What self-driving cars mean for the city
Spoiler: it’s about more than just traffic
There’s been a lot of autonomous vehicle (AV) discourseTM as of late, mostly on the question of whether AVs will solve traffic (they won’t) and what their impact on mass transit will be (it depends). As usual, though, there's more to the story.
AVs stand to make U.S. roadways orders of magnitude safer, especially for pedestrians. They're also a first step in a major change to the way we regulate public space (but more on that at the end).
Autonomous Traffic
I already mentioned that there’s more to this story than traffic, so I’ll be brief: Single-occupant self-driving cars, holding everything else equal, will make urban traffic worse.
As Salim Furth has pointed out, people already drive close together in urban areas. So, unlike with highway driving, AVs won't increase road capacity by allowing us to all (safely) drive closer together. They may also reduce the opportunity cost of being stuck in traffic. Some folks will find three hours working in a mobile office (or reclining while they scroll TikTok) far more tolerable than even one hour driving behind the wheel. So, individually more bearable commutes could easily lead to collectively worse traffic on the whole.
The solution, as Alex Armlovich has noted, is congestion pricing.
Traffic isn’t really a technology problem at its core; it’s an unpriced public goods problem. Getting people to internalize the cost of their consumption means putting a price on the space they’re taking up. That price should then incentivize things like vans and minibuses, which take up less space on a per capita basis.
So, AVs are awesome, but they’re not going to solve traffic or obviate the need for high(er) capacity transit. And as far as I’m concerned, that’s all settled debate. But there are other implications to widespread AV use that we ought to appreciate as well. Let's start with safety.
Building a better driver
Having lived in San Francisco, I’ve had the opportunity to actually ride in a Waymo. It’s a pretty neat experience and works as well as calling an Uber. That said, Waymo affects my life more as a pedestrian than as a rider.
The particular neighborhood I live in is blessedly walkable. However, like anyone walking anywhere in the United States, the risk of getting hit by a car is always a concern. Pedestrian deaths due to cars hit a 40-year high in 2022. The U.S. is also worse than other countries when it comes to protecting people from getting run over.
These types of stats are always top of mind as I’m crossing the street in front of a waiting car. But I’ve come to realize that this survival anxiety is mostly absent when walking in front of a Waymo.
Waymos don’t get tired or impatient. They don’t try to multitask on their smartphones. They definitely don’t drive drunk. And they most certainly don’t get distracted trying to fish out a french fry from their In-N-Out box because they waited too long to eat and are becoming increasingly hangry, but also can’t pull over because it’s San Francisco and who can ever find a place to park anyway.
Personal anecdotes aside, there’s evidence to suggest Waymos (if not all AVs everywhere) are safer drivers than humans.
did an excellent write up on the company’s crash data and found that Waymos get in 3x fewer accidents than human drivers (and of the accidents that did occur, most of those were a human's fault). Granted, AVs are currently limited to certain geographic areas with favorable conditions, but at least where they're already operational, they’re outperforming humans on safety.So–AVs don’t get distracted, angry, or impatient and that makes them safer drivers. There's another angle here, though, that not only has to do with safety, but also the future of how we govern public spaces:
Autonomous vehicles don’t break the law.
Asimov’s Laws of Urbanism
Okay, so for our non-U.S. folks, this needs some set up. Allow me to do that by explaining that Americans hate traffic laws — especially speed limits.
There’s something embedded within our political psyche that chafes at the idea that we shouldn't be able to exceed a posted speed limit whenever we feel the need. Generally observed norms of highway driving means driving ten miles over the posted speed limit. So on top of us having dangerous roadway design, we have a culture of selectively ignoring rules whenever we think we can get away with it.
AVs, in comparison, don’t selectively break the law. They don’t drive faster than posted limits, make illegal turns, or otherwise do things we tell them not to do. This adherence to rules contributes to safer public spaces. It also heralds a change in our paradigm for regulating public spaces.
Most people interact with rules by doing math that looks something like:
probability (of getting caught) * impact (of getting punished)
That’s not to say that people literally think this way, but it’s a decent description of how people behave with respect to following rules.1 Autonomous systems don't do this kind of calculation.
When we write rules for AVs, those rules are instantiated in code and directly govern the behavior of the thing moving around a city. A drone won’t enter a no-fly zone because, at the software level, it can’t. A delivery vehicle won’t block a non-designated loading/unloading curb because it knows not to — and knowing not to is enough. Increasingly, regulating public spaces will only require writing sensible rules; compliance will not require enforcement.
Now, I’m obviously being a bit hyperbolic here. None of this applies to human beings deciding to walk around being disruptive. We can’t write a rule in python to solve the problem of “practical joke” streamers harassing people for content. We may also dream up rules that are too difficult to completely articulate to a machine.
That said, I believe regulation via code will start with AVs and gradually be deployed to a broad range of technologies operating in public space.
Last year, there was a bill floated in the California legislature to programmatically prevent speeding. Had it passed, it would have obliged manufacturers to build vehicles with governors that dynamically cap a vehicle's speed. This isn't a one-off example, either. The Europeans have at least taken steps in that direction as well (which is probably where the legislator in California got the idea).
But this doesn't stop with AVs.
DJI consumer drones won't fly into restricted airspace. Lime electric scooters won't let you end a ride outside of designated areas.
Remember when we said we can't write a rule in python to stop hidden camera content Youtubers? Well, maybe we can. There's no technical reason we couldn't render smartphone cameras inoperable in specific places. If a device is aware of its location, it's possible to specify how it behaves based on where it's at.
If technology feels like magic, this approach to regulation starts to feel like it as well. Kind of like how wizards in the Harry Potter series can apparate (re: teleport), but wards on the school grounds prevent that from working at Hogwarts.
Society changes (even if people don’t)
For anyone about to respond with incredulity about public opinion, you're probably right —but about the wrong thing (sorry).
While I don’t think currently existing Americans are prepared to accept these types of regulations, I believe that will change in the future.
Imagine a teenager living thirty years from now. They’ve grown up in a world with widespread AV use for transportation and delivery. Delivery drones are commonplace as well. She's even grown up with self-piloting EVTOLs for regional transportation. The idea that these commonplace things have restrictions built into their behavior will be so banal as to not even register. Further, the concept that there’s something categorically different about a human-operated vehicle might require a sociology lesson on mid-twentieth century culture.
It’s often the case that society changes, but people don’t. Cultural drift is more about cohort replacement than persuasion.
Circling back to where we started, AVs will be the first scaled up example of regulation via code. And when we look back on this period, we'll see this as the moment where the introduction of automated systems into urban spaces changed the way we regulate technology in the public sphere.
Postscript
Between writing and publishing this draft, over a dozen people were killed in a terrorist attack on Bourbon Street in New Orleans. The terrorist began their attack by simply running over people with his car. A similar car-based attack took place in Toronto in 2018 and using cars to run down victims in a crowd is now a recognized terrorist tactic.
This reality highlights a few things worth repeating.
Laws, regulations, and norms of behavior that rely on fear of punishment do not work when dealing with motivated ideologues who don’t care what happens to them.
Building off that, preventing people from using technology to harm one another in public spaces will eventually have to involve rendering technology inoperable in the commission of violence, rather than just threatening to punish people after the fact.
When, where, and how regulation via code move forward is all a bit murky, but as a general rule, these things move forward one tragedy at a time. Hopefully it won’t take many more for us to develop better systems for keeping people safe.
Alright folks, that’s it for this week. Ending on far more somber note than originally intended, but such is life. Let me know what you think in the comments (especially if you’ve got a different take), and if you’ve really enjoyed reading Urban Proxima so far, consider leaving a tip (all proceeds go towards caffeine and research material).
Otherwise join us back here next week for the first installment in our four-part series on building new cities from scratch.
Rules we’ve internalized as worth following because they’re just the objectively right thing to do, notwithstanding.
I appreciate your thoughts on how more widespread autonomous vehicles will change the city in fundamental ways. The point on commute times is quite important as the gains from increasing the size of labor markets across large cities by reducing the discomfort and inefficiency of the commute would be a gamechanger.
But I disagree with the notion that even widely deployed AV tech will make using vehicles for mass murder any less likely. On the contrary, as we start to expect safer streets and fewer drunk drivers, it'll likely make such attacks *easier* to pull off. Even if we outlawed non-autonomous vehicles in parts of the city (very unlikely in the near future), a determined terrorist would bypass these protections or hack some vehicle out there to let them drive it into a crowd.
These technologies won't reduce the number of terrorists looking to use vehicles as weapons of mass murder anytime soon. If we want to reduce the incidence of such crimes the best way is to erect physical barriers or have people gather in places that vehicles any bigger than a golf cart can't actually reach.
When it comes to autonomous regulation enforcement (whether it's speed, cameras, recording, etc.) I think people's comfort almost completely depends on what they're habituated to. Everyone is comfortable having smartphones that record when and where you tell them too, and cars that drive at whatever speed you can manage (so long as you aren't caught), so there will be a lot of pushback.
Pilots on the other hand are completely used to the restrictions placed on when, where and how they can fly, so the introduction of new restricted airspace is met without fanfare, especially since the restriction is for the safety of all pilots. Even though it would be more convenient, and potentially more free, if pilots were able to fly wherever and whenever they wanted, the traditional regulations make it unthinkable for anyone to suggest deregulation. You can probably say the same thing with the EU and gun control.