Uber has announced the official start of self-driving tests in Pittsburgh. Uber has been running their lab for over a year, and had various vehicles out there mapping and gathering data, but their new vehicle is sleeker and loaded with sensors - more than on Google's cars or most of the other research cars I have seen. You can see several lidars on the roof and bumpers, and a seriously big array of cameras and other sensors.

In addition, recently it was announced that the GM-Lyft-Cruise combination will be offering rides in 2017 in a self-driving Chevy Bolt. Of course, there will be a safety driver in the car supervising it so it would be an empty taxi coming to pick you up, but it's a nice step.

These two announcements bring attention to two of the most important companies in the space, even though their technical efforts are much less mature than Google's or Daimler's. That's because of one key forecast that I have emphasized from the start:

A large fraction of the automotive industry is going to switch to be about selling rides, not selling cars

As we all know, Uber has already become the #1 brand in the world in selling rides in just a few years. It's a very important position to have. Lyft has #2 but other companies like Didi own China (and just got a $1B investment from Apple.)

As the owner of the ride brand, Uber has a lot of control. The brand of the car that drives you is less important and interchangeable. But that's not the only advantage these ride companies have:

• Ride companies have huge volumes of drivers on the road all day. They can be used as a resource for mapping, testing and verifying self-driving systems. Companies like Google had to pay staff and buy cars to do that.
• Ride companies can combine human driven ride service with robotic taxis, to take you from anywhere to anywhere any time. It just costs more if you want to travel where the robots can't.
• Uber and Lyft can fail in their research program and still win. They just have to find somebody else to sell them the cars. Of course that does mean a power trade -- it's very nice to own the magic sauce that makes it all work, but the ride companies are among the few would could have another provider and still have a lot of control.

At the same time, Lyft is now bound to probably work with GM, and Didi possibly with Apple, which leaves Uber with more flexibility among these.

The ride companies are already doing big experiments in real ride-sharing, ie. multiple independent passengers in the same car. Today, using UberPool is popular and can save significant money. A more interesting question arises when robotic taxi service is available for 30 cents/mile. I don't think people would share their ride to reduce the price from $1.50 to $1. Saving 50 cents does not move the needle for most people of even moderate income levels.

How will ride companies compete?

An important social question asks how many ride companies can compete in a market? Right now Uber has established a lot of dominance. In San Francisco, birthplace of Uber, Lyft and Sidecar, Sidecar shut its doors from difficulty competing with the other two. Is there room for only a few companies? That's bad, because competition is good for the public.

The first intuition is that fleet size is a big competitive advantage because you can offer faster pickup times and more choices of vehicle. Customers will care a lot about how long they have to wait for a ride. That will vary of course based on random positions of vehicles, and also how good the predictive positioning is in the fleet management system.

At the same time, it is possible to have a successful limo company today with just one limo. You only do scheduled rides (or ad-hoc rides booked via networks like UberBlack) but you have a business. It is not the size of your fleet that fully governs your wait time, but rather the ratio of the size of your fleet to the number of customers you have. Lyft has a smaller fleet but also fewer users, so I find it can often match or beat Uber on wait time, though neither wins all the time. There is a natural balance here -- the better your fleet-size/user-base ratio is, the shorter wait times you have, but that brings you more customers until the advantage starts reducing.

I'll have more analysis of this in the future.