About a year ago, I wrote this post saying that autonomous vehicles were already safer than human-driven ones. This claim was based on safety data from Waymo and about 22 million rider-only miles. (Rider-only means no human driver.) A year later, Waymo now has over 96 million rider-only miles across Los Angeles, San Francisco, Phoenix, and Austin (through to June 2025) — and their safety record is only becoming more compelling.

Here's an updated data set and a chart showing any-injury-reported crashes (average benchmark vs. Waymo):

What immediately stands out is that Waymo has virtually eliminated the most common and deadly type of crash: vehicle-to-vehicle crashes within intersections. Compared to the benchmark, they are down 96%. They've also reduced single-vehicle crashes by 96%, pedestrian crashes by 92%, motorcycle crashes by 89%, and cyclist crashes by 78%. These are all remarkable figures and evidence that we are solving one of society's greatest safety problems.

It also tells me that, as humans, our driving days are numbered. Pretty soon nobody will want us behind the wheel of a car. It will be viewed as too dangerous. And that's fine by me.

Waymo has just been granted approval to test its autonomous vehicles in New York City. The permit allows up to eight of the company's Jaguar SUVs to circulate in Manhattan and downtown Brooklyn. And according to the company, the plan is to start "immediately." This first approval only runs until the end of September, after which it will need to be extended — but I'm guessing that shouldn't be too difficult to obtain.

What's noteworthy about this announcement is that (1) New York City is a big and complex place and (2) it's the first city for Waymo that receives snow. The company currently operates in San Francisco, Austin, Phoenix, and Los Angeles.

That said, the company has been doing cold weather testing since, I think, 2012. And in 2016, they opened a 53,000-square-foot self-driving center in Michigan for this purpose. They've also run tests in Truckee, California, Upstate New York, and the Detroit area. So presumably its sensors are ready to melt snow and ice. But it's looking like the true test will be on the streets of New York.

Next should be Toronto.

Uber's stock has done exceptionally well this year. At the time of writing this post, it's up over 60% year-to-date. But at the same time, it remains unclear to me what the relationship will be between Uber and this brave new world of autonomous vehicles.

I mean, right now, if you're in Phoenix, I'm told you can order a Waymo car through Uber's app. But if you're in San Francisco, Waymo customers must use the Waymo app. It's all bit mirky right now, but Uber is just trying to put "as many cars on Uber's network as possible."

There's also an argument that, for the foreseeable future, ride-hailing networks are going to need some mixture of both human and robot drivers. I get this argument. But beyond the short term, I think there will be strong incentives to completely eliminate human drivers.

Last month, the New York Times announced that Uber is in talks with Travis Kalanick, the company's co-founder who got pushed out 8 years ago, to help him buy autonomous vehicle company Pony.ai.

It's a bit of an interesting story. Pony is a Chinese company, but because the US doesn't want Chinese tech to become too deeply embedded in the American economy — and has become increasingly hostile to such companies — it has been readying a clean US subsidiary of the business for sale.

This is what Travis allegedly wants to buy with the help of Uber. And it's particularly noteworthy because it could be an indication that Uber is worried about Waymo and wants to have its own AV unit (which it had previously, but then sold off in an effort to quickly reach profitability).

My sense is that Uber needs to do something along these lines. The risk of not having autonomous vehicle capabilities is simply too great.

Cover photo by Viktor Avdeev on Unsplash