Tuesday, December 21, 2021

Why and How to Transition to Self-Driving Cars

 by Michael C. Dorf

A recent story in the NY Times reveals how Time Magazine's Person of the Year Elon Musk repeatedly over-hyped the self-driving abilities of Tesla's cars and pushed engineers to design the company's "autopilot" features to rely solely on cameras even though most experts in the field believe that self-driving cars should combine visual input with radar and/or lidar (similar to radar but using lasers). In today's column, I want to raise some questions about the real and imagined virtues of self-driving cars and also say a few words about the transition to them.

If you're wondering what expertise I have for this exercise, the answer is: very little; I own and drive a car that has some electronic assistance and self-driving features (a fairly awesome fully electric Ford Mustang Mach-e, about which more below); and as it says right at the top of the blog, DoL provides commentary on law, politics, economics, and more. Today is a more day.

It's hard to predict what technologies will catch on. Facebook's coming metaverse strikes me as something between a dystopian nightmare and a bad joke. Google Glass was a bust, even leading its adopters to earn the nickname "glassholes," but you never know. Apple is apparently betting that the failure of Glass reflected a premature effort at implementation--launched before there was sufficient computing power and connectivity, with a rumored coming rollout of Apple Glasses. When mobile phone makers first started putting cheap cameras on phones, I remember thinking it was a stupid idea, like making a fork that's also a pencil; now only professional photographers and dedicated hobbyists use standalone cameras. 

There are four main reasons why self-driving cars could catch on: (1) safety; (2) environmental impact; (3) convenience; and (4) coolness. Policy makers have little reason to care about (3) or (4) but should be keenly interested in (1) and (2). That said, if we collectively reach the conclusion that all things considered a transition to self-driving cars would enhance safety and/or result in less harm to the environment, then their convenience and coolness matter because they affect whether and how quickly a fleet transitions to fully self-driving cars.

I'll address each of the foregoing four categories shortly, but I want to start by defining what I mean by a fully self-driving car. I mean just that--a car driven entirely by computer, so that there's no need for a backup driver. From the occupant's perspective, a fully self-driving car is equivalent to a taxi. You don't need to sit behind the wheel. Indeed, there is no steering wheel. You tell the car where you want to go and then you can take a nap in the back seat or watch a movie while your fully self-driving car drives you there. That is presumably the future that Elon Musk has been promising is just around the corner and that both tech companies like Waymo (part of the Google conglomerate) and the leading conventional car manufacturers are working towards.

Fully self-driving cars offer two sorts of potential safety advantages. First, computers don't get sleepy, their attention doesn't wander, and their reaction time is orders of magnitude faster than that of humans. Self-driving cars will eventually resemble computer chess programs--clearly better at the task for which they're programmed than any human.

Second, once self-driving cars are found on streets and highways in large numbers, they will be able to communicate with one another and even with smart-roads to avoid accidents. To decide whether to allow self-driving cars, regulators must determine whether they are at least as safe as human-driven cars, but the promise of substantially greater safety offers a reason for regulators to promote the transition.

At first glance, there's no obvious environmental benefit from self-driving cars. True, many of the cars that have self-driving features are also electric vehicles like Teslas and my Mach-e, and electric cars have a smaller carbon footprint than internal combustion cars--much smaller where solar panels or other clean energy charges the electric car. None of that is directly relevant to self-drivingness, however. You could put autonomous vehicle technology on an internal combustion engine car. But as a practical matter, self-driving technology will be developed for electric vehicles (as this GM promo explains), so more self-driving cars means more electric cars, which means less pollution and greenhouse gas emissions.

At least in urban and reasonably dense suburban areas, self-driving cars also reduce the size of the fleet because they reduce the need to own a car. You just summon a self-driving car when you need it. Now, you might think: hold on; self-driving cars will substitute for taxi, Lyft, and Uber rides, not car ownership. That's not entirely true, but even if so, that means fewer cars being manufactured. An overworked Uber driver can't drive as many hours as a computer that never needs to rest and only needs to stop to recharge (in a process that will be quick as fast-charging technology improves). If self-driving cars that are available through ride-hailing apps wear out more quickly (because driven more miles per day), that will keep the fleet more modern, which itself will reduce environmental costs.

To my mind, the troubling question is not whether to substitute self-driving for human-driven cars. Once the safety threshold is crossed, we should. The question is whether public and private investment in self-driving cars displaces investment in and use of public transportation. In an enlightened society, it would not. In the U.S., it might. But for now I'll accept the conventional wisdom that in the U.S. a transition to self-driving cars will reduce the environmental cost of transportation relative to the status quo, whereas a shift to much greater use of public transportation, which would reduce the environmental cost even more, is unlikely.

In light of the safety and environmental benefits of a shift to fully self-driving cars, any actions that undermine the shift are harmful. By over-hyping the self-driving capacities of Teslas, Musk has done harm. The Times story I linked at the top says that a small number of people have already lost their lives while Tesla's autopilot was engaged. It's possible that in each case the driver failed to take over when they ought to have, as Tesla argues. But insofar as Tesla encouraged people to believe that the technology could do more than it could, those deaths were avoidable. Moreover, even if Tesla in autopilot mode is as safe as or safer than a conventional human-driven car, every accident (and especially every fatal accident) that occurs while a car is in self-driving mode undermines public confidence in the technology. In order for self-driving cars to be accepted as safe, they need to be substantially safer than human-driven cars.

The fact that Teslas now apparently allow drivers to use the touchscreen to play video games while driving is surely a step in the wrong direction. So is the fact that whereas my Mach-e monitors where my eyes are looking while in hands-free "BlueCruise" mode, Tesla drivers can look anywhere they want.

Besides safety and environmental impact, what other factors will affect the transition to fully self-driving cars? Here I would note that what I called factors (3) and (4)--convenience and coolness--are in some tension. The ability to take a nap in your car as it drives you home from work makes it convenient, but napping isn't exactly cool. Until now, what has made a car cool are styling and some sense that driving it makes one powerful, like one is piloting a fighter jet or racing in the Indy 500. That's presumably why my Mach-e, which is a terrific car in a great many ways, is called a Mustang and has a ridiculous feature that enables "engine noises" to be played inside the vehicle to give the driver the illusion that he (let's face it, this feature is designed for men and boys) is driving a muscle car. Giving over control of the car to the computer so you can take a nap is the opposite of this kind of cool.

That doesn't mean that self-driving cars can't be marketed as cool. It does mean that for some time there will be different market segments. Tesla promotes what we might call (for lack of a better term) nerd-cool. They appeal to the same sort of people who are early adopters of other technology because they find the tech itself cool. What has made Tesla successful thus far is the ability to combine that appeal to nerd-cool with a simultaneous appeal to what we might call (also for lack of a better term) macho-cool, as exemplified by the sports car styling and performance of Tesla's cars.

Will that combination appeal to everyone? Probably not. For some time, we can expect various car companies to stay in their existing marketing lanes, with awkward exercises like the Mach-e's fake engine noise setting or whatever Porsche is doing to make its self-driving mode seem cool for people who want to drive a Porsche.

As someone who doesn't care too much about any kind of cool, I bought my Mach-e despite rather than because of its attempt at coolness. I got the Ford rather than a comparable Tesla (the model Y) mostly because there's a Ford dealership in Ithaca where I can get it serviced, whereas the nearest Tesla service center is two hours away in Rochester. True, I was repulsed by Musk's behavior early in the pandemic, but that was a bit of a wash. After all, Henry Ford was a union-buster, racist, and vicious anti-Semite who literally accepted an award from Nazi Germany. Having died over 70 years ago, Henry Ford is not currently running Ford Motor Company, but the company could and should do more to acknowledge its founder's profound flaws.

People will find their own reasons to want or not want self-driving cars. In the meantime, less hype and more actually useful features would be nice. I suppose it's an impressive feat of computer engineering that my Mach-e can parallel park and head-in park automatically, but really, should anyone who can't head-in park without assistance be driving at all?