When language models acquired a body problem

Google DeepMind’s Gemini Robotics announcement marked a shift from describing the physical world to acting within it. The system was presented as a vision-language-action model, with a companion model for embodied reasoning and robotics programming.

The announcement’s own terms reveal the difficulty. A useful physical model must generalize to new objects and instructions, react to changing surroundings and manipulate objects with precision. None of those properties can be reduced to a language benchmark.

Embodied autonomy is consequently a systems story. The model, robot, sensors, worksite, operator and recovery procedure share responsibility. A demonstration can establish a capability claim under stated conditions. It cannot by itself establish dependable work in an uncontrolled environment. The body problem is the integration problem.