Re ""American Exceptionalism" basically means we allow ourselves to commit war crimes with impunity."
The AI option will just add another layer of legal protection to the contractors and US government.
The "computer did it" seems to work wonders with new staff and staff retention.
Re "One of the things that has consistently mystified me about Americans' complacency with drone warfare is the underlying assumption that our current monopoly on drones is going to last forever."
The US still has the bandwidth to each drone to get the huge flow of image data back in real time and the international networks to encrypt commands.
Re "For all intents and purposes, we're already using killbots, and the really important point here is that airborne killbots can be used (for now) with impunity across borders."
A page from UK history during the
https://en.wikipedia.org/wiki/... and hopes AI drones will allow for the total grid like control over vast areas of the world? Worked well for the UK in the early 1900-30's back with total signals intelligence.....
The problem is the signal flow up and down so an AI will slow that command link tampering issue by not allowing any confusing new commands in while been totally aware of its mission, location and all local movement.
What is been sold on is another U2, D-21
https://en.wikipedia.org/wiki/... to try and out fly, out smart all other nations for decades.
The problem is how much "smarts" can the US afford to lose in a real crash or super computer projected blind SAM event?
The US mil has always been very hesitant to load up any device with too much technology, better to use a crew or sat control allowing the drone hold 1980's optics any nation can buy and a sat link back to push up all the collected data and a simple weapons system.
The loss of a commanded drone with its design, sat dish, optics and engine is no real loss. The recovery of the AI drone with its total encryption package is a loss that could be difficult to recover from.
Over time 'random' AI flight patterns could become super computer friendly and then its just the need for a big, fast SAM thats never been turned on in the past.
Its a race over decades, who can control a drone to get it to land anywhere or predict an AI flight path in time?
How much randomness can be designed into every unique AI drone for every mission approach, mission and then return?
Nations will use any networks other passive radar like system to look up and track.
What can a AI drone be fooled into expending all its limited payload on? Low cost, fancy emissions and optical fooling artwork in vast empty open areas lure in vast numbers of very complex AI drones?