Understanding Self-Driving Cars

Let’s Make Better Cities, Not Just Better Cars

In 1988, I visited the computer science department at Carnegie Mellon University. As a 16 year-old prospective student, I was shown many of the projects that the department was working on at the time. Computer vision — enabling a computer to “see” with cameras and interpret the results — was one of their big projects. One of the moonshot ideas that they hoped to enable with this technology was a computer-controlled automobile, which was in the very earliest and most primitive stages at that time.

This technology has now been popularized as Google’s “Driverless Car” project, and many in the world of technology have hailed it as a panacea for all that ails us: congested roads, distracted drivers, driver error, fuel efficiency, and more. Designer and engineer Bran Ferren recently suggested that the self-driving vehicle would be the most important technology of our era. And he may be right. But we cannot consider this technology by itself or in a vacuum: there are other seminal shifts occurring simultaneously that we need to understand.

The End of Ownership

Car-sharing of various kinds is exploding. Services like Zipcar, RelayRides, Car2Go, and GetAround are revolutionizing how cars are used. Rather than owning a car and using the same car for all your trips, you will have a car app. You’ll have access to a car (or a van or a convertible or whatever) wherever you happen to be. If the car is ever more than a short walk away, it can drive itself to you. You’ll pay for a car as a service, rather than as an asset to own.

Research firm Alix Partners indicates that car-sharing programs will displace 1.2 million new vehicle sales by 2020. For each shared car, 32 new cars will go unsold. This will usher in a major shift in the world economy.

The Shift to Cities

The world is moving into cities in record numbers. Millennials are choosing cities over suburban living almost exclusively. In urban centers, you really don’t need a car; many people actively figure out how to avoid owning one due to expenses related to parking, insurance, and maintenance. And in an urban environment, a car (self driving or otherwise) is less important than walkability and living close to a variety of resources.

Commuting Is Dead

The notion that a self-driving car will be a godsend because it will enable people to be productive during their long commutes into the city is an idea only a boomer could love. No one will want to commute, in a self-driving car or otherwise. Even if self-driving cars can improve the throughput and safety of our freeways (which they can), our roads are in need of serious reinvestment. They are at capacity, and we really don’t have the budgets or space to build more roads in most built-out urban centers.

But besides all that, spending time commuting at all is just dumb. We can’t overcome distance. No matter how we travel, commuting takes time — time that can’t be spent doing something truly productive or pleasurable. And yeah, you could “do work” in a self driving car (or on a train), but wouldn’t you rather just walk 10 minutes to work and grab a coffee on the way in? Why spend your time in a tin box of any kind?

And whether your commute is powered by gas or electric propulsion, any distance you travel contributes to your carbon footprint. Why not simply move around less?

Self Parking > Self Driving

The most dramatic effect of autonomous vehicles is the possibility of a self-parking car. The implications here are enormous. Imagine arriving at a party in the rain and allowing the car to go park itself. Or going shopping for furniture, being dropped off in a convertible and leaving with a minivan, summoned from its parking spot with an app.

The truly revolutionary change here is the potential to reshape land use. We waste an inordinate amount of land on parking. In urban contexts, parking has had a dramatic negative effect on land use patterns over the last 50 years in particular, and it’s only recently that we have begun to figure out that devoting land to parking is, in fact, toxic to urban vibrancy.

If we shift away from car ownership to car sharing, we’ll have dramatically fewer parked cars. If we have fewer parked cars, we can devote less land to parking. If we devote less land to parking, we can increase density in urban environments, which leads to wealth creation.

The Tech Is Really Hard

The technology to create a truly safe self driving car is difficult, to say the least. It’s entertaining to hear people talk about it like it’s a certainty — an inevitable entitlement of our age. The fact is that this is going to be really hard to implement at scale (which is why Google calls it a moonshot, remember.) And it’s going to be even harder to regulate.

Techno-fundamentalist cheerleaders love the idea of the self driving car. But many have never actually studied the problem in detail and consider it to be no harder than a cool smartphone app or anti-lock brakes. It’s an incredibly difficult problem. They like to say, “computers can process far more information than a human can” — but that’s not true. Computers can process less information more slowly; what they can do is pay attention to very specific signals more consistently and reliably. But they will often need our help to understand the world around them, and they can’t interpret new information in context anywhere as quickly as a human can.

They like to say, “it will save 15,000 lives in the US annually, and a million lives worldwide.” But that’s a fallacy, because that assumes a static world where we blow up all the millions of conventional cars and replace them with robot cars overnight (which we won’t), and that machine drivers won’t make entirely new mistakes on their own (which they will). The fact is that auto deaths are declining anyway — partially because young people are driving less.

Instead, we live in a dynamic, complex world, where we’re going to slowly start to replace the mistakes of some human drivers with the brand new and cognitively different mistakes of robot drivers. Just wait until a robot car “kills someone” with a bad sensor or bad decision. It will happen. The outcry will be enormous. How we deal with the transition between human drivers and robot drivers will be very challenging. Coming up with a legal framework to manage it in a sane way will be quite difficult and frought with emotion. Don’t expect it all to go smoothly.

Transit Is More Important

We already have a really incredible technology for moving large numbers of people at scale that can also create large-scale economic growth: it’s called mass transit, and it’s the single best investment that we can make in our urban centers. It works at both long-haul and short-run scales.

A mag-lev train line from Washington, DC to Boston would create massive economic growth in all the cities it touches, effectively merging DC, Baltimore, Wilmington, Philadelphia, Trenton, Newark, New York, and Boston into one giant supercity with incredible density and potential for wealth generation.

Improving transit in cities like Baltimore (whose aborted subway system sits as a monument to shortsighted planning) would dovetail with such an effort, quickly transforming its economics and creating more opportunities for everyone.

All In Context

As the suburban dream fades and we shift into cities in ever larger numbers, it’s important to understand the self-driving car in context. It is part of a constellation of technological upheavals that will be happening all at once, not a singular change to be dropped into the America of the 1960's.

It’s as if someone were to look towards the future in 1990 and declare, “Wow, we’re going to be able to run Windows 3.0 so FAST in 2014!” Clearly that misses the point. We could do that. But would we ever want to?

Technology is all about “how.” But we need to ask “why.” Because the real reason self-driving, self-parking cars are cool — when we finally do figure them out — is because they stand to make our cities much more vibrant, beautiful, and interesting by allowing us to infuse them with even greater density and diversity. Ultimately this should allow everyone to spend more time doing what they love and less time moving around.

We should be designing places where 90 percent of what you need on a day-to-day basis is within a 15 minute walk or 10 minute bike ride. We did this 100 years ago. The need for cars — self driving or otherwise — is simply a symptom of a design failure: a sign that we have failed to design functional places. So rather than exalting the self-driving car as a salve for this otherwise terminal disease, let’s figure out how to use this technology to help us create better places.