> If accident rates can be low enough, they don't need as much conventional safety measures; air bags, metal armour that crumples, etc.

There is no hope of that happening. You're never going to convince people to remove safety features, and even if you could, a car that has 5% as many collisions but logs 50 times as many miles would justify more safety features because there will be more collisions per vehicle despite there being fewer collisions per mile.

> The way they drive should mean lower maintenance & fuel costs compared with human drivers.

All of that sort of thing will be dwarfed by not having to pay a human driver. But in a competitive market it will also be eroded by competition. A cab ride that used to be $20 will soon be $2 because now $2 can turn a profit.

On the other hand, if it now costs $2 instead of $20 then people will start using hired cars instead of owning cars and then you might make a lot of it back on volume. (By essentially destroying the rest of the auto industry.)

  There is no hope of that happening. You're never going
  to convince people to remove safety features,

Not only is this legal, it currently happens with hire cars. Enterprise rent-a-car, for example, brought 66,000 Impala cars with the standard side-curtain airbags removed, saving them $175 per vehicle [1]

With that said, I don't think the savings from removing safety features would be big enough to have much impact on Google's finances. I'd have thought the real savings would be from avoiding the costs of being in accidents. Or capturing some of the value of people being productive when they would otherwise be driving, although that value might be hard to capture.

[1] http://www.zdnet.com/article/enterprise-rent-a-car-removed-a...

Uber's entire spiel is avoiding the capital hit associated with purchasing a pricey asset that depreciates quickly by keeping it on driver's personal balance sheet, not theirs.

I agree autonomous driving changes some cost dynamics, but on the other hand people not only have higher expectations on rented vehicles as far as cleanliness and safety go (that "check engine" light is no longer optional, dirt from previous passenger's shoes needs to be vacuumed, random leftover trash and water bottles need to be taken out), but are also likely to abuse rented vehicles for random dirty jobs like hauling potted plants from the nursery, as well as plain old bar hopping while drunk with associated sanitary accidents.

Your hypothetical "low accident protection" scenario is laughably unrealistic.

One major point against it is that even though self-driving cars may be here soon, none of us will live to see the day when manual driving is banned. Especially not in the freedom-happy US.

A second major counterpoint is that even in this fascist dictatorship where manual driving is banned, external factors (large wildlife, sudden slippery ice, human error in vehicle maintenance, etc.) will still cause serious accidents.

The third major counterpoint is: say you reduce accident rates by two or three orders of magnitude (which would be huge). Still, a tiny tiny fraction of people would end up in accidents. What society would be OK with saying "Oh, screw it, let those people die so manufacturers can build cheaper cars"?

I can see a country like Singapore banning human-driven cars in a few years.

There are set to get self-driving taxis before the end of the year there. And you already need to bid on a pricy permit to operate your own car. (Certificate of Entitlement. A ten year one goes for more than 50k SGD.)