Just as in the case of the "Hyperloop": in theory, it is perfectly possible, but anyone who got as far to build a few meters of it (or something resembling it) quickly realized the practicality problems, and what would happen if you scaled those problems to a few hundred kilometers...
On the other hand, autonomous driving is not a solved problem, not even in theory. One could argue that it would require some sort of a generic AI, which we don't have, and nobody knows if we will in the future. So selling anything based on that is simply fraud, in my opinion.