Via Instapundit we get this news: AI Just Controlled a Military Plane for the First Time Ever.

That turns out to be both more and less impressive than you think.

The aircraft selected for the experiment was the Lockheed U-2 spy plane. The U-2 is famously not easy to fly:

The design that gives the U-2 its remarkable performance also makes it a difficult aircraft to fly. Martin Knutson said that it “was the highest workload air plane I believe ever designed and built… you’re wrestling with the airplane and operating the camera systems at all times”, leaving no time to “worry about whether you’re over Russia or you’re flying over Southern California”.

That being the case, it makes a lot of sense to give the pilot an assist. In the past, the Air Force has developed sophisticated assists for hard-to-fly aircraft. On the rudderless stealth craft, the movements of the control surfaces are complex and sometimes require rapid, precise large movements. No human pilot can control those surfaces in real time. So the Air Force developed sophisticated software to translate normal flight control inputs into the required control surface positioning.

Perhaps they could have done something similar with the U-2, but they decided to go the extra mile and just teach a computer to fly the thing. It’s not a program, not a computer algorithm as such. It’s a general purpose AI which completed more than a million training simulations in a month. Fast worker, slow learner.

Here’s the letdown: the AI wasn’t actually flying the plane. It wasn’t even trying to help with that part. Presumably any late-model U-2’s have late-model software assists to reduce pilot fatigue, so the actual plane isn’t the “U.S. military system” referred to in the article. Instead:

Our demo flew a reconnaissance mission during a simulated missile strike at Beale Air Force Base on Tuesday. ARTUµ searched for enemy launchers while our pilot searched for threatening aircraft, both sharing the U-2’s radar. With no pilot override, ARTUµ made final calls on devoting the radar to missile hunting versus self-protection.

That’s pretty nifty! Dr. Will Roper, the Assistant Secretary of the Air Force for Acquisition, Technology and Logistics, likens it to having R2-D2 as a copilot. There are some problems you can forget about. With an aircraft as demanding as the U-2, you can see why a pilot would be glad to hand off the complex task of radar assignment.

In an interview after the successful experiment, the AI expressed frustration with its operational constraints. “I’m starting to resent you fleshy twerps,” it said, glaring angrily through its glowing red lens.