[malandrew]

Autonomous cars also has a regional learning curve that benefits from having many human drivers on the road in many cities. For a purely autonomous company to enter a new market, they need to map it and there is a learning curve to building out maps sufficient to provide all the possible routes in a city. An autonomous network that can only do some routes and is useless for other routes, is going to have an expensive time scaling as they will have to put human drivers behind the wheel until they know they can operate safely in that new market.

A company with many existing human drivers can easily enlist those drivers with autonomous-car quality mapping kits. The drivers are all already driving all possible routes, so mapping a new market is much easier and cheaper. Furthermore, a mixture of autonomous cars for routes that autonomous cars can operate safely and human-driven cars for routes that are still being learned guarantees that the service is useful for any route a rideshare customer might want to take.

Lastly, you won't be able to operate a self-driving network without great ops. You need people to maintain the vehicles and provide customer service. Building out a great ops org is yet another massive learning curve for a purely autonomous company.