Outgoing transit agency head Jeffrey Tumlin reflects on how self-driving cars from Waymo, Cruise and Zoox affected the city’s other transportation modes.
David Zipper, Bloomberg News
SAN FRANCISCO
EnergiesNet.com 01 15 2025
Few Americans have had more direct experience with self-driving cars than the residents of San Francisco. For four years, robotaxis from companies like Waymo, Cruise and Zoox have traversed the city’s famously hilly streets as they conduct test runs, transport passengers and travel empty toward pickup locations (known as “deadheading”). Anyone can request a driverless ride by tapping an app on their smartphone, much as they would an Uber.
While some San Franciscans have celebrated robotaxis for heralding a safer, easier future of urban mobility, others complained that they obstruct emergency response, block traffic and entrench transportation networks already biased toward automobiles. In 2023, an activist campaign encouraged Bay Area residents to disable robotaxis by “coning” them — placing traffic cones atop their hoods, which froze the vehicles in place.
As the director of the San Francisco Municipal Transportation Agency since 2019, Jeffrey Tumlin has been immersed in his city’s self-driving saga. With unified oversight of transit, taxis, curbs and streets in San Francisco, SFMTA is more powerful than most urban transportation departments. Still, Tumlin and his colleagues struggled to handle robotaxi companies that are accountable only to state and federal regulators, not city officials. Clashes over self-driving cars were a hallmark of Tumlin’s tenure at SFMTA, along with his handling of the Covid-19 pandemic and controversies around dedicated lanes for bikes and buses.
Tumlin departed SFMTA at the end of 2024. CityLab contributor David Zipper spoke with him about his attempts to manage San Francisco’s robotaxis, and what his experience portends for other cities where self-driving cars could soon arrive. Their conversation has been edited for length and clarity.
You had worked in the transportation sector for decades before joining SFMTA, and you were already familiar with autonomous vehicles. What have you learned about self-driving technology in the last few years that most surprised you?
As a regular Waymo user, I have watched the cars become better than I am at seeing pedestrians hidden from view and predicting their behavior. I didn’t think that was going to happen. I am surprised by how sophisticated they are with erratic human behavior, which I had assumed would be very challenging.
Overall, are robotaxis a positive or negative for San Francisco?
So far, there is no net positive for the transportation system that we’ve been able to identify. The robotaxis create greater convenience for the privileged, but they create problems for the efficiency of the transportation system as a whole.
What do you mean by that?
What I like about Waymo is that the user interface design works well. I don’t have to talk to a human, and the vehicle’s driving behavior is slow and steady. I think robotaxis offer the potential for significant upsides for personal convenience, but it remains to be seen whether they offer any overall benefit to the transportation system.
How would you respond to those who say robotaxis are making San Francisco a better city because the experience of using them is superior to other ways to travel?
I agree that there are qualities or Waymos that outperform other modes. The vehicles are very nice. The driving behavior is slow and steady and predictable, and there is chill music.
But those are qualities that you can replicate in any mode. If we mandated speed governors, passenger cars can be slow and steady. If we regulated taxis in order to optimize for user convenience and safe driving behavior, taxis could emulate those same qualities as well. Similarly, if we had massive private funding, we could achieve the same level of quality in public transit.
Was there a robotaxi moment over the last several years that prompted you to think, “I cannot believe this is happening”?
In my first meeting with the original CEO of Cruise [Kyle Vogt], I was in City Hall trying to lay out the case why good data supports public trust, and explaining why I didn’t expect robotaxi companies to be perfect. San Francisco streets can handle a good deal of chaos, but I can’t have dozens of Cruise vehicles being immobilized in traffic and blocking my train lines.
He leaned across the table from me, pounded his fist on this heavy oak table, and said, “Jeff Tumlin, you are the single greatest threat to the American autonomous vehicle industry.”
I can handle myself in almost any situation, but I was just tongue-tied. I was trying to be a good bureaucrat while supporting the industry — being a partner with prudence. I didn’t even know that he had any idea who I was. So it was a very strange first meeting, and I think it set the tone for the challenges that we had following that with Cruise.
[After a San Francisco woman was badly injured on Oct. 2, 2023 by being dragged beneath a Cruise vehicle, California regulators accused company officials of withholding crash information and rescinded its operating permits. General Motors announced in December that it was terminating its investment in Cruise, and that the company would cease offering robotaxi service. A Cruise spokesperson declined to comment.]
Waymo recently ran a promotion offering riders a discount if they traveled to one of several Bay Area transit stations. For public transportation, are robotaxis an ally or a threat?
One of the things that I find fascinating about the robotaxi industry is it appears to have learned nothing from the experience of Uber and Lyft.
Less than 1% of Uber and Lyft trips in the Bay Area connect to transit. But in the late 2000s, the companies made all these lofty promises that they were going to be a critical first/last-mile connector to public transportation. They believed that the existence of ridehail would allow people to shed cars because there would always be a vehicle there when you need it.
Not only did none of those promises come true, the opposite actually occurred. The San Francisco County Transportation Authority did an analysis in 2018 showing that almost half the increase in vehicle miles traveled in San Francisco was due to Uber and Lyft. They were displacing transit, walking and biking trips, they were adding to deadheading miles, and they were significantly increasing traffic in the most congested parts of the city.
If San Francisco eventually has 20 or 50 times as many robotaxis as it does today, how would the city be affected?
Again, there is fundamental tension in transportation between user convenience and system efficiency. Taking an Uber or Waymo is super convenient for me. I get to come and go when I please, I’m in a weather-protected little vehicle, and I don’t have to deal with people I don’t know or don’t like.
Read more: The Case for (and Against) Robotaxis
At the same time, it’s an incredibly inefficient use of roadway space. Flooding the city with Waymos would significantly increase congestion and allow the streets to move fewer people, which threatens San Francisco’s economic vitality. That’s particularly true if those Waymo trips are substituting for walk, bike or transit trips.
By the way, since robotaxis are funded by venture capital they could significantly underprice Uber and Lyft and put them out of business. That’s exactly what we saw with ridehail companies 15 years ago, when they undercut taxis and raised price once they dominated the market.
How many robotaxis are operating in San Francisco right now?
We don’t know. The city has regulatory authority over taxis, but not over robotaxis. We get very little data from state and federal authorities.
So how does San Francisco gather information about robotaxis on its streets?
The bulk of the relevant information that we get on robotaxi performance comes from 911 calls and social media. This has been a source of great frustration, since robotaxis have obviously been tremendously impactful to the transportation system.
We’re still unable to get consistent data. For example, we’ve asked our transit operators and first responders to fill out paper forms when they observe a robotaxi incident. Then a bureaucrat has to type it into Excel to analyze. It’s the worst possible way of trying to do data analytics.
To what extent do robotaxi companies voluntarily share data with the city?
We are given information only when we request it after an incident. You have to go to the company and say, “We know this occurred. Give us information.” Then they will.
The greatest challenges have been interactions at scenes like a police incident or building on fire. When the autonomous vehicles get confused, they become immobilized and get in the way. We had many such issues.
Autonomous vehicles also have a hard time interacting with humans who direct traffic at busy intersections. The vehicles get confused, particularly during special events and motorcades. They are also flummoxed at construction sites. A Waymo vehicle just a few days ago drove into wet cement.
Ideally, what robotaxi data should be collected and shared with cities?
The most important is safety data. Unfortunately, state and federal regulators as well as the robotaxi companies themselves define safety as “collisions.” As long as the robotaxi is not colliding with other road users, there’s no safety data.
For example, it’s not considered a safety incident when a robotaxi ignores police caution tape or becomes immobilized in the middle of a complex intersection. As long as the robotaxi itself isn’t hitting anything, it’s also not considered a safety incident — even though it may create significant safety problems for other roadway users.
I think we need comprehensive data about the tax robotaxis impose on overall transportation system safety, as well as other key characteristics. For instance, to what degree are robotaxis allowing us to move more or fewer people on our roadways? To what degree are robotaxis able to serve the needs of people with disabilities, particularly people in wheelchairs? Are robotaxis merely creating greater convenience for the wealthy, or are they providing essential mobility for people with fewer choices?
Another big one: What are the systemic impacts to greenhouse gas emissions? In San Francisco, all robotaxis are electric, but their computing requirements are consuming vast quantities of energy that get in the way of using the existing power grid to electrify other vehicles.
The historian Peter Norton has argued that self-driving cars threaten public urban space as much as early automobiles did a century ago, when they prompted the invention of “jaywalking,” the implementation of crosswalks and the narrowing of sidewalks. Is that a legitimate concern, or is it overblown?
Well, I have sat in on panels at transportation technology conferences discussing the “pedestrian problem.” There have been many such conversations asking, “If we have ubiquitous autonomous mobility, why would anyone ever need to walk again?”
Waymo has at least demonstrated that we don’t need to install chips in all humans so that AVs can identify pedestrians. That’s good news; industry is not having that conversation anymore. But yes, I am very worried that so much of the joy of the remaining walkable cities in the world could be at risk from self-driving cars.
At the end of the day, do you think AVs belong in dense, multimodal places like San Francisco?
One story that has not been well understood is that the two big Waymo cities are San Francisco and Phoenix. Phoenix is very happy with their experience with autonomous vehicles.
Self-driving technology seems to work really well in a system like Phoenix that was designed for cars. The level of complexity that AVs have to deal with in a city like Phoenix is far lower than in a city like San Francisco.
AVs might thrive accommodating automobile-dependent cities in a way that’s not soul-destroying. Being stuck in traffic congestion while having to operate a motor vehicle every day is exhausting. If you no longer need to worry about operating the vehicle and can make productive use of that time, it could create an interesting new economic model for cities like Phoenix.
Fair enough, but the densest cities are often the most enticing robotaxi markets. Many residents are affluent and willing to pay a lot for relatively short trips in a place like San Francisco — or New York, Seattle or DC.
It goes back to this fundamental tension between convenience for the wealthy versus the needs of the transportation system as a whole. For private AV companies to be viable, they need to target wealthy folks in dense locations where vehicles don’t spend all their time on the road deadheading.
Is there a regulatory solution to nudge robotaxi companies toward the Phoenixes — the places where their upside seems greater — and away from the San Franciscos?
I think there is, assuming regulatory structures are focused on public outcomes while supporting technology development. The challenge we have right now is those regulatory functions are diffuse, and they tend to be focused on narrow questions rather than larger ones like “how does this transportation technology help us move more people in the same amount of roadway space?” or “how does this technology advance overall transportation system outcomes?” Part of the problem now is that our transportation regulatory structures were designed for the automotive era.
How often do city leaders from other parts of the world ask you for guidance about robotaxi policy?
All the time.
What do you tell them?
If they’re from East Asia and Europe, where there is a greater willingness to use regulation to uphold the public good, I focus on early establishment of data protocols. I suggest they let industry expand as rapidly as it can demonstrate that it’s ready. That expansion can be done in several dimensions. It can be managed by geography, time of day or number of vehicles.
What if those seeking advice are from an American city?
Unfortunately, we found in San Francisco that there’s very little that we can do because most US cities have no regulatory authority over self-driving cars. One area where cities do have some control and ability to partner is the curb. We’ve had some good voluntary partnership with the autonomous vehicle companies when we say, “Please do not pick up and drop off in the bike lane, or in front of the hospital or the fire station.”
If I read between the lines, it sounds like part of your advice to American cities is to plead your case to the governor and state legislature, if they’ll listen to you.
That’s right.
bloomberg.com 01 09 2025
