From a reader:
Dear Jonah,
All due respect to our shared fascination with tech-porn, robot soldiers can’t possibly be cost-effective. Airborne drones work because airborne targeting (when it even works) is as expensive as drone technology. On the ground, an $18.2M robosoldier will never be a match for an $18 Louisville Slugger (or a free 60-pound rock dropped from a fifth floor window.) Grotesque as it may be, human cannon fodder is simply far cheaper and far more cost effective. How many enemies would such a robot have to kill before it is taken out in order to be considered a good value? Far more than it could conceivably destroy, I’ll bet.
Me: I’m not so sure. Twenty years ago, if I said that computer chips would be so inexpensive they’d appear in Happy Meal toys and cheap greeting cards, people would’ve laughed. I have no idea what robosoldiers would look like or how much they’d cost 50 years from now, but the economic story of the last 50 years suggests that amazing technology gets cheaper and cheaper and (American) human lives get more expensive. The insurance policies on soldiers alone are amazingly costly. Add-in training, lifetime health benefits and the political costs of casualties and it becomes less than obvious that a particularly lethal robosoldier (which might simply be a pencil-sized floating death dealer of some kind for all I know) will be more expensive than a flesh-and-blood one.