The Pentagon Is Testing AI That Can Predict Conflict 5 Days in Advance

A few years ago, reporters listened to a U.S. Air Force general describe something that sounded strangely familiar on a humid afternoon inside the Pentagon briefing room. It’s familiar in the sense of science fiction films, but not in the military sense. On the surface, the concept seemed straightforward: artificial intelligence that could identify conflict indicators days before it actually starts.

The project is part of a set of military tests known as GIDE, or Global Information Dominance Experiments. The idea is surprisingly useful. Algorithms analyze massive streams of data, including satellites, radar, intelligence reports, and even subtle indicators like vehicle movements near military facilities, rather than depending on analysts who spend hours staring at satellite images.

It’s difficult not to think of the movie Minority Report while you watch the explanation play out. However, Pentagon officials maintain that’s not the main point. They claim that the system cannot read the future or predict crimes. It’s identifying patterns that people might miss. Nevertheless, the concept seems uncomfortably near.

The system reportedly detected anomalous activity near a fictitious strategic location during one of the demonstrations. There were no missiles firing or airplanes scurrying, so the signal wasn’t spectacular. Just minor details: more cars in a parking lot, an odd logistics movement, and indications that equipment was being set up. It was all meaningless on its own. But when combined and processed at machine speed, the system saw it as a military maneuver preparation. And it did so days before it could have been discovered through conventional intelligence analysis.

The head of U.S. Northern Command and NORAD, General Glen VanHerck, gave a remarkably straightforward description of the capability. He clarified that the objective is to change military strategy from responding to threats to foreseeing them. Commanders could take action sooner rather than later by preparing defenses, deploying forces, or sending diplomatic signals.

This strategy has a subtle logic. You suddenly have choices if you know something might happen in five days. However, there are a number of unknowns hidden behind the optimism.

The system depends on combining enormous amounts of data from all over the world. Cloud servers receive images from satellites that orbit hundreds of kilometers above the planet. Movements across oceans are monitored by radar stations. Sensors keep an eye on communications networks and airspace. A part is also played by commercial data sources, some of which are routine, such as shipping traffic.

After that, the AI looks for correlations, patterns, and anomalies. It may now only take minutes to do what once required dozens of intelligence analysts.

However, it has always been difficult to forecast human behavior, especially in geopolitics. Numerous intelligence systems throughout history appeared plausible until they weren’t.

A disturbing historical parallel is mentioned subtly by some analysts. RYaN, a program run by Soviet intelligence during the Cold War, used dozens of indicators to try and predict a nuclear strike by the United States. The system issued frightening alerts that almost forced Soviet leaders to make disastrous choices.

Of course, modern technology is much more sophisticated. Machine learning systems are able to identify patterns in datasets that are too big for humans to handle. The Pentagon is interested because of that promise. However, skepticism also starts to seep in at this point.

Although geopolitics doesn’t always follow historical norms, algorithms trained on historical data might be able to identify patterns. A erroneous forecast could be caused by an unexpected military drill, a sudden change in politics, or a misread signal.

Furthermore, inaccurate forecasts are not innocuous in military settings. Pentagon officials appear to be cognizant of this tension. They stress time and time again that people are still in charge of making decisions. They claim that rather than giving orders, the AI system offers analysis and warnings.

Another question is whether that distinction holds true over time. Rival nations are vying with one another globally to incorporate artificial intelligence into their military plans. China has made significant investments in defense systems powered by AI. In its command networks, Russia has experimented with automated decision-making tools. Predictive intelligence technologies are being tested in even smaller countries.

The Pentagon’s endeavor is therefore a component of a larger change. Algorithms are processing battlefield data more quickly than any human team could, making warfare more and more data-driven.

The underlying signals are frequently so unremarkable. When machines piece together these small bits of activity, which were previously lost in vast amounts of intelligence data, they take on significance.

The true power is found in that stitching process. Observing this technological change from a distance gives the impression that military thinking is changing in a small but meaningful way. Intelligence spent decades looking for threats. Forecasting them is now the focus.

It’s still unclear if machines can accurately predict conflict. National politics, human decision-making, and unexpected crises are rarely predictable datasets.

However, it is evident that the Pentagon thinks the attempt is worthwhile. After all, an algorithm that can provide commanders with even a few days’ notice could alter the start of wars or possibly stop them completely.