I think the point is, only a human should: 1) authorize the kill, 2) choose the target, and 3) specify the limit of acceptable collateral damage. The missile'

I think the point is, only a human should: 1) authorize the kill, 2) choose the target, and 3) specify the limit of acceptable collateral damage. The missile's AI should never decide 1, 2, or 3, but only how to execute its orders in compliance with the standing rules of engagement.