yes for missions that does not change underway or where the objectives don't contain loosely defined objectives with moral ambiguity, and for the vast amount of

yes for missions that does not change underway or where the objectives don't contain loosely defined objectives with moral ambiguity, and for the vast amount of those you don't really need anything beyond and 80ies computer system running relatively conventional procedural code, which is why they have been performed using cruise missiles or artillery for decades. The problem is that the missions flown requires a lot more judgement then modern machine learning is capable of providing, or that we are likely to entrust solely to an computer anytime soon. Yes the skynet scenario is increasingly likely but it's also something military planners are aware of especially doing periods of diplomatic tensions like the ones we are under now where you really don't want an trigger happy automated defence system in charge of fully armed missiles. There is also the issue that all we know is that the next war will have it's own entirely new politically defined strategic constraints, where the most effective tactical strategy is might be blocked out due to long term diplomatic/political consequences.