Two weeks in a driverless taxi: the robo-car works – but it breaks the traffic rules

What it’s about

Since August 2023, San Francisco has been the first major city in the world where commercial robotaxi services run without a safety driver. Waymo, the Google subsidiary, and Cruise, a General Motors offshoot, received an approval that seemed unthinkable just years ago. For this self-experiment I let these cars chauffeur me around the city for two weeks. Twelve rides, plenty of insights. The most important: it works. But it is also unsettling.

How it feels

“Hello, Stephan,” the robotaxi greets me. The display shows the name and destination. The driver’s seat is empty, the steering wheel turns as if by ghostly hand. After two rides you get used to it; after five you enjoy it. The cars drive cautiously, keep distance, brake earlier rather than later. They behave like a nervous student driver in their first month – and that makes them predictable in dense San Francisco traffic.

Where it grates

The exceptions are the interesting part. My Waymo once crossed a solid line to avoid a roadwork sign. Another time it misread a one-way street. Both situations in which a human driver would, by reflex, have decided differently. The systems learn, but they aren’t error-free. The question is whether society wants a system that causes fewer accidents than humans, but different ones.

The bigger story

Behind Waymo sits one of those Silicon Valley bets the Valley is famous for. Sebastian Thrun, the German who built Google’s X lab, started the autonomous-driving research early. From that lab came Waymo, plus projects on internet balloons, wind-power kites and quantum computing. Not all of them worked. The robotaxi, however, seems to be arriving on the market 14 years later. What that means for cities like Berlin or Munich depends less on the technology than on the regulation.

I wrote the full piece for Handelsblatt.