The vast majority of aircraft crashes are a direct result of PILOT ERROR.
It's more complex than that.
I am also of the opinion that this, (the "certainty" of pilotless, commercial airline flight), is the dead-horse..., but the arguments 'for' pilotless flight
must be engaged and examined and not dismissed. History is unkind to such dismissals, but it also forgets when one is right because by then it is ordinary.
While the industry has largely solved original, direct causes of accidents from the 30's on to the mid-80's which were primarily mechanical, system and/or engine failure, weather, navigational error, cockpit/instrument design, CFIT, ATC, mid-air collision and airports & runway design, human factors have understandably been a very difficult nut to crack, particularly post-automation beginning in the mid-80's. When one adopts, "the mind of the machine", one is changed, and in ways that are by their nature, invisible to one. That is what has occurred here, in aviation. That neither necessitates nor obviates pilotless flight. Any examination must be detailed and as much philosophically-conceived as technically and psychologically-conceived.
There is both a logical and a practical inconsistency in argument for pilotless flight that cannot be got around.
No system of human conception is "Spock-like"; any human system will always and already exhibit human limitations, if not in the cockpit crew then from those on the ground. Human factors accidents exist everywhere, not just in the cockpit.
There is no magic, no opportunity for perfection in pilotless flight. There is however, billions to be saved in salaried employees, training, checking, simulators and all that goes with managing a workforce including liabilities and insurance, and that is an entirely separate discussion. History shows us however, that most times, there are "opportunity costs" and outcomes that can remain unforeseen if one is too convinced, that can balance any such changes in the foundations.
Such design and implementation obviously is improved upon through collective work such that the problem of "missing something" in the attempted anticipation of every problem that has/does/will occur in commercial air transportation may be minimized - that is the principle of aircraft manufacture today and is obvious. But such factor cannot be eliminated, and can exceed the human limitations for a working complexity; there will always be certain threshold which will always be slightly less than the very best solutions from the very best minds. At some point, the outcomes will be at the level of committee work instead of insightful, really workable solutions. Bear in mind that aviation has been and continues to be designed for "average", simply because there are very few naturals or "top guns" - cockpit, system and airfoil design must be manageable by "average", whether such solutions are in software protections or technical design. "Average" here, means the Mean of the small set of all airline pilots,
not of all pilots, period. Despite recent accidents, it's my opinion that that "average" remains very high indeed in our cockpits - so far.
This is not an argument for or against pilotless flight, but it is an argument against improvements beyond a
10^-9 probability of occurrences of the kind with which we are presently familiar. Whether present in the cockpit or resident on the ground either in an engineering lab or in a drone-like compound, the human factor within complex, potentially-rapidly-changing, high-risk enterprises is essentially the same. What keeps commercial aviation safe is the adoption of a very great distance built between what is operationally acceptable in commercial aviation and what is sometimes necessary in military operations. That factor alone may mitigate against pilotless flight, along with the necessary computing power and equivalent, flawless, (much higher than
10^-9 capability), network reliability.
The "solution" then must submit not to technical argument but in-service arguments and the economics and even the politics of pilotless flight.
In this, I could offer that these advances, (and I accept them as very possible, but not probable) will not take a just few years because of their acceptance by designers, regulatory authorities and even the flying public. Nor is it "simple" to take "all the data" of events and occurrences from which a cognizant digital system would/could "decide
anticipatorily", (parse all occurrences in the database then parse the future, as airline pilots routinely do every day), based upon programmed "experience" of all such factors and provide a guaranteed-successful outcome of say, QF32 as described in an earlier post.
Nor can one just enter into the discourse a remark that computers will soon mimic human thought and will, as such, be better than pilots. In aviation, even imagination must have cranes, not sky-hooks, otherwise the argument is merely interesting, but not convincing.
It is a fact that most accidents now are the result of a network of causal pathways largely though not exclusively related to human factors. It is also an obvious fact that pilotless flight is ready now, today, for individual or select missions and tasks. The key is in how congenial may the marriage be between these two key components.
Certainly the discussion must remain light on its feet, for technological change is exponential. But the cost of such overall capability as described above for what could be a small reduction in human-factors-fatal-accidents may outweigh the cost of the present system "as-is".