Do you actually read the things you are writing? So a human pilot can monitor a degraded system making slow, large errors, but couldn't monitor a perfect system with "micrometer" accuracy making no errors.
It's amazing the faith people have in automated systems - and the "sci fi" ideas they have of their capabilities.
They don' t make slow, large errors.
They wait very long before acting and they act very slow.
The problem is that the system requires it to be like this, because the system is built with the philosophy that the machine is not reliable and that the pilot must remain in full control of the machine and be able to react when the machine goes wrong.
But if you don' t make it that way, the pilot has no perception of what the machine is doing.
That is for instance what happened with that B737-800 in AMS. Because the pilots came in too fast and that the A/T was supposed to be idling, they did not realise that the A/T had entered the autoland flare mode.
That philosophy is why you need double and triple A/P's for CAT3 approaches.
This philosophy comes from the nineties when most of today' s aircraft' s technologies were developped. These machines can be built today with such level of redundancy and low error level that they could fly themselves with immense precision without being monitored by the pilot.
However, with aircraft there are far too many variables. For totally automated flight without pilots, we would have to redesign just about everything in our aviation system. Firstly, taxiing from the gate to the runway is difficult. Either every plane would have to be totally automated and interconnected or none at all. Any plane that did not exactly follow the taxi instruction, messes your whole system up. Then there is evaluating issues like Rejected Take Offs etc. The only computer powerful enough is the brain. End of. To be able to compute information in that time requires excellent situational awareness and experience.
Interestingly, in the article I quoted it was mentionned that ATC would directly control the aircraft FMS. Taxiing from the gate is not an issue.
It' s not easy but you just need to position some lines on the ground and there are already proven systems (see Lexus, Citroen, Opel, etc...) that can optically detect lines and symbols.
Take-off situations can be monitored in operations centers where every aircraft' s systems can be monitored by several UAS pilots who can make the abort call in the event that the computer fails to detect or react to a critical situation (which is very unlikely).
There will be people needed to detect that flock of birds or traffic on the runway. On the other hand, in a fully autonomous and monitored system, accidental runway incursions caused by pilots will become a thing of the past.
If a pilot' s wage is an average $50K per year, the wage saving on each aircraft would be at least ten times or $500K per year, even if UAS pilots will be needed to monitor.
The savings on operations will be at the very least another $1 million per aircraft per year through cheaper insurance, lower operational cost due to increased flexibility in terms of cruise altitudes and climb/descent profiles, increased efficiency through improved aerodynamic design achievable from the removal of a cockpit and increased revenues because the square meters wasted by the cockpit will no longer be.
We' re going through the denial phase, but we must face reality.
How many people were not skeptical about replacing the flight engineer by an EICAS and before that how many people were not skeptical about replacing the navigator by an FMS?
15 to 20 years from now, we will be heading into that direction and 50 years from now we will be there.