This is all math here. If a computer is statistically better at driving than a human you can ask the same question as "Would you give up control to decrease the chance of an accident? Or would you risk your kid's life just to feel in control?"
>What happens when the first car has a BOSD and the backup human driver isn't paying any attention and the accident happens within seconds of the crash?
People will probably die. The the same thing that happens when a human crash a car which is already happening everyday.
You know they actually have a physical car? That car was able to drive itself from Mountain View to Tahoe.
As of August 2012, it has logged 300,000 miles.
http://www.theatlantic.com/technology/archive/2012/08/google...
But given the conditionals of my first sentence, the conversion may not start in earnest for a decade or more.
We do, in many places - your car is one such place. Another is every single hospital in the civilized world. You know, there's a continuum of world between HA datacenters and your smartphone.
The technology pundits are ignoring the human factor, giving up control, trusting your kid's lives to a computer, the same technology that crashes on the desktop, and on our smartphones. What happens when the first car has a BOSD and the backup human driver isn't paying any attention and the accident happens within seconds of the crash? These things will likely kill a bunch of people and will go the same way that the C-5 or the Segway did.
Maybe I'm too cynical, but until we have HA computers that exist outside of a datacenter, I doubt a proof-of-concept by Google will go anywhere in this or the next decade, let alone disrupt anything. The idealism and flying cars optimism is well placed and I'd love to see it happen, but the future nearly always takes much longer to happen than people predict...