Around the corner from where I live, they're building a new subway station. It's part of the first major expansion of the regional subway network in decades. The entrance to the station is on a road that gets used as a cut-through by car commuters looking to avoid the highway. The speed limit on the street was lowered from 25 to 20, but cars going 40mph or more remained common. The city installed chicanes: drivers speed down the middle of the street, over the middle line. The city installed speed bumps: drivers gun it between the speed bumps. Crossing the street to get to the subway entrance is undeniably safer than it was, but it's still unpredictable, because of cars trying to evade the traffic calming.
Drivers resist interventions that make driving more difficult, even if those interventions make cities more pleasant and livable. They resist the interventions politically, but also they try to find ways around restrictions that have been put on their movement. Most of the efforts championed by the current generation of urbanists make it more difficult to drive, by design. The overproliferation, excessive speed and danger of cars are the problem that they’re trying to solve. Whereas if you're commuting by car, it is natural to think about ways to make that commute easier and faster. Anything that makes driving slower or more difficult feels like a problem you should be solving. The goals of car commuters and the goals of urbanists are directly opposed.
In some ways, that tension does not exist with autonomous cars. The people deploying these vehicles can't take the risk of programming their vehicles to break the law. As with the overwhelming risk1 associated with causing a crash, the risk to these companies' brands of deploying vehicles that are not law abiding is too high to be worth it. Tesla, always an outlier in terms of playing fast and loose, was forced to recall a version of their "full self driving" package where the human in the car could choose an "assertive" mode. In this mode the car would roll through stop signs if there was no other traffic around. This is, for the avoidance of doubt, unambiguously bad, not least because being sure there’s no other traffic isn’t always straightforward. It is nonetheless something that human drivers do more or less constantly. If you’re designing a car to drive like a human, that’s what it should do. Yet the release of this feature caused an enormous outcry. Tesla was faced with intense regulatory and governmental scrutiny and rolled back the feature within weeks.
If you are trying to get cars to do things—like slowing down—which don’t come naturally to human drivers, the comprehensively law-abiding nature of autonomous cars is potentially beneficial.
The sheer robot-ness of autonomous cars is potentially beneficial as well. Driving through areas that have been partially reclaimed from cars is frustrating. Stopping and starting, yielding to slower pedestrians and bicycles, keeping your speed below 20 miles per hour, navigating a narrow lane winding amid chicanes: all of these make driving less pleasant. If your goal is to get people to drive less and use other modalities more, that's a feature. But it's a feature that makes people mad, and people who are driving and mad are unpredictable and dangerous. Autonomous cars don't get frustrated. They are designed to accept whatever constraints exist in the world around them—obstacles, speed limits, road conditions—and plot a path from A to B within those constraints. The specifics of the constraints, and the difficulty of finding a solution within those constraints, is just a fact. The autonomous car is incapable of being bothered by them. It's the same algorithm either way.
In addition, there are some driving skills where autonomous cars are better than humans. They are better at staying strictly to a marked path. They’re better at maintaining a consistent, slow speed. They are quicker to stop for something in their path. All of these skills become more relevant as roads get narrower, slower, and more heterogeneous.
So in some ways it seems like autonomous cars—and autonomous vehicles like delivery robots—are a perfect complement to the kinds of spaces urbanists want to build. That's a vision that has been promoted by auto companies. Toyota's Woven City is envisioned as a fully mixed-use, walkable utopia. Most people are on foot or riding bicycles. The motor vehicles that are present are delivery robots and autonomous shuttles. Personal, private cars are nowhere to be found. Woven City isn't just a set of visualizations. Toyota really is building a city designed to illustrate its vision on the slopes of Mount Fuji.
That vision, of autonomous cars as part of the transformation to a more human, human-scale city, is something I promoted as well. When I was representing Perceptive Automata, when I talked about the [link ]nightmare scenario, I would always contrast it with the utopian path. If we make autonomous cars that interact well with people, I argued, we can unlock these new kinds of spaces. Autonomous cars and urbanism are not at odds. They can be part of the same positive transformation of the world.
I never quite believed in the utopian path, even as I laid it out. There are big problems with that vision as an achievable goal. Toyota is building Woven City, and putting real resources behind it. But that's not how they make money. How they make money is selling cars. Increasingly large, increasingly luxurious cars. The Toyota Tacoma pickup is incredibly popular and, like every other pickup, has grown steadily more enormous model year by model year. The Highlander SUV, even more popular, has a curb weight pushing 5000 pounds.
Toyota seems to be preparing for a future where they're in a very different business. They're making an argument, with Woven City, that they're still relevant in that future. But that future isn't the business they're in today, and it's not at all clear that future represents a business they would want to be in. Because the successes of the car industry have been built on making cars that people want to be in. And when you're talking about what benefits or unintended costs the arrival of autonomous cars might bring, that's the crux of the issue. Autonomous driving in cities makes cars in cities more pleasant to be in.
The fundamental goal of urbanism is to improve cities by getting people out of their cars. Pleasant, safe, human-scale spaces are spaces where car travel is largely unnecessary and other forms of travel like walking and cycling are practical and safe. The point of urbanist interventions like road diets and chicanes and woonerfs, modifications to the landscape which attempt to create less car-centric spaces, is not just to make walking and cycling more pleasant. It's to make traversing those places by car less appealing. The fundamental goal of the auto business is to sell more cars by making them more enjoyable and rewarding to use. Making driving a worse option, or making it a better one.
These goals are, again, directly opposed.
That’s why autonomous cars, for all the ways they could theoretically complement pleasant, human-scale urban places, are instead overwhelmingly likely to imperil them. An autonomous car that is capable of comfortably conveying its passengers through an environment designed to reduce the appeal of commuting by car subverts the goals of that environment. It makes driving more pleasant again. It makes car commuting easier and faster. It's a workaround. Just like the cars driving in the middle of the street to avoid slowing down for chicanes, the hypothetical perfectly functioning autonomous car becomes a way to treat livable, low speed, multi-modal urban spaces as a problem to be solved. Autonomous cars are taking the car commuter’s framework and optimizing around that, whatever rhetoric of transformed and more livable cities might surround them.
I talked about this previously.