Satellite autonomously finds targets using AI

An Earth observation satellite has successfully identified targets autonomously using an onboard vision-language model (VLM), a significant first. This advancement, powered by Google DeepMind's Gemma 3 and NASA JPL's software, allows satellites to process data in orbit, reducing the need for extensive ground analysis. The technology enables natural language queries to direct the satellite, potentially leading to more efficient data triage and new applications like continuous monitoring. This milestone paves the way for more sophisticated AI infrastructure in space, enhancing the capabilities and value of space-based sensors.