[justinclift]

There are some obvious problem scenarios though, which are unfortunately likely to happen:

* The "no human in the loop" combat robot is deployed into an area with non-combatants, but wrongly classifies them as enemy combatants.

* The "no human in the loop" combat robot gets damaged while in active combat, forgetting the concept of "friendlies"

* The "no human in the loop" combat robot gets hacked. All sorts of ways this could go badly. If they're hacked en-masse that's potentially amplified dramatically.

[invalidname]

100% true and concerning. My main problem is that people treat this as a simple black and white "killer robot is evil" situation. It's not. It's far more nuanced.