What's a robber going to take from an Uber driver at gunpoint? Obviously not the cash he's earned from the rides, so... his personal wallet and the Uber smartphone? Most drivers have a wallet and a phone; why are Uber drivers a substantially better target than any other driver out there?

An average driver does not stop to pick up passengers ever n amount of time. A Taxi/Uber driver does. Speak to some taxi drivers and ask them how many times they've been robbed. Its pretty common. Not necessarily at gun point either. They get a lot of knives pulled on them.

An Uber driver doesn't have the cash from fares, wheras a taxi driver does, though.

They do have valuables. Their own cash, maybe tips, the cellphone, etc. Robbers go for the easy targets.

The drivers dont have more than any average driver... in fact they usually are just average everyday drivers. The pickup and payment is done entirely through the mobile app and now includes pictures of both drivers and riders.

I really would say it's far safer than a taxi. Someone planning on robbing someone probably wouldn't use a GPS capable device to first have the potential crime scene come to them with continuous tracking the whole time, only to steal what they could get from basically anyone on the street.

they don't get tips in cash. They seem marginally more likely to be robbed than other drivers.

I bet Uber can make autonomous taxis happen. They operate in urban areas. Uber can create the very detailed maps of these areas that autonomous vehicles need.

Right now, we have two conditions where autonomous systems work well: on the highway, where you can use products like Delphi's sensor package to do adaptive cruise control and lane assist, and in more urban environments with very detailed maps (see the Google autonomous cars).

I actually think there's a good chance Uber can create very detailed maps of places like Manhattan and San Francisco, and make autonomous vehicles that can use these maps to drive around.

Uber is under such a microscope when it comes to drivers that it makes a lot of sense for them to sink the R&D into this problem. All the cost going into paying people, paying to find people, and paying to clean up when people make a mistake, goes away.

Map data is not the main issue. I'm talking from experience here. Its the sensors and software. Good sensors are über expensive (mind the pun), not easily integrated in vehicle control systems, and currently not awfully precise (for the kind of duty daily automated driving requires). The software just needs more time to cook.

One thing people forget is that building cars is hard and expensive. Autonomous vehicles is not only about mounting some sensors and computers. Its about interacting with a slew of systems that were not designed for autonomy. A fully autonomous car that can drive people around like a Taxi requires a complete engineering overhaul from the wheels and up.