A satellite just learned to find things on its own — here’s what that means
📰 ArticleTim Fernholz
First vision-language model deployed on a satellite autonomously identifies targets in orbit.
For the first time, a vision-language model (VLM) has been deployed and operated on an orbiting satellite, enabling autonomous target identification. The mission, announced by TechCrunch, involved Loft Orbital's YAM-9 spacecraft, which carries an Nvidia Jetson Orin AGX GPU. NASA JPL's NAVI-Orbital software packaged Google DeepMind's Gemma 3, a lightweight VLM designed for edge applications. The satellite responded to natural language queries such as 'find areas where natural environment meets human development' and 'identify infrastructure around railway hubs' by analyzing sensor data in real time and returning relevant imagery. This capability drastically reduces the volume of raw data that must be downlinked to Earth for analysis, as traditional satellites rely on ground-based processing. Loft Orbital plans to scale to 50–100 such satellites for global real-time coverage. The achievement also paves the way for larger AI infrastructure in space and potential applications like lunar or Martian exploration assistants.
💡Highlights
- ├─First VLM in orbit (Gemma 3)
- ├─Autonomous natural language queries
- └─Nvidia Jetson Orin AGX on YAM-9
🎯For
- ├─Space industry engineers
- ├─AI researchers
- └─Satellite operators