Google and NASA test an AI robot for deep-space support

Google and NASA are testing an AI-powered robot for logistical and medical-support tasks in space, including when communication with Earth is delayed. In an early laboratory test, Google says the system found and moved a medical bag successfully 85% of the time.

Watch Desk analysis

What happened

The project brings together NASA’s Human Research Program, Johnson Space Center Robotics Team and Jet Propulsion Laboratory with Google Public Sector, according to The AI Insider’s report. The robot is intended to help astronauts retrieve equipment and medical devices, while leaving decisions to the crew.

The reported test used an open-source, four-billion-parameter Pi0 vision-language-action model. Google says it was fine-tuned in about 2.5 hours and can run locally on compact edge hardware. The team also used Google DeepMind’s MuJoCo simulator to train and test movements before trying them on physical hardware. The report describes an early proof of concept at Johnson Space Center, not a deployed spacecraft system.

Why it matters

A Mars mission could face round-trip communication delays of up to 44 minutes, according to Google. A robot able to interpret instructions and act locally could therefore help crews handle routine tasks without waiting for Earth to respond. The reported 85% success rate is an early laboratory result, not evidence that the robot is ready for medical work in space.

Our read

This is a concrete test of AI moving beyond screens and into a demanding robotics setting. The useful detail is not just that a model can guide a robot, but that the team is testing local operation under the weight and power constraints of a future mission. That is promising engineering, with plenty still between a lab demonstration and a dependable crewmate.

What to watch

• Whether later tests improve on the reported 85% success rate.
• Which edge devices can run the model within spacecraft power and weight limits.
• How the team tests reliability and astronaut oversight before any mission use.

Discussion spark: For deep-space missions, should an AI robot be allowed to handle routine medical-support tasks on its own, or should every action require an astronaut’s approval?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.