US Army should ditch tanks for AI drones, says Eric Schmidt

But I don't think that's a necessarily terrible thing. When you talk about military procurement a bunch of pork-cutters love to piss and moan about $5,000 toilet seats and $500 machined spurred nuts made out of an alloy nobody uses. There's this general impression that government appropriation of these industries would bring the cost down, but looking at communist and socialist analogs this doesn't seem to be the case. Worse yet, every Boeing that America buys has to be bailed-out directly by it's owner instead of being allowed to die naturally in a free and competitive market. Forcing DARPA to design all this stuff instead redirects manpower from important tasks to instead optimize a price tag that might not even be realistic in the first place.

The most important part of deploying a novel technology in war is integrating it properly. Artillery doesn't work without targeting personnel and radio officers communicating what and where to strike. Submarines aren't scary unless command & control can coordinate targets and update the crew in the vessel. Drones and glide bombs rely on robust targeting information that must be maintained and updated to hit a mobile target like a tank. The US takes the time to figure this stuff out because they have the enviable position of accepting "No" for an answer from their advisors. If something is too dangerous, expensive or undeveloped to field in warfare, a rational army should not operationalize it.