Breaking News

Google DeepMind pushes humanoid robots closer to the workplace

Google DeepMind and Boston Dynamics are moving humanoid robots beyond staged demos, training Atlas robots with Gemini AI to perform industrial tasks in real-world factory settings. The collaboration combines Boston Dynamics’ Atlas, a six-foot-tall electric humanoid, with DeepMind’s Gemini AI, giving the robots the ability to understand instructions, perceive their environment, and adapt to unforeseen situations.

The partnership is focused on practical deployments in factories, rather than demonstrations, enabling robots to handle complex, repetitive, and safety-critical industrial tasks. Hyundai, which owns Boston Dynamics, has been providing production units for training, with AI intelligence integrated through Gemini Robotics.

Gemini AI Enables Intelligent Action

Gemini Robotics merges vision, language, and action, allowing Atlas to interpret spoken commands, navigate cluttered spaces, manipulate objects, and plan multi-step processes autonomously. For instance, the robot can locate a requested item, avoid obstacles, sort parts, assemble vehicles, or clean factory areas without pre-programmed choreographed movements.

Powered by Gemini 2.0 multimodal AI, Atlas can simultaneously process video, speech, and sensor data, giving it human-like awareness for real-world tasks. Visual-language-action (VLA) models allow the robot to see its surroundings, understand commands, and act physically, learning from mistakes along the way. Safety features and rigorous simulation tests ensure Atlas can operate around humans and machinery without causing accidents.

Practical Deployment and the Road Ahead

The initial deployments target manufacturing, where Atlas receives live sensor feedback from the environment that informs Gemini AI, improving task execution over time. The combination of Boston Dynamics’ advanced hardware and DeepMind’s AI intelligence aims to address labour shortages, increase productivity, and enhance workplace safety.

This collaboration is also setting the stage for comparisons with other humanoid initiatives, such as Tesla’s Optimus robot, highlighting different approaches to building practical humanoid machines. Beyond factories, such AI-driven robots could eventually be applied to warehouses, medical facilities, or even residential tasks, though the rise of working humanoids also raises ethical and labour-related questions.

By integrating Gemini AI with Atlas, Google DeepMind and Boston Dynamics offer a glimpse of the near-future workplace, where humanoid robots may become regular partners in industrial operations as early as 2026.