In 2016, the president of Lyft, John Zimmer, offered a rosy and ambitious vision of technological progress, predicting that soon most of his company’s customers would be cruising around in autonomous taxis. Car ownership would decline dramatically, he said, as people would be able to summon robo-taxis on demand, all paid for by the mile or via a Netflix-style subscription. (“Don’t drive very often? Use a pay-as-you-go plan for a few cents every mile you ride,” he wrote. “Take a road trip every weekend? Buy the unlimited mileage plan. Going out every Saturday? Get the premium package with upgraded vehicles.”) The idea was utopia: Eliminating parking spots would make room for more public space, and taxis themselves would be hybrid office-entertainment venues, ferrying blissful passengers safely between destinations. This wasn’t just remaking transportation—it was transforming society. (“Transportation doesn’t just impact how we get from place to place. It shapes what those places look like, and the lives of the people who live there.”) All this would happen by 2021, he predicted.
None of it has come to be, obviously. Last week, Lyft announced that it had sold its autonomous vehicles division to Toyota for $550 million. Waymo, Google’s self-driving division, just shuffled its executive leadership. Late last year, Uber sold its self-driving arm to a start-up. In a recent Securities and Exchange Commission filing, Tesla acknowledge it may never reach its ambition of fully self-driving cars. Across the industry, autonomous vehicle efforts have proven to be stubbornly difficult to bring to fruition, consuming billions of R&D dollars and thousands of engineering hours. Meanwhile, the infrastructure needed to support autonomous cars—government investment in roads, highways, communications, along with proper regulatory oversight—remains inadequate to the challenge. Additional issues present themselves around the law, consumer trust, and even the computational limits of deep learning and artificial intelligence. Despite these enormous obstacles, tech leaders preach continued faith—and venture capital investment—having spent the last decade promising that our blissful self-driving future is just around the corner.
At a time when fighting climate change and providing alternate forms of mobility should be core parts of urban and transport policy, we seem to be going backward. The federal government has largely left it to states to regulate autonomous vehicles, leaving a patchwork of laws that companies like Tesla, which has taken an incrementalist approach to rolling out self-driving features in a series of software updates, have exploited to do what they want. What little federal policy remains seems ineffective: The recent Biden stimulus plan laid out huge investments in electric vehicles, rather than placing a renewed emphasis on public transport.
What we need are not more energy-efficient cars or self-driving cars or, as Zimmer fantasized, privately owned fleets that are available for hire, but fewer cars entirely. We need people biking, walking, taking buses and trains and subways, or otherwise riding in something besides a free-ranging, 3,000-pound metal exoskeleton with an error-prone operator, digital or human. If safety is often cited as a prime reason for developing autonomous vehicles—about 36,000 Americans die in car accidents every year—then perhaps a better way of saving lives is to have fewer cars on the road, replaced by mass transit and other public options.
It may be that autonomous vehicles, as once conceived, aren’t possible, at least not without giving over our roads entirely to driverless cars. Elon Musk, the Tesla CEO and industry clown prince, allowed a note of doubt to creep in recently when he tweeted, “A major part of real-world AI has to be solved to make unsupervised, generalized full self-driving work, as the entire road system is designed for biological neural nets with optical imagers.” One implication might be that “full self-driving” and human drivers simply can’t coexist, at least not on this “road system.”
Any uncharacteristic circumspection Musk displayed may have been prompted by a recent Tesla crash in Texas, in which two people died. (It’s one of 27 federal investigations into Tesla crashes, according to The Wall Street Journal.) After initial reports claimed that the car was engaged in self-driving mode when it crashed, despite no one being in the driver’s seat, Musk denied that the car was even self-driving-capable. Further reports have been muddled by the way a Tesla’s self-driving features work, since they incorporate two separate systems, one for cruise control and one for steering, which must be activated separately. It’s possible that the car owner had activated one but not the other, causing the car to crash into a tree and burst into flames.
The ongoing uncertainty around this event—to the point where we cannot even trust the Tesla CEO’s word on the matter—signals how unready autonomous vehicles are for prime time. That is not to say that private test tracks can’t host autonomous vehicles, but these vehicles are simply not capable of tackling the “novel situations”—an animal crossing the road, an overturned truck, a pedestrian crossing somewhere besides a crosswalk—that a human driver encounters regularly. “Human drivers meet novel situations all the time but can handle them thanks to their understanding of how the world works in general,” noted the writer Ben Dickson. “But the AI’s actions will become unpredictable when faced with edge cases,” which are unlikely to be found in the data set on which an autopilot was trained.
In contrast to all the hype and expense of autonomous vehicles—along with the regulatory and technical and economic uncertainty—bicycles, sidewalks, and public transport work exceedingly well. They are known quantities that can be strengthened and made more accessible with a fraction of the resources that have been poured into A.V. research. They are safe and enhance the urban experience. They contribute few, if any, emissions and don’t enrich a handful of oligarchs at the expense of the common interest.
Perhaps we should see the autonomous vehicle dream for what it is: a Disneyland-style spectacle that can’t live up to its sci-fi imaginings, a series of very expensive and glitzy pilot projects that can’t cut it in the real world. Naturally, Elon Musk offers a useful example. Recently, the Boring Company, one of Musk’s many ventures, unveiled a “Convention Center Loop” in Las Vegas. Surrounded by colorful LED lighting, a Tesla eased through a narrow tunnel that seemed to leave little space for an emergency exit. The project was patently ridiculous—slower than advertised, low capacity, expensive, dependent on human supervision, and possibly dangerous—but it received all the usual hype of any Elon Musk production. Still, it doesn’t matter if it works as promised: The Boring Company already has contracts with the city of Las Vegas to build tunnels all over town—just don’t ask him when.