Pentagon believes it can predict events several days in advance using AI

Governing is planning and it is to react more quickly to a threat or to start discussions rather than waiting for the complications that the Pentagon is betting on artificial intelligence (AI). This is what emerges from a recent speech by General Glen D. VanHerck, head of the North American Aerospace Defense Command (Norad) and theU.S. Northern Command. He explains that the United States has completed the third test in a series mixing sensors civilians and military (satellites, communications, drones…), AI and cloud computing. According to him, this set would predict the turn of events several days in advance. The program called Global Information Dominance Experiments (Gide) gathered eleven military commands during the exercises.

The system is based on machine learning to process raw data collected by sensors (satellite, radar, communications, etc.) in real time. So far, nothing new, but where theartificial intelligence intervenes, it is on the identification of very small changes that would go unnoticed. This can be the start of a submarine, a parking lot that is unusually full of vehicles, small movements of troops that may seem insignificant, for example. The AI ​​can then predict if these “details” would be likely to present the beginnings of a threat a few days later and could launch an alert.

The future can be read in AI

This is the kind of work that would take hours, if not days, for analysts, since they have to process and sort through a lot of data before they can pinpoint a possible threat. Instead, with the Gide experiment, it only takes a few seconds to identify it and raise the alarm.

Finding signs of a possible threat is one thing, but knowing what is really going on is another. On this last point, AI shows its limits and humans remain strongly involved after anomalies were identified by the AI. It is the analysts who then take the lead. But for General VanHerck, AI is proving to be very useful in anticipating a lightning attack and prepare without wasting time for a defense operation, or even initiate diplomatic discussions instead of entering into conflict. The platform could be operational very soon for military certification in spring 2022.