I could see a trim package, but if autopilot hardware is only a few thousand dollars, it would behoove Mercedes to put it in all of its vehicles. At that point, any startup could build an interface to allow any Mercedes owner with the hardware to lease out their car through a P2P ridesharing platform.
As it turns out, going from sensors to working self-driving car is a pretty difficult task, and Uber has some of the best self-driving researchers in the world working for them.
Right, but Tesla has vehicles on the road, now, with a high level of autonomy (driving across the country autonomously 96 percent of the time: http://jalopnik.com/we-set-a-cross-country-record-in-a-telsa...) that are gathering massive amounts of data (40 million/miles/month) from vehicle owners as they drive.
Its an uphill climb for Uber. They will need to fund their vehicles; Tesla funds their battery manufacturing, vehicle manufacturing, and autopilot telemetry corpus buildouts using the vehicles owners have paid for.
1. I was responding to your Mercedes/Uber scenario; the comment I was replying to literally didn't mention Tesla (which fits more aptly to Apple's position in the desktop computing analogy).
2. Interstate driving autonomy and city driving autonomy are different beasts. Gathering a ton of data about task X isn't necessarily helpful for task Y; ML isn't magical. If anyone, Google is ahead of the pack in the market Uber is vying for (automated taxis).
Bye Uber.