The issue with FSD is depth vs. width. MB decided to go deep, i.e. higher autonomy (L3) in limited use cases. Others decided to go wide, i.e. more use cases but lower autonomy. Tesla wants both, which is admirable, but it means that they will be in beta mode longer than the rest, while removing valuable driver assistance features.
In reference to the aerospace, autonomy there is a much easier problem to solve because of lower complexity and stronger regulation (FAA requires every aircraft to have a transponder, for example). Even then, we have pilots because autonomy has not reached the depth (complexity) and width (variety of cases) of a human. That makes me very skeptical about robotaxis in the near future, except in very limited cases - well defined boundaries, no inclement weather, pre-defined trajectory, etc. Unfortunately, the promises of FSD muddy the water.