How AI Is Transforming Military Event Prediction and Thermal Facial Recognition

21CTO Digest: On January 4, the U.S. Army employed artificial‑intelligence technology to predict major global events.

Artificial intelligence is accelerating efficiency and capability across logistics, daily life, and the battlefield, prompting militaries worldwide to speed up AI research.

U.S. military drones combined with AI technology

On January 4, 2019, DARPA launched an AI model called Kairos, aimed at event prediction. DARPA, the Defense Advanced Research Projects Agency, develops high‑tech solutions for military use.

The concept of event prediction traces back to Jean Piaget in 1923, describing how contextual and temporal information can be combined to infer future events and generate corresponding AI‑driven charts.

For example, a person leaving home on a Saturday to shop might later attend a football match; DARPA is less interested in such benign scenarios and focuses on detecting criminal pathways.

Consider a potentially dangerous individual who first posts heated criticisms on social media, then calms down, and later engages in unusual activities (large purchases, travel) before an attack. The AI model aims to detect such patterns and flag high‑risk actors despite time gaps.

However, the approach has limits: it cannot predict everything, such as the onset of popular trends or large‑scale protests like the recent French demonstration against fuel taxes.

It also struggles with long‑term analysis, being better suited for short‑term forecasts; predicting a country’s economic collapse a year ahead is beyond its current data scope.

AI technology that can recognize faces on walls

AI technology increased the U.S. Army's learning speed by 13×

The U.S. Army is also researching a machine‑learning method to identify faces captured by thermal‑imaging cameras, enabling identification in complete darkness.

ARL scientists Benjamin S. Riggan, Nathaniel J. Short, and Shuowen Hu released a white paper detailing how to apply facial‑recognition techniques to thermal images.

The main challenge is matching captured thermal faces with a database of traditional visible‑light facial images.

These devices are mounted on Apache helicopters, armored vehicles, and other platforms, allowing infrared sensors to detect low‑visibility targets.

Deploying cameras with AI lets the system process images without relying on conventional lenses; thermal images are fed to a neural network, which returns both the presence of a person and their identity.

The technology is still emerging but rapidly maturing; by aligning thermal facial data with existing databases, real‑time battlefield identification of allies versus adversaries becomes feasible.

Beyond facial recognition, the Army employs numerous machine‑learning algorithms for adaptive learning and computation in resource‑constrained environments.

Nevertheless, technology alone does not replace effective management.

Apple’s global facial‑recognition implementation, widely accepted in China, demonstrates how such technology can enhance security and privacy, reducing account theft and financial fraud.

Compiled by: Li Zhiliang