I think you're overshooting with your response.

My statement had nothing to do with LLMs, or targeting Iranian schools. I was simply responding to the previous statement that a computer will fail you.

Computers don't fail people. They're precision instruments, and anything short of a cosmic ray bit-flip, or hardware failure, will not result in a failed execution of its instructions. Therefore the computer didn't fail YOU - YOU failed to provide adequate instructions for it to do what you wanted.

Even in your reply you specified that these neural nets are being given specific information to train target selection on. Sure, the AI is nowhere near the keyboard (well, it actually is, but more on that later), but then again, WHO fed the AI the training data? WHOSE decision resulted in the AI spitting out a school as a viable bombing target? Ultimately a human sits at the top of the chain, even if it was multiple automated AI systems that collated, labelled, sorted and managed the training dataset. The reason why those highly advanced neural nets didn't work the way they were expected to (even accounting for the NN black box effect), is, ultimately, down to human failure.

Mind you by failure here I am simply talking about expected outcome vs what happened. The expected outcome being the AI providing 100% viable combat targets, active combatants, military bases, etc., not a school full of children. Because while the cheetoh in chief might be a raging narcissistic child rapist bastard, I doubt most of the rest of the US armed forces agrees with bombing children, so the original goal of the AI had to be to provide said viable targets. Therefore there had to be a human component that provided the data to skew the targeting, and that human component was most definitely sitting in a chair in front of a keyboard...