I probably need to read the report to get more information, but I'm just having trouble visualizing this hypothetical future where we need a swarm of autonomous

I probably need to read the report to get more information, but I'm just having trouble visualizing this hypothetical future where we need a swarm of autonomous agents to identify their own targets and threats at all level without any human intervention, and where any manual control would be too limiting. There are circumstances, like CIWS or CRAM systems automatically shooting down incoming munitions, where that comes in handy. I'd also be more comfortable with putting autonomy in the hand of anti-aircraft systems (either SAMs or UAVs). In that case you can program in parameters and I can trust a machine to reliably stick to those parameters. But if we're talking about a machine gun mounted on an unmanned ground vehicle? If you really need to shave off that second or two of human reaction time, that's not a real battle, that's a showdown from a Western movie. I don't think it's too much effort to send a quick picture to a human, or have a human being specify where they can engage and for how long for area suppression. I think it will be multiple decades before AI has progressed to the point where I'd trust it to correctly identify enemy soldiers vs non-combatants as well or better than a human being. Fortunately, I think it will be even longer before that's something we need.