I confess I don't understand why the focus is so strongly on the taxi-replacement market; it's interesting, sure, but largely unviable outside metropolitan areas. It can't serve people who go shopping and want to use their cars as lockers during the shop, or people who leave sports gear in the car while they work then go to the gym later, or a number of other cases. Who pays for them? It'd be an expensive business to set up and you'd have regular arguments about taxi licenses that'd make the fuss about Uber look tiny. Who'd clean them? You bet people would do unsavoury things inside them. No; it's a small, awkward to serve.
Fit a 'self-drive' button to my next mainstream car though, and you have a killer product. You can keep it just like your normal car whenever you want, but when you don't - just press the button, program the navigation and do what you want. That night I'd have slept. This evening I'd probably have read on my way home from work. The technical challenges are no greater and it's much easier to set up.
That's a killer app, not an autonomous taxi. Your car, with an invisible chauffeur on demand. The first manufacturer to properly pull it off won't be able to keep up with demand.