I've been intimately involved with AP testing and development for automotive applications for some years now. I can safely say that it is a far more complex problem than automating airplane movements and flight - simply because aviation takes place in a controlled environment. There are no children running across a runway unexpectedly (nor would the autopilot need to tell the difference between said children and a blowing by paper bag). Further, there's very little one can do to avoid obstacles on a runway even as a human pilot. If that deer crosses when it wants to, you will hit it, you can't brake and swerve. There will be little anyone can do to prevent pilots from becoming obsolete - we do after all account for about 80% of accidents. It simply makes sense with the exception of the irrational fear we have about not having a fellow human being present in the same vehicle we are in - irrational is the keyword here.
The critical junction Airbus is now at deals with how to implement complete automation (level 5 autonomy) without making things LESS safe during implementation. Going single pilot would be an absolute facepalm decision - and I don't think for a second they really are considering this. What they will want to do is not add new variables. Leave all as is, but start relying on the pilots for backup only. It will take several years of fully automated flight to demonstrate that the autopilot handles all abnormals as well (or better) than a crew would. Regulators will follow, the rest is history.