The Costs and Benefits of Artificial Intelligence

Naval forces keep watch on international shipping lanes from the Red Sea to the Arabian Gulf with a widespread network of floating, flying and underwater sensors.

These valuable eyes on the water, which hunt for smugglers, terrorists and other criminals, collect millions of images. Sorting such a vast volume of data tests the skills of Soldiers and Sailors. 

A new partner on the battlefield — artificial intelligence (AI) — is providing the sorting and sifting power to overcome what would otherwise be analytical overload.

Depending on parameters set by their human programmers, AI systems can winnow the millions of images into a few hundred that represent potential maritime threats. That allows militaries to use ships and aircraft more effectively and efficiently. 

“We’re able to move at speeds that were previously unimaginable,” said U.S. Navy Vice Adm. Brad Cooper, deputy commander of U.S. Central Command. 

Like commercially available electrical power generation in the 19th century, AI is a general-purpose technology that can be harnessed for potentially thousands of applications. The ability to sift large amounts of data faster than the human brain is among the military benefits of AI systems.

But regional and global adversaries enlist AI as well, most prominently in disinformation campaigns designed to destabilize societies and weaken their resolve to defend themselves. Manipulated audio and video, some using synthesized voices and images, are becoming harder to distinguish from the real thing. 

The disruptive power of AI as it relates to global strategic competition was the theme of a conference at the University of South Florida in the United States in March 2024. Sponsored by the Global and National Security Institute, U.S. military commanders, AI experts and academics gathered to discuss AI’s role in international security.

Advantages of AI

The goal of AI — the ability of computers to learn based on fresh data and experiences — is smarter decision-making for its human users.
In that regard, it has also found uses beyond defensive applications such as naval operations in the Red Sea. Its ability to sift through reams of intelligence reports, for example, can help militaries select targets more precisely. 

Experts agreed on criteria to assess AI’s usefulness for militaries. Does it save time?
Does it make a Soldier’s job safer? Does it provide better results? 

“That is the only measure of success,” Schuyler Moore, chief technology officer at
U.S. Central Command, noted at the conference. 

Military leaders cautioned that unleashing AI to operate autonomously is a mistake. To avoid tragic errors, human operators must remain in the loop. AI still doesn’t make complex decisions. In that sense, it’s a dumb technology, only as good as the people who programmed it. Its usefulness is more for quantifying data, less so for evaluating the quality of that data. When it comes to ballistic missile defense, an AI miscalculation could have disastrous consequences. 

Reliance on a flood of AI data during battles could lead to hasty, reactive decisions when patient strategic thinking is required, said Dr. Peter Bovet Emanuel of the Swedish Defense University’s Centre for Special Operations Research.

AI models are useful tools to predict adversaries’ intentions, but adversaries often try to operate outside these models in a game of competitive one-upmanship, so military officers stress the need for human intuition to detect when AI systems provide faulty or misleading results.

AI disinformation

AI’s ability to synthetize language, images and audio can have large repercussions for national and international security. Some countries have tried to further their aggressive aims by simulating the images and voices of rival military commanders to weaken resistance among opposing troops. 

On an even larger scale, AI-generated cognitive campaigns featuring floods of fake information could destroy a country’s will to defend itself before the shooting has even begun. With advances in technology, fake news articles created with AI are often nearly impossible to detect. Presidents of nations have been among the victims.

“We are going to see a lot more examples of this,” said Dr. Todd Helmus, a behavioral scientist who has worked with the U.S. military to study militants and radicals in the Middle East and South and Central Asia. 

Policing malign uses of AI is difficult. Most AI innovations in image and audio manipulation come from content creators who may not be linked to any national government.

In the wrong hands, artificially created media could become “weapons of digital mass destabilization,” said Dr. Vilma Luoma-aho, a Finnish expert in the field of AI manipulation. 

Helmus described a world in which unsavory adversaries churn out unlimited streams of false content with automated AI. The U.S. and its partners need to contradict the worst of this fake information, but there are limits to that type of defense. Experts urged countries to focus on teaching students critical thinking skills. That would allow individuals to assess the inevitable streams of computer-generated fakery. 

“Detection is not going to save us,” Helmus said. 

Predictive power of AI

Another strong military use of AI is war-gaming, predicting threats and how to counter them. AI’s ability to collect and sort data in every language allows military planners to create simulations and scenarios in a fraction of the time it once took. 

AI helps war-gamers forecast distant threats and prioritize them in order of importance. “Five or six months is too long to design scenarios,” said Walter Kulzy, a senior operations research scientist at Johns Hopkins University in the U.S.

Here are steps in AI-assisted war-gaming:

Collect information that challenges the status quo. These “signals” and “disrupters” are clustered into trends.

Generate stories based on these trends and forecast alternative outcomes. 

Project future impacts tailored to the needs of whatever organization is seeking the forecasts.

Assess courses of actions in response to trends and threats.

Ensure informed decision-making, whether it be designing war-gaming scenarios, acquiring new weapons and defensive systems, or enlarging or reducing military forces. 

As with other uses of AI in the security field, a system is only as good as its human oversight. A good AI system will be evidence-based, allowing military planners to trace information back to its source to check for reliability. Such a system should also generate sufficient scenarios to mimic the complexity of the world.

“Humans are more important than hardware,” said Gen. Bryan Fenton, commander of U.S. Special Operations Command.