" Whats sad is that we have to train pilots to do their jobs. "
In one sense that's exactly the problem, but in another that's a very complex problem. Most of us do the job such that the accident/serious incident remains extremely low. That doesn't mean there aren't close-calls, it just means very few end up badly.
Though largely in the background, the reversal of the role of automation has been considered but not largely explored.
The role of automation (in general) was established by circumstance after WWII during the time when "automatic" household appliances began making their presence (and benefits) known and understood by all.
The role of automation in flight is an extension of the concept and installation of autopilots. So it was natural that extensions of automation such as autothrottles and FMCs, (this latter in the early 80's, initially by Sperry) should extend to control of complex aircraft performance and navigation capabilities.
The notion that this was not a natural way to proceed because humans make terrible monitors of ordinary, interminably non-stimulating processes, permitted autoflight systems to be "the" way that both the airlines and the manufacturers naturally chose to go.
The attraction for airlines was a reduction in crews from four, (captain, co-pilot, F/E & Nav), to three then quickly, two pilots. In fact, I submit that this is the main (really, the only true) impetus behind fully-autonomous commercial flight.
However, what about reversing the role of automation from active to passive?
That means, crews would employ normal automation just as is done today but would have the added role of monitoring (really, auditing) manual flight.
Much needs to be considered of course. Automated protections already do much of this, but in my view, they typically engage long after the fact in terms of what a pilot would do when realizing that something was starting to come off the rails, so to speak.
I've seen discussions on this elsewhere and know there are both technical and philosophical limitations to the notion. However, role-reversal for automation seem a logical alternative to considering fully-autonomous, regularly-scheduled commercial flight.