AI will soon replace humans for most ATC, due in part to the high cost of affirmative action government clock punchers with power tripping bureaucratic attitude. Current AI can already handle the speech part, and could learn the rest of the trade in 4 hours. This can be rolled out to small, uncontrolled airports first. It’s even possible to have a second layer of peer to peer AI powered ATC for backup.
Why would you want to put AI control at an uncontrolled field, which by definition has no ATC controller? Auto-ATIS and reported traffic like some FBOs do on unicoms?
I am very skeptical that AI is the solution. Just off the top of my head, how about planes taxiing in fog and communication errors. Where do you go to get that dataset of experience to pour in in four hours?
AI might be able to have helped with the fog related collision scenario and the miscommunication scenario (Houston story) iff you had more sensors (such as the ground radar), and technical improvement to pilot-ATC communication. (all instructions echoed on a screen to eliminated “what was that”, “what did he say”.) and all the sensors and comm gear always was working.
Aviation is only so safe as it has been in America because it was a lifelong calling and discipline. Trying to make it a great job for people who didn’t feel the calling is the mistake. Humans make the most reliable system under real world failure scenarios. Rickover demonstrated that.
I predict getting rid of the human controller will really require getting rid of the human pilot as we know it.
The big reason for AI is distributed and diluted responsibility, opaque algorithm/rules, and plausible deniability. Secondary reasons, easier and cheaper for a reduced level of quality. And those are the features, not the bugs.