This right here.

US IC already using pattern of life analysis for target selection, wouldn't surprise me if they decided to use ML with "95% accuracy" to do that.

Best outcome: really good slaughterbots become cheap and every country gets them. Then the battlefield is mostly slaughterbots slaughtering each other. But still, when one side wins...

I mean, you gotta understand, this whole "oh no, civilians died" thing is fairly new. Japan, Germany and Russia in WW2 wouldn't have minded slaughter bots that use visual and hear signature based human identification to enable slaughterbots that simply kill anything that lives until their battery runs out.

I hear solar is cheaper and more effective than ever, these days.

In case you want your slaughterbots to last longer, I mean.