The question is whether the problem is one of insufficient engineering optimization or whether it requires a step-function in technology that does not exist now

The question is whether the problem is one of insufficient engineering optimization or whether it requires a step-function in technology that does not exist now. It appears to me that self-driving cars are of the former type, while flying cars are the latter. Current-resolution lidar, cameras, and radar seem to provide sufficient sensor input. The costs are too high by a long shot, but that may just be a question of getting economies of scale established. Current PC graphics hardware has sufficient bandwidth to process those sensors. I don't think you can just throw current neural net training at the problem and get Type 5 autonomy out of it - there will be lots and lots of engineering hours in figuring out what to do with that sensor data - but that's just a problem of doing many man-years of straightforward work. Flying cars don't have adequate power from a current-gen internal combustion engine running on petroleum, and especially not enough power from lithium-ion batteries and electric motors. If you could get a power source that provided an order of magnitude or two greater power density than the best of those technologies, flying cars would be viable. Until then, no amount of engineering hours will make it work.