Particle.news

Google DeepMind Enhances Robots with Advanced Gemini AI

New capabilities enable robots to navigate and perform tasks using extended context from videos and human instructions.

Overview

  • DeepMind's robots use Gemini 1.5 Pro to understand and navigate real-world environments.
  • The AI model's long context window allows for more complex task execution and multi-step instructions.
  • Robots are trained by watching video tours and absorbing extensive environmental details.
  • Tests show a 90% success rate in following over 50 different user instructions in a controlled environment.
  • Potential applications include home assistance, healthcare, and various service industries.