As recently as March, Daimler promised to put 10,000 autonomous taxis on the streets by 2021. This week, however, Daimler Chairman Ola Kallenius announced that the company was making a "reality check" on the project and focusing on self-driving long-haul trucks instead.
It's fine that self-driving cabs aren't coming as fast as some expected -- and it's even better that Silicon Valley-style big talk appears to be going out of fashion.
Kallenius's "reality check" has some solid business reasons: Daimler is cutting costs and can't commit to a large, capital-intensive project without a clear idea of what kind of first-mover advantage it might confer. But mostly, it comes because of a long-obvious technical problem.
Making sure self-driving cars aren't a menace in city traffic is a job that will take more than a couple of years. Investigators are still trying to get to the bottom of the March 2018 accident in which a driverless Uber killed a pedestrian in Tempe, Arizona, and it appears Uber's cars had been involved in dozens of previous nonfatal incidents in the course of the same testing program.
No one wants to be in the same situation as Uber so General Motors subsidiary Cruise won't be launching self-driving taxis in San Francisco this year, as previously promised, and maybe not next year, either.
There have been lots of news stories about Waymo, an Alphabet subsidiary, launching a self-driving taxi service in Arizona, and in April, it even put an app for it on the Google Play store.
But in September, Morgan Stanley lowered Waymo's valuation because of delays in the commercial use of its technology, and last month Waymo CEO John Krafcik said driverless delivery trucks could come before a taxi service.
For European carmakers, which have to deal with older cities not laid out on a grid, launching autonomous taxi services appears even more daunting than for Americans. They know it's a long way from Tempe to Amsterdam or Rome. That's one reason Volkswagen, a latecomer to self-driving development, isn't worried about being overtaken.
Alexander Hitzinger, chief executive of Volkswagen's autonomous-vehicle subsidiary, said in a recent interview that even an industry pioneer such as Waymo was "a long way away from commercializing the technology" and that Volkswagen's autonomous vehicles would be developed by the mid-2020s.
That time frame may be no more realistic than the previous hype about big 2019 and 2020 launches. Autonomous car developers can complain all they want about unpredictable human drivers and pedestrians who are causing all the accidents with their flawlessly superhuman creations, but nobody is going to clear the cities of people to give self-driving cars a spotless safety record.
And making sure that, after millions of hours of training, artificial intelligence is finally able to drive like a human after a few hundred hours on the road, is not all that's required for robotaxis to be viable. There's still the whole matter of figuring out how to reduce rather than increase urban congestion by using cars that don't "think" like humans.
It's also dangerous to adopt any kind of specific framework for the launch of automated truck services, even though that's an easier project than taxis because the routes are fixed. The presence of humans in what is still a predominantly human world has rather unpredictable consequences for robot behavior. And the first movers have an obvious disadvantage: Like Uber with a taxi, they can get burned in ways that could set the whole business back years, and the earnings potential is unclear.