Waymo is in 4 cities and nobody crashed

Waymo now operates in Phoenix, San Francisco, Los Angeles, and Austin. Millions of autonomous miles driven. Tens of thousands of paid rides per week. Zero fatal crashes attributable to the autonomous driving system.

Zero.

I keep turning that number over in my head. Not because zero is surprising for a single company in a single year. But because the comparison is devastating. Human drivers in the United States killed approximately 40,000 people last year. That’s 109 per day. About 5 per hour. In the time it takes you to read this post, a human driver will probably kill someone.

Waymo’s safety data shows their vehicles are involved in significantly fewer injury-causing crashes per mile than human-driven vehicles. The numbers aren’t even close. On a per-mile basis, the autonomous cars are dramatically safer.

The inversion point

There’s a moral argument that we don’t talk about enough.

If autonomous vehicles are statistically safer than human drivers, and we have the data to prove it, at what point does choosing to drive yourself become the reckless option? At what point does the act of driving, of putting your imperfect human reflexes and attention span in control of a two-ton machine, become the thing we need to justify rather than the default?

We’re not there yet. Waymo operates in four cities. The weather conditions are limited. The operational domain is constrained. But the direction of the data is clear. And the direction matters more than the current position.

What four cities means

Four cities means different traffic patterns, different road infrastructure, different driving cultures. Phoenix drivers are different from San Francisco drivers. LA has its own particular brand of chaos. Austin is growing so fast the roads change between map updates.

The fact that the same system works across all four, with zero fatalities, suggests the underlying technology generalizes. It’s not a trick that only works in the Arizona suburbs. It handles hills, fog, construction zones, and the specific insanity of Californians changing lanes without signaling.

What I notice when I watch the cars

I’ve ridden in Waymos in San Francisco. The thing that strikes me every time is how boring it is. Not boring in a disappointing way. Boring in a “this is how a car should drive” way. It stays in its lane. It uses turn signals. It yields to pedestrians without the aggressive creep-forward that human drivers do. It stops fully at stop signs. Fully.

When’s the last time a human driver came to a complete stop at a stop sign? Be honest.

The Waymo drives like a responsible, patient, slightly conservative human driver who is never tired, never distracted, never angry, and never in a hurry. That’s the product. Not flashy. Reliable.

The politics of this

Self-driving cars are weirdly political. Some people see them as a threat to personal freedom (the government will control where you go). Some see them as a labor issue (what about taxi drivers and truck drivers). Some just don’t trust computers.

I understand all of these concerns. They’re valid. But 40,000 deaths a year from human driving is also a concern. It’s the concern that we’ve normalized because we’ve been living with it for a century. We don’t think of driving as dangerous because we all do it. We’ve absorbed the risk.

Waymo is demonstrating that the risk is optional.

Not today, not everywhere, but the trajectory is clear. The NHTSA data will keep accumulating. The gap between human driving safety and autonomous driving safety will keep widening. And at some point, probably within a decade, we’ll have to answer the uncomfortable question: is it ethical to let humans drive when machines do it better?

I don’t have an answer. But I think the question is coming faster than most people realize.

Four cities. Zero fatalities. The data is quiet and relentless.

Related thinking: