Our spring issue, “Pandemic Politics,” is out now. Get a discounted subscription today!

Against Self-Driving Cars

Instead of spending billions developing driverless cars we should be building sustainable people-centered transportation.

An Uber self-driving car drives down 5th Street on March 28, 2017 in San Francisco, California. Justin Sullivan / Getty

Uber lost its license to operate in London last week. Transport for London determined that the ride-sharing company hadn’t done enough to tackle fraud on its platform, including fraud committed by drivers who had managed to game Uber’s system, driving under false identities after they had been dismissed or suspended.

It’s distressing news for Uber executives already under pressure to boost the company’s sagging share price. But Uber has a plan. It’s going to get rid of its drivers altogether and run its business through a fleet of self-driving cars.

The plan to eliminate drivers isn’t exactly new. Former CEO Travis Kalanick promised back in 2015 that the company would operate an entirely driverless fleet by 2030. Kalanick was sidelined in Uber’s quest for respectability but his dream lives on. In 2018, the company spent $475 million on autonomous fleet development.

Uber is not alone. There is a veritable frenzy among automotive assemblers and tech companies to create and deploy a viable commercial self-driving car, particularly in California, which has more than sixty companies testing driverless vehicles. The battle lines in the war for driverless supremacy are not drawn between automotive assemblers and tech companies, however. Instead, tech companies and auto giants are teaming up.

Waymo, Google’s self-driving car division, buys cars from Jaguar Land Rover and Chrysler. Apple and Uber have both purchased autonomous vehicle start-ups and Uber is working with Volvo. Honda has attached itself to Cruise, General Motor’s driverless division, while Ford and Amazon are working out a partnership. Huawei is providing the AI muscle for a number of joint ventures, while Hyundai has invested in roughly twenty companies working on autonomous driving, some of which are direct rivals of Uber. Behind the scenes, SoftBank, through it’s Saudi-backed $100 billion Vision Fund, has dumped billions into autonomous technology development at Toyota, General Motors, and Uber.

These partnerships are not quite equal. Whoever creates and controls the information and communications technology that directs these driverless cars (at this point the tech companies rather than the assemblers) will take home the lion’s share of profits. Nor do all of these companies share the same vision for a driverless world. Right now there are two main motivating stories at work in the drive for autonomous vehicles.

The first is a story about safety. Over a million people are killed by cars every year, and experts say nine out of ten of these accidents are caused by human error. Speed demons causing mayhem in school zones, elderly people in denial about their diminishing reflexes, idiots getting behind the wheel when they’re three sheets to the wind: self-driving car advocates say the solution is to take humans out of the equation.

“To the extent we can demonstrate that we’re driving more safely than the average human then,” Dan Ammann, head of Cruise, told the Financial Times, “by definition, every time we deploy one of our cars on the road we’re making the world slightly safer.”

Experts say that autonomous vehicles — equipped with lidar sensors, high-definition maps combined with GPS, and artificial neural networks — will finally make driving safe.

The second story extols the efficiency of autonomous vehicles. Nine out of ten American households own a car, yet these cars are idle most of the time. Driverless-car advocates say this is a wasted asset. Instead of letting our cars sit in the driveway or on the street they could be out making money for us as driverless taxis.

Tesla boss Elon Musk, for example, promises that his company’s cars will be fully autonomous by the end of 2019 and by the end of 2020 be positioned to participate in Tesla’s driverless taxi service. In a win-win, Tesla owners will send their cars out each day to work as taxis while the company gets a healthy cut of the fares.

The efficiency story is often melded with a sustainability story. Autonomous cars, it is imagined, will all be electric and people will share them, thus reducing the overall number of cars on the road.

These narratives are repeated over and over, by corporate executives, city planners, transportation experts, and tech journalists. Autonomous cars are just around the corner, we’re told, which is great because apparently, we’ve always wanted them. And even if we didn’t, they’re coming anyway, so we need to redesign our cities and roads to accommodate them.

The stories we tell about self-driving cars, however, don’t hold up.

For one thing we have no good evidence that self-driving cars are safer than human operators. Thanks to an incredibly business-friendly NHTSA, the data we get from companies testing autonomous driving tech is mostly voluntarily submitted self-assessments that “read like marketing brochures.”

Companies brag about all the miles they are clocking in their autonomous vehicles, but these are not quality miles. Instead, they are primarily miles generated in geofenced areas that eliminate factors such as inclement weather or poor lane markings.

But the solution is not to add more miles — it isn’t possible to add enough miles using current development strategies. As a recent report argued, to determine, for example, whether self-driving cars were even 20 percent safer than human-operated vehicles (with 95 percent confidence) would require autonomous vehicles to be “driven more than 11 billion miles to detect this difference. With a fleet of 100 autonomous vehicles being test-driven 24 hours a day, 365 days a year at an average speed of 25 miles per hour, this would take 518 years — about a half a millennium.”

As Michael DeKort, a former aerospace systems engineer argues, “it is not possible, neither in time or money, to drive and redrive, stumble and restumble on all of the scenarios necessary to complete the effort.” At the same time, he argues that “the process will cause thousands of accidents, injuries and casualties when efforts to train and test the AI move from benign scenarios to complex and dangerous scenarios. Thousands of accident scenarios will have to be run thousands of times each.”

Moreover, scientists aren’t even sure how to proceed beyond “Level 2” driverless tech in which drivers are for the most part along for the ride as the car does its thing, but are expected to intervene quickly if something goes wrong. In tests conducted by numerous companies professional drivers (even if there were two drivers in the car) kept falling asleep, and even if they were alerted, took a considerable time (upwards of forty-five seconds) to regain situational awareness — far too long to prevent a tragic accident.

Part of the safety story makes sense. Deep learning, sensors, and cameras, should all be used to make vehicles and driving safer. Volvo (the company that standardized three-point seatbelts in passenger vehicles), for example, advocates cars equipped with driver-facing cameras and AI that will alert the driver if they are looking away from the road for too long, or can even stop the car if the vehicle’s software determines the driver is driving under the influence of alcohol or drugs. These are interesting ideas that should be explored. But there is simply no good data at this point demonstrating that driverless cars, developed using any existing strategies, will lead to safer outcomes.

There are also much easier, cheaper, and more sensible ways, of reducing vehicular deaths: reducing or eliminating cars in areas with high pedestrian traffic and creating more car-free zones; investing in more and better green buses, shuttles, and trains, so it’s easy to weave public transportation into our travel plans; paying a decent wage and enforcing strict health and safety rules for drivers, particularly long-haul drivers.

The second story we’re told, however, that driverless cars are an efficient solution to wasted resources, makes no sense whatsoever.

One fairly obvious problem is that while cars are big, heavy, and expensive, they are not particularly robust. If people rent them out all day while they are at work, they’ll depreciate rapidly. Also, why would people be more willing to share their car just because it was an autonomous vehicle? Granted, there has been some interest in carsharing, through apps like Getaround and Drivy, but carsharing hasn’t taken off for the simple reason that most people can’t risk being without transportation or the possibility that a stranger might destroy their vehicle.

A far bigger problem, however, is one of vision. Self-driving car advocates are remarkably oblivious to the developmental imperatives of a landscape characterized by looming climate catastrophe, underinvestment in basic lifesaving science, resource depletion, and yawing inequality.

Instead, they zero in on ordinary working people — drivers — as the problem and support dumping hundreds of billions of dollars into dubious projects that operate on the assumption that if we can just figure out how to eliminate drivers from the equation, poof, we’ll be able to leapfrog the hard political work of developing sustainable transportation.

Those who question the wisdom of the driverless vision, who ask whether it wouldn’t be smarter and safer to make sustainable people-centered investments in technology, need to make their voices heard. We’ve let automobile assemblers and tech companies take the wheel for far too long.