AI Targeting, Used in US Airstrikes, Is Just The Beginning

Retired Air Force General Jack Shanahan predicts that it will be five years before the US military feels comfortable enough to start using AI engines to recommend the ideal weapon or best order in which to shoot enemy targets during battlefield conflicts.

He says it will take time not only to develop algorithmic systems that could figure out the most effective way to run a war, but also to become confident about using them.

“If it is used in an automated mode and something goes horribly wrong, people will not use it for a long time,” he told me this week.

Shanahan was the first director of Project Maven, the Pentagon’s big push since 2017 to bring AI to the battlefield. Over the last few months I’ve looked into what happened in the years since, for this Businessweek story.

...

Shanahan told me the Centcom confirmation is the “most forward-leaning statement that I’ve seen in the past couple of years” about Maven and the use of AI in combat. He thinks that reflects a broader trend – that military use of AI is slowly being normalized.

“There’s a recognition more and more that these capabilities are out there and available and making a difference,” he said, adding people are getting more comfortable with machine learning systems, which are often opaque and need continuous tweaking and oversight. Confidence will come from experimentation in exercises and improvements in parallel with real operations, he argues.

Read the full story and more from Bloomberg.