Google's Gemini Robotics 2 Pushes AI Into the Physical World
In one line: Google DeepMind unveiled Gemini Robotics 2, its latest model built to control humanoid robots in the real world.
Key points
- Google DeepMind extended its Gemini family into physical robot control with Gemini Robotics 2.
- The company frames it as a jump toward so-called "physical AGI" — moving beyond text and images to manipulating real-world objects.
- Wired notes that placing AI into the physical world brings added safety and control risks.
Why it matters
When large models move from language and images into physical action, robots gain far broader general-purpose capability. But because a malfunction can translate directly into physical harm, the moment demands safety validation alongside the performance race.
Read more
How this story unfolded
- Google DeepMind Eyes July 17 for Gemini 3.5 Pro After Full Architecture Rebuild
- Agility Robotics to Become First Pure-Play Humanoid Robot Company on Public Markets
- Google DeepMind Brain Drain: Six Senior Researchers Exit in 10 Days, Alphabet Sheds $270B
- Google Loses Two More AI Researchers to Anthropic in Week-Long Talent Exodus