latentbrief
Back to news
Launch2w ago

Robots Learn Dexterously With Gemini Robotics 2

DeepMind Safety, Search Engine Journal1 min brief

In brief

  • Gemini Robotics 2, a cutting-edge AI model, is teaching robots to master complex tasks with whole-body intelligence.
  • Unlike traditional robots that rely on pre-programmed instructions, this new system allows robots to adapt and learn dynamically in unpredictable environments.
  • For instance, it enables humanoid robots to perform intricate movements like walking, crouching, and cleaning a cluttered room-all while collaborating seamlessly with other robots to complete tasks faster.
  • The innovation lies in three key models: - Gemini Robotics 2 controls full humanoids and other advanced robotic arms, offering precise dexterity.
  • - Gemini ER 2 enhances communication between robots and humans, allowing them to understand their surroundings and plan multi-step tasks efficiently.
  • - Gemini On-Device 2 optimizes performance by running locally on robots, enabling quick adaptation to new environments within just a few hours.
    • This breakthrough brings us closer to robots that can operate independently in diverse settings, from homes to industrial sites.
  • As the technology evolves, we can expect even more versatile and autonomous robotic systems in the future.

Terms in this brief

Gemini Robotics 2
An advanced AI model designed to teach robots complex tasks using whole-body intelligence. Unlike traditional robots that follow pre-programmed instructions, Gemini Robotics 2 allows robots to adapt and learn dynamically in unpredictable environments, enabling them to perform intricate movements and collaborate with other robots efficiently.
Gemini ER 2
Enhances communication between robots and humans, allowing them to understand their surroundings and plan multi-step tasks effectively. This model improves the ability of robots to work alongside humans by enhancing their understanding and task planning capabilities.
Gemini On-Device 2
Optimizes robot performance by running locally on the device, enabling quick adaptation to new environments within just a few hours. This model allows robots to adapt swiftly to different settings, improving their operational efficiency and versatility.

Read full story at DeepMind Safety, Search Engine Journal

More briefs