Needing lane markings, or to follow the car in front, are a fairly primitive form of autonomous driving. The more advanced form is using may identifiable objects to place the car with precision on a 3D map.
Ah, no. Very much no. Where do you get this? Plenty of demos have already demonstrated what you describe. It's a good looking demo. But it is a dead end developmentally for level 5 automation. It can be a shortcut for level 4 on well described routes that don't change... i.e. Disney amusement park ride. Doing a demo for VC money. But in terms of ADAS development, any project that pursued this hits a brick wall and has to do all the real hard work of vision and path in a different way.