Military AI operating in uncontested areas will offer value to the military while eliminating the risk of attacks, partial data, and misunderstanding contexts, among others.
Fremont, CA: Artificial intelligence (AI) is useful in image recognition, recommendation systems, and language translations. It can also understand the context of a text, objects, and multi-task.
AI can identify a T-90 main battle tank in a satellite image, recognize high-value targets in a crowd using facial recognition, translate text for open-source intelligence, and text generation for use in information operations.
With the military hoping to implement AI into its systems, there are specific issues that need to be addressed. Most AI systems are trained using data that are labeled by a human. Capturing these data and sharing it is difficult, particularly for organizations that want to classify data and restrict access to it. AI systems can also become large and slow, which results in being sensitive to dimensionality issues. Although AI systems work well in uncontested, controlled domains, it will lead to errors under adversarial conditions.
An AI’s image processing ability is not strong when it comes to images that are different from its training sets. Therefore, the model may fail to recognize the content correctly. The systems also struggle to define between correlation and causation. The military needs to be concerned with adversarial attacks as attempts will be made to fool image-recognition engines and sensors. Cyberattacks will try to elude intrusion-detection systems, and logistical systems will be provided with modified data to block the supply lines with false requirements.
However, AI has many application areas in the military to optimize productivity, lower user workload, and operate faster than humans. Military AI working in uncontested areas will offer value to the military while eliminating the risk of attacks, biased data, and misunderstanding contexts, among others.
Check out: Top Aerospace Technology Companies