I'm betting on drones, and it's not even close.
Self-driving research is focused squarely on predicting human behavior similar to Minority Report. That's a tall order. The raw computing power alone required to make it happen is at least 50-60 years away. If and when predicting human behavior is possible, we'll see it in the stock market before we see it in a car.
Imagine needing to account for a jaywalker, a person driving erratically, a paint can falling off the back of a truck, unpainted lines on the road and torrential rain. Now imagine accounting for all of those things happening at the same time. It's a study in chaos theory, not physics. Can it happen some day? I think the better question is: why does it need to happen in the first place? Self-driving tech is a band-aid fix that leverages the existing transportation infrastructure in this country. People in Venice travel by boat, people in America travel by car. If we had to solve the "personal transport" problem from scratch using near-term technology, how would we do it?
Human-sized drones. The technical challenges are far more surmountable. Imagine standing on a pad and ordering a drone to pick you up. Two minutes later, a Mini Cooper-sized drone drops from the sky like a meteor, lands in front of you, and opens its door. You hop in; the door closes; lift-off. Five minutes later, you're 20-miles across town getting dropped at the front door of your destination. It's the "flying cars" dream come true.
What's technically preventing this from happening today?
Air traffic control. Personal toy drones like the DJI Mavic are already fairly reliable GPS, but they're put to shame by aircraft-grade GPS technology. Everything from commercial jets to private Cessna 172s to military aircraft rely on GPS every day to travel safely through clouds and rain at speeds well above the speed of sound. What's more limiting is the human aspect of air traffic control. This is already changing; new rules for the FAA are being enacted that allow for more automated overall control, but these rules are in place because humans are involved. If there is an automated queueing system and exclusive "drone traffic lanes," the system would run itself.
Safety. What if a motor fails? The answer here lies in what happens in other forms of transportation. In a car, you pull over. In a commercial airliner, you've got built-in redundancy in the other engines. In a military jet, you eject. In a private plane, you pull a parachute. All of these options (save pulling over) is available to add to drones today. Add an extra motor to allow one to fail. Add an ejector seat, or a parachute in the event of multi-engine failure, or power loss. We've applied this tech in other areas for decades. Add to that the fact that these would be electric, not gas-powered, and the reliability goes up while maintenance requirements go down.
Acceptance in society. Collective trepidation about Tesla's autopilot is a good proxy for our reluctance to give control to a computer. But that's what will be necessary for drones to succeed. This is largely a marketing problem. Make drones incredibly convenient and people will trade almost all other freedoms for it.
Energy technology. To be able to fly a two to four person payload for just a few minutes requires a massive battery. A battery so large, the drone wouldn't be able to leave the ground. UAVs have been around for a while. Rocket launches, which are much more complicated than a DJI Phantom, have been fully automated since the Apollo program. But there's a reason we don't have a battery-powered rocket: Newton's third law.
We need a few more turns of the battery tech wheel for drones to reach a capacity where it's viable as a transport platform. This is a real limiter of today, but it will be solved within our lifetime.
While we're busy waiting for a self-driving car, automated drones could be whisking us to wherever we need to go in less time.