landscape of building the next generation of intelligent machines.