Robotaxis: Waymo vs Cruise (vs Uber)
The world of autonomous driving isn't developing the way I expected! But it is making real progress in San Francisco.
Last week I spent a couple of days in San Francisco. I was there on business, but I decided it was time to get to know Waymo. I think everyone has kept track of autonomous vehicles (aka robotaxis) in San Francisco, just because of the press coverage. Quick summary: Cruise was first to get licensed to sell rides in cars with no drivers. Waymo was second. And both were allowed to sell those rides 24 hours a day within the city limits of San Francisco last August.
While it dabbled in autonomous vehicles, Uber ultimately withdrew from that race and went all in on cars with drivers (to transport both people and food). Now, as everyone likely knows, General Motors, parent of Cruise, stopped offering robotaxi rides in any of its regions in October, only two months after being allowed to operate 24 hours a day in San Francisco.
N.B. I was an investor in Cruise and made a significant amount of money from its sale to General Motors, like 50x my original investment in the seed round. 💰 Making money from a company makes me loyal to that company. So I could be accused of a conflict of interest in this case, but as you’ll see, I feel that Cruise betrayed me (and all of its customers) and am not feeling so loyal anymore.
In two days last week, I took six Waymo rides. I’d had one in a previous trip. And I had four trips in Cruise before they withdrew and before they were allowed to operate 24 hours a day. That makes me an expert in robotaxis! (I can’t count how many Ubers I’ve taken all over the world, but this is only incidentally about Uber, as the standard for comparison.) You can’t ride in Cruise anymore, at least right now. General Motors decided to take their AVs off the roads in San Francisco, Phoenix and Austin.
So I was already pissed at Cruise for having screwed up their approach to implementing robotaxis. When I invest in a company, I am all in. Cruise was started by one of the four co-founders of the company that became Twitch, which was sold to Amazon. That investment was “only” 20x, but it also spun out Socialcam, which was 32x. So I already had plenty of reason to care more about Cruise than any other robotaxi company, not to mention that Cruise was the first company to propose applying autonomous driving to ride-sharing, rather than personally owned cars. And the founder proposed doing robotaxis in San Francisco, precisely because it is one of the most difficult environments for computer intelligence to deal with. If it works in San Francisco, it will work anywhere, he said.
Well, Cruise didn’t work in San Francisco. And Waymo does. My 7 rides in Waymo persuaded me that Waymo does work in San Francisco: up and down hills; through six-way intersections; in and among crazy bicycle riders and inattentive pedestrians; in rain and fog. I am absolutely certain that a Waymo will kill someone someday (which Cruise did not do, but Uber did do in Phoenix) and probably run into a fire truck (which Cruise did do in San Francisco.) It’s just odds that these things will happen.
But the odds also are that robotaxis (and ultimately autonomous personal vehicles) will kill and maim fewer people than human drivers already do for every mile driven. More than 40,000 people die every year in traffic accidents in the U.S., for any reason. It’s inherently obvious that autonomous vehicles don’t drink and drive, and that they can be programmed to be careful in ways that humans cannot. But people will still die inside autonomous vehicles. The problem will be dealing with human reactions to computers being imperfect until there are no more human-driven vehicles, say in 2099. Meanwhile, it will scare the bejesus out of a lot of people that computers are driving them around.
For now, I can tell you that I will not hesitate to take a Waymo wherever it is operating. (So far just San Francisco, only on surface streets, and Phoenix, on both streets and highways, with plans for LA and Austin.)
What’s the difference between Cruise and Waymo? Apart from the publicity and competing story lines, I have to think that it is product management. Waymo has balanced the risk inherent in doing something new with the need for its customers to be reassured that it’s not risky. When I rode in a Cruise robotaxi, it had a cute name and cool looking graphics, which were clearly intended to reassure customers. But when you actually rode in the car, it would jerk around unexpectedly and sometimes stop or start unexpectedly. It was not reassuring.
In my experience in Waymos, the car always drove smoothly. In my first ride last week, it was raining heavily. I was somewhat apprehensive, particularly when I realized that the care wasn’t using its windshield wipers. I couldn’t see through the windshield. But the car didn’t act any differently; it was using LIDAR and all of its other sensors to see its surroundings as though it was broad daylight. This actually made complete sense, and I relaxed.
Other small bits of my experience helped me relax even more. The Waymo paused for a few seconds at every stop sign to take its time to assess the situation, including seeing pedestrians or bicyclists arriving or departing the same intersection. Waymo did not use predetermined stopping places (as Cruise did) and was clearly using reasoning to figure out where to stop safely or when to drive around a double parked car. (I was a little upset once when the car went around a corner before finding a place to pull over; I had to hoof it around the block to get in. I’m not sure that was a problem.)
In San Francisco, I still had to use Uber to get to and from the airport or anywhere outside city limits. And I did notice, on my way into the city from the airport, a Waymo pulled over on Route 101 with its lights flashing, but I was past it before I could see whether it had stranded a passenger or not or see why it was on a freeway in the first place.
From this instant expert, I tip my hat to Waymo for managing its product really well. With this sort of performance, you start to think that robotaxis are inevitable. (And, just as a footnote, I did sell all my GM shares some time ago.)
Great--dare I say journalistic--rundown of the state-of-the-art for those of us who haven't sampled the experience. I wonder if the quality of your experience changed your back-of-the-napkin sense of timing for when broader rollout could be accomplished. For instance, how long until a Santa Fean might be more likely to pick a driverless car over an Uber?