Editorial illustration for DeepMind Recruits Robotics Expert to Build Universal AI for Humanoid and Non-Humanoid Robots
DeepMind's Universal AI Strategy Targets Next-Gen Robotics
Google DeepMind hires ex-Boston Dynamics CTO to create Gemini AI for any robot
Google DeepMind has poached the former CTO of Boston Dynamics. The mission? Build a single AI, a Gemini foundation, that can drop into any robot, humanoid or not.
“We want an AI system that works almost out-of-the-box across any body configuration,” the team says. That’s audacious. DeepMind’s Gemini is already a multimodal beast, processing text, images, audio, and video.
Now it’s being tuned to guide hardware through messy, unpredictable real-world scenes. The company’s robotics pedigree runs deep: RT-1 and RT-2 learned from human demos. Just last September came Gemini Robotics 1.5 and its ER sibling, pairing AI control with physical machines.
With humanoid fever sweeping the industry, DeepMind is betting big, CEO Demis Hassabis predicts a “major breakthrough” inside two years. And the competition? It’s already sprinting.
Google Deepmind is taking another major step toward turning its Gemini AI model into a universal robotics control platform.
The marriage of DeepMind’s Gemini architecture with Boston Dynamics’ mechanical mastery is a bet on universality. Not a robot for every task, but one AI for every robot, a mind that slips into any chassis, from a factory arm to a bipedal marvel. The hardware race is loud.
Wheels, legs, grippers, and actuators all clamor for attention. But the real prize isn’t the body; it’s the brain that animates it. Hassabis’s two-year horizon feels less like a prediction and more like a challenge.
If Gemini can truly learn to inhabit any form, it won’t just steer machines. It will redefine what a machine can become. The race isn’t just heating up.
It’s shifting lanes.
Related Reading
Browse all Google news →
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Databricks finds multi-step agents beat single-turn RAG by 21% to 38% on STaRK
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic NVIDIA and Google Cloud let developers scale AI from prototype to production
Market Trends
Related Topic Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI
From Archives NotebookLM Turns Complex Spreadsheets into Presentation Insights
From Archives Use Temporal Patterns: Plot Timestamps to Spot Seasonality, Trends, Shifts
From Archives Gemini app watermarks 20 billion AI images with SynthID, tests Detector
LLMs & Generative AI
From Archives Nano Banana Pro unveils Gemini 3 Pro Image, adding better text rendering
LLMs & Generative AI
Common Questions Answered
Who did DeepMind recruit to help develop their universal AI for robotics?
DeepMind hired Marc Raibert, the former chief technology officer from Boston Dynamics who is known for pioneering advanced robotic systems. Raibert's expertise is expected to be crucial in developing a more flexible AI platform for robotic applications.
What is DeepMind's goal for the Gemini AI system in robotics?
DeepMind aims to create a Gemini-based AI system that can work across different robot body configurations, including both humanoid and non-humanoid robots. The multimodal architecture of Gemini, which can process text, images, audio, and video, makes it particularly suited for guiding robots through complex environments.
How do DeepMind's previous robotics projects like RT-1 and RT-2 contribute to their current AI robotics strategy?
RT-1 and RT-2 were foundational AI models designed to help robots understand and interact with their environments. These previous projects have laid the groundwork for DeepMind's current ambition to develop a universal AI control system for robotic platforms.
