Artificial intelligence

Meet the AI system that can control a space satellite

This news comes to us from Wuhan University in China, a group of researchers have allowed an AI system to take control of a satellite in Earth orbit for 24 hours, with the aim of exploring how an AI would behave when operating an object in space, without human intervention.

The result of the experiment has been published on the website of the journal Geomatics and Information Science of Wuhan University and has been defined as a milestone in space exploration.

The AI selected several locations on the planet on its own, instructing the Qimingxing 1 satellite to take a close look at those locations, including the city of Patna, in northeastern India and home to the Indian Army's Bihar Regiment, which in 2020 clashed with China's army in the Galwan Valley along the disputed border. The AI also selected Osaka, one of Japan's busiest ports that receives U.S. Navy ships operating in the Pacific.

China has to its credit more than 260 remote sensing satellites in orbit, some operating dormant in space with a limited lifespan, the researchers surmise that allowing an AI to control a satellite can improve efficiency in detecting foreign objects or activities and warning relevant institutions.

To improve its effectiveness, such a system requires a deep understanding of the world, not only in terms of recognizing natural and man-made objects, it must also encompass the complex and changing connections between them and the many human communities.

The AI decision-making process is extremely complex; a variety of factors, such as real-time conditions, camera angles, target value, and satellite mobility limits, must be considered when planning daily work.

While this experiment creates a variety of attractive possibilities for the future of space exploration, it must be asked, what is the limit of optimization offered by AI systems on space resources, and how far a machine can be allowed to decide actions alone in space, no doubt the possibility is very large, but we must not leave aside, that the compiled data must have its proper use.

16 de Mayo, 2023