The problem with that argument is that the current weak spot is enemy jamming capabilities, and most especially if satellites (easily-jammed low-power antennae) are relied on as relays. Predator UCAVs function over Pakistan only as a result of tacit agreements made at ministerial level - they would go deaf over seriously contested airspace even if they could evade radar. 

 

Until there is a swarm system of nodal comms relays (made up of dedicated UAV nodes etc) then UCAV airpower can quickly be reduced to nil, we are still a long way away from taking humans out of the engagement process i.e. you hardly want your "deaf and dumb" UCAV to use "AI" to make decisions that could cause a major conflict. 

 

And how do you solve the problem of GPS jamming? Terrain Recognition? Star Sighting? The point about having an intelligent observer at the platform level is that he/she is able to rely on limited communications and commands, to replicate that 10,000km away you need to provide the same level of sensory input that the pilot on location enjoys, given the relative weaknesses (bandwidth, resilience, redundancy) of those communications relays that currently exist there needs to be an enormous amount of equivalent development in those communications systems before you can comfortably assert that manned platforms are past their sell-by-date.

 

Not to say it won't happen to some degree, of course it will, but there's enormous barriers to overcome that often get overlooked.

 

R