It's not just about the overall fatality rate per mile. Self-driving cars also need to not have any accidents that humans would have easily avoided. Until they demonstrate that, they're not going to be widely accepted.

Lidar sees a 3D cloud around the car with almost a cm accuracy. Better sensors could have most definitely prevented the accident.

However, to add to tesla's point. I have a car with adaptive cruise control. Not having to constantly check distance with car head has probably saved me once or twice from a crash.

Good auto-pilot's are definitely going to reduce human deaths on roads. Not only just deaths, but the amount of traffic all of us have to endure because someone just didn't keep the right distance from the car infront and bumped into them.

I'm all for open smart driving technology.

I see so many people (like the one two replies above) push the argument for the latter using the advantages that even the first would give.

Why? If they're overall much safer, then that seems like a net win to me—even if their failure modes are not identical to human ones.

There will, of course, be exceptions, as we've already seen (poor guy).

What are you comparing? a 70,000$ car with better safety standards to a 10 year old car! A car that's aimed to be completely autonomous (with driver assistance features) compared with a car completely human driven?

Can you not think beyond extremes? Here's a simple example. An always-on "Autopilot" (or any other driver assistance system) that only intervenes if the driver does a tragic mistake. We still have the important general intelligence of a human and the faster and reliable response of a machine.