Friction losses aren't significant. These cars are limited by aerodynamic drag power -- this scales as speed^3, and dominates all other losses.
No.
"Of the energy output of fuel in a car engine, 33% is spent in exhaust, 29% in cooling and 38% in mechanical energy, of which friction losses account for 33% and air resistance for 5%. By comparison, an electric car has only half the friction loss of that of a car with a conventional internal combustion engine."
So, meh, even if magic self-driving cars had zero air resistance, they're only 5% more efficient than non-self-driving cars.
I mean future self-driving cars travelling at bullet-train speeds. For these, losses are dominated by drag.
[1] The economics DO NOT work this way yet, because batteries degrade too quickly (Tesla is what, 500 charge-discharge cycles?). Heavier use means proportionally shorter life; no advantage. Solve battery life, and you've solved economics.
I could reclaim about $1000 a week in billable time with a self-driving car (if I billed hourly, which I don't, but I think you get my point).
Social forces will drive the former, economic forces will drive the latter.
Also: the physics of high-speed cars makes the concept of cars traveling as fast as bullet trains totally ridiculous. The fact that a car is self-driving doesn't change the physics of tire/road friction.