Issac Asimov observed that while predicting the automobile was straightforward, foreseeing the traffic jam was difficult. Our predecessors could envision the car by picturing a locomotive, but with tracks swapped out for a paved road. Having conceived it, they built and mass-produced it. The car then proceeded to revolutionize human mobility, as expected. Immediately thereafter, the car quickly constrained this newfound mobility with gridlock, which its creators had not anticipated.
After inventing the traffic jam, the unforeseen changes kept coming. The automobile created drive-through restaurants, shopping malls, and sprawling suburbs. After remaking the built environment in its own image, the car changed the atmosphere, filling the air with exhaust. The car changed the global economy and geopolitics, as the world’s growing need for gasoline made oil the world’s most important consumer commodity. At the same time, the car also changed social life, commutes, vacations, and even sex lives. Before cars, teenagers had nowhere private to spend time with each other. Widespread car ownership, along with the Pill, is widely believed to have helped catalyze the sexual revolution.
Now, the classic automobile may soon be a thing of the past. Today, self-driving cars are available in San Francisco, Los Angeles, Austin, and Phoenix. Imagining a self-driving car is easy: picture a regular car, but swap out a human driver for a computer. Soon, robotaxis will be widespread; from across California, Arizona, and Nevada, to the whole U.S. Sunbelt, and ultimately everywhere. Just like its predecessor, self-driving technology will revolutionize the movement of people at scale. What else will it change?
First Order Effects
Self-driving cars make the roads safer. Some changes are obvious. Most importantly, a world where most cars drive themselves is a much safer one. Every year, more than 2 million Americans are injured in road incidents. More than 40,000 of those victims die. Road deaths in the USA peaked at roughly 55,000 annually in the period from 1969 to 1973, and then began a long, slow decline, reaching a low of roughly 32,000 deaths in 2014. The latest year for which we have fully-compiled data is 2022, and that year was particularly bad. While 2022 had slightly fewer fatalities in total than the year before, cyclist fatalities increased by 13%, reaching the highest level since record-keeping of cyclist deaths began in 1975. And fatalities involving drivers aged 65 or more increased by 4%.
Why numbers have spiked in recent years isn’t clear, but it’s likely that distracted driving is the cause: 2014 is also the first year that a majority of Americans owned a smartphone.
Unlike humans, self-driving cars don’t text or use social media while on the road, or take their eyes off the road at all. Nor do they become drowsy, or get drunk, or high. The sensors on self-driving cars give them 360-degree awareness; most self-driving vehicles use LiDAR, a remote sensing technology which uses lasers to build three-dimensional maps of a car’s surroundings within an inch of accuracy. (Tesla is a notable exception: their “Full Self Driving” vehicles use only cameras.) LiDAR works even if it is foggy, smoky, or dark. Self-driving cars do still struggle with rain, and even more with snow and ice. But in time, they will master these conditions, too.
Since the inception of self-driving car testing in the American Southwest, there has been only one recorded fatality involving a self-driving car. (this number does not include incidents involving Tesla vehicles that aren’t self-driving under the definition. They still require a human, watching the road and with hands on the wheel.) During this same period, human drivers have caused thousands of deaths; just Arizona in 2022 alone had 1,302 traffic fatalities. Some prominent academics have argued that “the safety advantages of self-driving cars are aspirational but have not been proven”: since the current safety benchmark for human drivers is one fatality per 100 million miles driven—a standard that self-driving vehicles haven’t had the opportunity to meet, with Waymo’s fleet having logged only 7 million miles by the end of 2023. However, focusing solely on this mileage threshold misses a crucial point. Waymo vehicles already show 85% fewer injury-causing incidents compared to human drivers.
While we cannot definitively prove that self-driving cars will reduce traffic fatalities, their superior record in preventing injuries is already clear. And we should note that self-driving systems today are, as the old Silicon Valley adage goes, the “best they have ever been, but the worst they will ever be.” Once self-driving cars are able to take full advantage of the pattern-matching skills of contemporary AI—also the worst today that it will ever be—it seems reasonable to expect that self-driving cars will only get safer. If all car movements are under automated control, a car crash could become as rare as the crash of a commercial jet.
Cleaner air, quieter streets
A world of self-driving cars is also a world of electric cars at scale. Today, all commercially-available AVs today are also EVs, and it is likely that trend will continue: EVs have fewer moving parts and require less maintenance, which is crucial for fleet vehicles, like robotaxis. They also offer precise control over acceleration and braking for which self-driving systems are best suited.
As self-driving EVs become common, we can expect the air to become cleaner, reducing the emissions that cause global warming. Today, greenhouse gasses emitted from personal vehicle use account for about 10% of all global emissions. This figure is even higher in North America: per the California Air Resources Board, 28% of the state’s GHG emissions are from personal vehicles. Shifting all personal vehicles to electricity will slash almost all of these emissions, although some of the gains will be slightly offset if the energy grid draws on coal or natural gas.
As tailpipe emissions are eliminated, smog will disappear, changing the local climate for the better. Researchers at USC’s Keck School of Medicine recently found that for every twenty zero-emission vehicles per 1,000 people in an area, asthma-related emergency visits dropped 3.2%.
Another benefit of increased EV adoption will be a reduction in noise pollution. Electric vehicles are near-silent, compared to noisy gasoline-powered cars. As self-driving EVs take over neighborhoods, streets will become dramatically quieter, except for the low-pitched hum that EVs are required to emit to warn nearby pedestrians of their approach.
Roads for robots
Another change that is easy to foresee with the advent of self-driving cars will be changes to the design of streets themselves. Our built environment is currently designed to accommodate human drivers, with all of their foibles. Since humans can’t work out when it’s safe (or fair) to proceed, we have traffic lights at every intersection to tell us when to stop. Since human attention can lapse, our lanes are wide to prevent collisions between drifting cars. And since a car can’t move once a human leaves it, we provide space to park cars near every place that a person might want to go––even if that space could be used for better purposes than vehicle storage. And so on.
In a world where cars can drive themselves, the built environment would no longer have to accommodate the limitations of human drivers. We could eliminate traffic lights and narrow lanes; at the same time, we could replace parking spaces with patios, offices, or green space. At the launch event for the Cybercab—Tesla’s forthcoming self-driving car—Elon Musk showed a video in which the acres of blacktop surrounding a sports arena morph into a forest, under the caption “This Is the World We Want”.
Given the stickiness of our built environment, it might not be so simple to turn parking lots into parks in practice. But given the hunger among American urbanists for walkable cities, the opportunity presented by self-driving technology to reclaim space from cars should not be dismissed.
A New Look
Form follows function. As human drivers are taken out of the loop, the design of cars themselves will change to accommodate their new role. Today, a car’s function is to ‘facilitate human operation’. But tomorrow, once human operation isn’t required, the function will change to ‘facilitate human comfort’. This means that personal cars will become less like a vehicle, optimized for travel, and more like a living room on wheels, optimized for riders to enjoy themselves.
In the future, occupants will want to read, work, eat, watch movies, play games, take meetings, and more in their cars (remember what 1950s teenagers with cars got up to). Gone will be the steering column, the dashboard displays, and all of the other visual and physical accommodations for a human driver. Likely the next to go will be the classic seat configuration, with all passengers facing forward. This configuration is necessary today because cars need windows to permit the driver to see. As a result, passengers need to face the direction of travel, since facing sideways or backward in a moving car with windows can provoke nausea. But in a self-driving car, the windows, and the forward view itself, may be eliminated as unnecessary, just as we’ve done away with hand cranks and the clutch pedal in the cars of today. Even if some future self-driving cars do have windows, smoother, more predictable computer driving may help to reduce motion sickness.
Zoox, Amazon’s in-house AV company, is exploring custom vehicle designs, including pod-style conversational seating and entertainment screens. They expect that the shape of vehicles will evolve too; without the need to house a gasoline engine or accommodate human sightlines, cars could become somewhat shorter and wider. And who knows what features privately-owned self-driving cars may ultimately have: reclining chairs, fold-out tables, mini-fridges, or coffee-makers are all possible.
Meanwhile, at the same time that personal vehicles become increasingly luxurious, robotaxis will likely optimize for utilitarian rides, with no more comfort than a bus seat. While today’s rideshare vehicles, where the driver-owners often make them comfortable to earn better ratings, robotaxis will operate as part of anonymous corporate fleets. Without personal owners invested in passenger experience, these autonomous vehicles will likely maintain only the most basic amenities needed for their brief, utilitarian journeys, and will optimize for frequent and easy industrial cleaning. Expect chrome fittings and vinyl upholstery everywhere.
Most contemporary self-driving firms are relying on conventional vehicle frames, which are cheaper to obtain and don’t face special regulatory hurdles. Just as the first automobiles looked like ‘horseless carriages’ before developing their own form, today’s autonomous vehicles still mostly look like traditional cars without a driver. But over time, we should expect new aesthetics to emerge that are specific to the self-driving use case: ones that privilege passenger leisure, productivity, comfort, or economy.
Social Upheaval
What subtler changes might the self-driving car unleash? The answer may depend on where we live: the cities, or the suburbs.
Today, suburbanites own their own cars. They have to, as car ownership is a prerequisite for suburban life. In the suburbs, everything is so far apart, and the transit service is so poor—limited, sporadic, or unpredictable—that almost every trip must be taken by car. By contrast, urbanites generally have access to cheap, fast, and flexible modes of transportation. Many do not own their own cars, and constrain their travel to the pre-determined routes of public transportations within cities.
Self-driving technology will not change these patterns of use. Instead, it will accelerate them.
Urban renaissance
In cities, self-driving may help to bring about an urban renaissance.
It may not seem that self-driving adds much to the urban transportation menu. Isn’t a Waymo robotaxi functionally identical to an Uber ridehail? But while riding in a Waymo may not be a significantly different experience from riding in an Uber today (aside from not having to listen to the driver’s annoying music), the lack of a driver means that automated ridehail, at scale, will be much cheaper.
The economics of the robotaxi are still hard to grasp definitively. Still, it seems reasonable to expect that removing human drivers and their wages from the equation will make rides much cheaper.
Imagine a tourist visiting San Francisco on a Saturday afternoon. She’s just finished taking a photo at the Golden Gate Bridge Welcome Center, and now wants to travel to Salesforce Tower, a distance of about six-and-a-half miles. Her Uber ride might cost $26. With a 15% tip, that’s about $30 in total cost to the user.
Now let’s further imagine that this is the future when robotaxis––such Waymo, Zoox, the Tesla Cybercab, and any new entrants––are available at scale. We don’t know how much each vehicle will cost to operate, but let’s suppose that it might cost around forty cents per mile. (This is double Elon Musk’s estimate, but since he tends to make overly-optimistic predictions, forty cents per mile seems like a safe bet). Using that estimate, our San Francisco tourist’s trip to the Salesforce Tower would cost the robotaxi firm something like $3––a tenth of the current charge.
So what will this future tourist pay? We can’t know for sure. But we can immediately see that robotaxi firms would be able to reduce prices relative to today’s baseline enormously and still make handsome profits.
Public transit on the edge
The urban self-driving renaissance won’t all be sweetness and light. As cheap as robotaxis might be in the future, they still won’t be cheap enough for most city-dwellers to use them all the time––meaning that poorer urban residents will be more reliant on public transit than ever for their daily travel. However, these transit systems may be less reliable, and worse funded, than they are now.
The economics of public transit might become even more precarious in a robotaxi world than they are today. As middle-class riders switch to robotaxis offering ever-more-affordable direct trips, transit agencies will suffer declining ridership and fare revenue, particularly on local bus routes. This could even trigger a ‘death spiral’ where reduced income leads to service cuts, which drive away more riders, which further reduces income, and so on and so on.
Death spirals are possible, but so too is a new golden age. As robotaxis flood the streets, grade-separated rapid transit—subways, elevated trains, commuter rail, and light rail operating in its own corridor—may actually become more attractive than it is today. The sheer volume of trips in self-driving vehicles will likely lead to gridlock on the streets. A train bypassing surface congestion entirely would be an attractive alternative, and the reliable, predictable travel times of grade-separated transit might be more appealing to riders than the variable, congestion-dependent journey time of a robotaxi.
Local bus service is unlikely to be so fortunate. Buses stuck in traffic with robotaxis will become less competitive, especially if robotaxi fares drop significantly. But if cities want to ensure that poorer residents, who are unable to afford regular robotaxi use, still have the benefits of good bus service, there will be options. The same software and sensors that can help a cab drive itself could do the same for a bus, reducing operating costs. Transit agencies could then use that surplus to run more frequent bus service, even if wealthier residents still eschew the bus for the robotaxi.
Some cities may go further and reserve whole lanes on busy corridors for fast, frequent bus service. More dramatically, they might cancel underused bus routes and replace them with automated shuttle buses that would arrive promptly when riders called them, creating a hybrid between traditional transit and robotaxis. Much will depend on the support that transit can command among a city’s most influential residents: if all of a city’s wealthy residents defect to robotaxi, the experience of using public transit could become significantly worse than today.
New jobs appear, old jobs fade away
Even if cities’ working classes can survive the upheaval of public transportation, the advent of the robotaxi will pare away a whole class of job: drivers. When cars drive themselves, there will be less need for truckers and delivery people, and certainly no need for ridehail drivers.
Robotaxis will, of course, create new jobs that don’t exist today. Teleoperators will remotely ‘check in’ with robotaxis experiencing problems, and will facilitate solutions to problems the AI can’t solve. Like air-traffic controllers today, these will be high-speed, high-stakes jobs, requiring sound judgment and great skill in both critical thinking and interpersonal communication, often in high-pressure environments. On the bright side for human workers, cities will need hundreds of such teleoperators, who will be paid well given the complexity of their work. However, jobs like these won’t be open to the tens of thousands of drivers that the robotaxi will displace. One of the ladders that allowed immigrants and low-skill workers to make a living and put down roots in the United States will disappear.
Like so many new technologies, robotaxis will improve the lives of users, and make a new set of workers more prosperous. However, some people will certainly be made worse off.
The Great Divergence
Overall, urban life will likely thrive in a self-driving world. Walking will remain both a pleasurable and practical transportation option when destinations are close. Robotaxis will accommodate occasional cross-town trips, or more common first-mile-last-mile trips to and from those stations. The city, practically and psychologically, will become defined by the extent and range of its rapid-transit lines. Trips in that space will be simple and cheap; trips outside of it will be more difficult and expensive. Friends, families, and communities could sort themselves around 15-minute neighborhoods, or failing that, conveniently-located transit stations.
As robotaxis make urban communities easier to traverse than ever, people who enjoy and prioritize in-person encounters will increasingly become urban people. But while urban communities become more tightly knit than ever, urban and suburban life will increasingly diverge.
In today’s suburbs, there are few reasons not to drive. The chief constraint on driving is how mentally taxing it is: driving takes time, and all of that time must be spent focused on the road. Most trips, being stop-and-go on city streets or congested freeways, demand the driver’s constant attention. But when every personal vehicle can drive itself, this deterrent will disappear. As a result, people who live in suburbs will likely drive much more often than they currently do.
Early studies indicate that suburban families will drive more once they have access to a self-driving car. In 2017, researchers gave thirteen Bay Area households access to free chauffeur service for a week. The researchers tracked the subjects’ travel for the week, comparing it to the preceding and subsequent week as a control. The idea was to see how people’s behavior might change if they had access to a car that they did not have to drive.
The result was striking: families with access to the ‘self-driving’ car used the car much, much more. On average, study participants traveled 83% more miles than they had previously, with some participants increasing their weekly travel by over 300%. Individual trips themselves got longer, with 91% of users increasing their number of weekly trips of over 20 miles. And families often sent the car out with their children alone, trusting the car (or really, its chauffeur) to drive their children without supervision. In an exit survey, one participant reported that “I definitely felt the benefits of a self-driving car. I noticed that I reach work less tired, I noticed that I can do work on my way back home and not worry much about traffic jams, and I noticed that my commute overall feels more pleasant.”
The implications are clear: the principal cost of suburban car use is the attention that must be paid to do it safely. Removing that cost makes driving psychologically ‘cheaper.’ And as Economics 101 teaches us, if a good becomes cheaper, more of it will be consumed. Therefore, we can expect that, as personal self-driving cars proliferate, more and longer trips will be taken.
Self-driving cars might change real estate markets too. If you can work productively during your commute, why not live further out, where houses are cheaper and yards are larger? As the old joke has it, the three factors that most influence a home’s value are “location, location, and location.” Historically, location has meant access: to supermarkets, good schools, walkable downtowns, and, when necessary, to the highways that connect home and work. But when distance can be filled with useful activity, the cost of that distance is much smaller. As the apparent time-cost of travel falls, more people might start to prefer larger, nicer houses, even though living in those houses will keep them farther away from everything else.
The key word here is apparent. Travel, even when filled with useful activities, has other personal costs that are less obvious. No matter how luxurious self-driving cars become, certain activities will still be impossible in-vehicle: spending time with large groups of friends and family, for instance. A three-hour weekly poker night in person will not be feasible if the trip to and from the party is 90 minutes each way. Perhaps the poker could be played virtually in the car, over Discord; but in that case, why take the trip at all? Dinners with your sister’s family, pottery class, and impromptu get-togethers are all impossible in the confines of a self-driving car. Moreover, if every car trip means climbing into a mobile entertainment center, the temptation to just stay home—or rather, to stay connected virtually—only grows stronger when your destination is so far away. Those who find such a car-driven, isolating lifestyle horrifying will probably choose city life over the suburbs.
Self-driving cars will also carry insidious costs to health. Suburban life is already quite sedentary; automated driving will make it more so. More time in cars means more time spent sitting, an activity we already engage in too often. Personal vehicles might get larger, but not so large as to accommodate a yoga routine; rights-of-way and turning radii impose hard limits on how wide a car can be. Add an increasingly sedentary lifestyle to even less in-person social contact, and the result of self-driving cars could be increased decay to our bodies and, perhaps, our spirits.
Predicting the Traffic Jam
Finally, we would be remiss to miss Asimov’s warning, and fail to predict an obvious potential second-order effect of self-driving cars: more traffic jams than ever. This claim may seem anything but obvious. After all, won’t self-driving cars have all sorts of features that will help mitigate the effects of congestion? They are less likely to cause road incidents that slow the flow of traffic. And when accidents do happen, self-driving cars won’t slow to rubberneck. Self-driving cars can follow other cars more closely without sacrificing safety. And they will make delivery cheaper, so more people will stay home and order things from the Internet, rather than leaving their homes.
Perhaps we will be so fortunate. But perhaps not; 150 years ago, the English economist William Jevons observed that as British industry became more efficient in its consumption of coal, consumption actually rose, as falling costs to operate machinery induced more production of goods. This phenomenon is now known as the Jevons Paradox, or among transportation planners, as ‘induced demand’: when a thing becomes easier to access, latent demand for the thing is spurred, increasing consumption. Historically, building more roads to decrease traffic has only brought more cars on the road, making traffic worse: Los Angeles has terrible traffic despite its immense amount of freeways. By making car trips much less costly in attention and focus, we will likely make car trips more common, and worsen congestion.
The original automobile reshaped our cities, our economy, and our sense of ourselves. Self-driving vehicles offer to do the same. Just as the first cars did, they will begin by revolutionizing transportation, and proceed to revolutionize everything else. Self-driving cars offer great promise: they will allow us to build cities that prioritize human interaction over vehicle storage, and suburbs that balance affordability, privacy, and connectivity. And soon, our streets may never see another preventable death.
Will the costs of self-driving cars outweigh the benefits? A confident prediction would be foolhardy. However, it does seem possible that thoughtful planning could bring very positive outcomes. If we’re willing to take our hands off the wheel, we could build a world where travel is almost effortless, and costs nothing but time.
