AI is not autonomous driving. There can be a component to it, but in order to use AI for untethered tasks like driving, I'd need a LOT faster cellular network connections than current 5G standards provide. Of course, the required technological leaps are currently being tested. Implementation isn't imminent, but it does exist.
In a world where I can just miracle stuff up, I'm envisioning roadways imbued with sensors. The vehicles that travel on them would be purely autonomous, electric (it's far easier for me to dump electric energy into the effort than maintain IC vehicles), and probably not privately owned - although there's no reason why they couldn't be.
Right now, the issue with autonomous driving is all the other non-autonomous vehicles doing unpredictable stuff. If I eliminate those and use a pathway with identifiers, sensors, and regulators, I could deploy vehicles with even opaque "windows" (they could be video screens of outside if people really wanted them) that you could summon up via phone app. You'd tell it where you wanted to go, get in, and the grid would optimize the route at the highest speed possible. Urban traffic could max out at 50-70mph. Interstate travel could easily hit over 100mph. There's be no need for "safe following distance" since the grid would know the planned velocities of all objects. Each roadway could support many more vehicles by eliminating this space.
Yes, I know my proposal will never work. I can't design an extra wide vehicle with "Punisher" stickers that back into parking spaces, since the spaces won't be needed. It'd be psychologically damaging to exit such an autonomous vehicles wearing wraparound Oakleys.
