A committee of tech insiders say AI-controlled weapons will avoid harming civilians.

I can imagine that an AI-controlled weapon would do a better job of following its rules of engagement. Thus, whether it would harm civilians less, or harm civilians more, depends on what those rules of engagement say.

Would an AI-controlled weapon be better at saying, "We shouldn't be fighting a war here"? Would it be even occasionally capable of asking that? Would it be capable of asking, "Sergeant, do you think we're the redcoats?"

I seem to recall reading that the helicopter crew that stars in the collateral murder video killed in accord with the letter of the rules of engagement; that they took advantage of loopholes in it to justify some of their killings.

Some AI programs are capable of such reasoning. Would these weapons do that?