In aircraft, some versions can do full autoland, including zero visibility taxi to gate. That certification, called category III C, requires aircraft, crew and airport certification plus regular recurrent testing and recertification. It is VERY expensive so is used in very rare situations.
The other extreme in autopilots, analogous to Tesla autopilot beta, will fly straight and level, following navigation directions in the air but not for approaches.
The IIIC idea is much like Google cars today. It can only be used in pre-certified situations, with thick data on specific routes and is expensive to build, operate and maintain.
The straight and level analogy, for Tesla, is quite easy to use, saves stress and boredom, but NEVER can be trusted to act independently. In Silicon Valley terms, Beta.
Seems to me the aircraft analogy is apt. We should expect nothing close to full autonomy until/unless technology and processing logic have advanced far beyond their current state.
FWIW, I'm thrilled with my Tesla autopilot, and I'm not expecting major advances anytime soon. I am expecting significant advance making current functions work better, but nothing remotely close to IIIC without analogous high cost, maintenance and training intensity.