[narrator]

No fully automated slaughterbots. There has to be a man in the loop that at least clicks a mouse every time the robot kills an enemy combatant. No one should ever say: "Oh yeah, we had a bug in the auto-kill algo and our killer robot shot a whole bunch of innocent civilians. Whoops! That's nobody's fault. Nobody's responsible. The algo did it all by itself!"

[blotter_paper]

In the context of a military that accepts collateral damage as par for the course, I don't really see the distinction. Did anybody but Manning go to prison over that Collateral Murder video? We have humans approve airstrikes that we know will kill unidentified civilians when they're close to an identified target. We have humans approve airstrikes that target unidentified individuals which we probabilistically identify as combatants based on movement and cellphone metadata. What difference does it make if a human pulls the trigger --err, presses the button?

[merpnderp]

I'm completely with you, but eventually "we" won't have a choice when adversaries go full autonomous.