Tech leaders promised we'd be riding in driverless vehicles by now. What happened to the autonomous auto revolution?

When did work start on these cars? 
Almost a century ago. In 1925, an experimental remote-controlled car cruised down Manhattan's Broadway in front of bemused onlookers — only to crash into another vehicle. In the 1950s, the Radio Corporation of America built a prototype highway embedded with wires that could control a car's steering and speed, with the aim of letting drivers relax behind the wheel and giving authorities better control of traffic flows. But officials balked at the cost: up to $200,000 per lane-mile. Interest in self-driving cars accelerated along with advances in computing and wireless communications, and from 2010 to 2021 auto and tech giants such as Google and General Motors collectively spent an estimated $100 billion on driverless vehicles. Yet despite grand promises from industry leaders like Tesla CEO Elon Musk, who in 2016 called autonomous driving "basically a solved problem," no self-driving system has yet achieved the goal of "full automation" — driving to any destination without any human input. "This may in fact never be possible," said Ragunathan Rajkumar, a professor of computer engineering at Carnegie Mellon University, "or it is at least many years or decades away." In a sign of the difficulty of making self-driving cars a reality, Apple last week killed its decade-old autonomous electric car project. 

How do self-driving cars work? 
Most autonomous driving systems, including those developed by Google's Waymo division and GM's Cruise, use cameras, radar, and lidar — light detection and ranging — sensors to construct a 3D map of a vehicle's surroundings. (Tesla is an exception: Its vehicles rely solely on cameras.) Data collected by onboard sensors and cameras is fed into an artificial intelligence model that directs the vehicle and decides the best and safest way to navigate around obstacles, including other cars and pedestrians. But while humans are generally good at dealing with the unexpected, the AI must be taught how to react to countless potential hazards, which has slowed the evolution of the technology. How should a driverless car respond, for example, if a dog runs into the road, or if a bicyclist is wobbling dangerously, or if snow obscures white line markings? "The problem is that there isn't any test to know if a driverless car is safe to operate," said Mike Ramsey, an analyst at market researcher Gartner. 

Is the technology safe? 
The industry says self-driving cars are up to seven times less likely to get into crashes leading to injury than normal cars, which means the technology could help cut the U.S.'s 43,000 annual road deaths. That seems to make sense, because AI doesn't drink and drive, and never dozes off. But road users in cities where driverless vehicles are being tested say the cars often behave erratically, such as by braking inappropriately and failing to move aside for emergency vehicles. A string of accidents has also dented confidence in the industry's safety claims. In 2018, Elaine Herzberg became the first pedestrian to be killed by a self-driving vehicle, after she was hit by an Uber test car in Tempe, Ariz. It's thought the car became confused when Herzberg stepped into the road while pushing a bicycle with bags on its handlebars — an array of objects it couldn't interpret. And last fall, a woman in San Francisco was left critically injured after she was hit by a human-driven car and knocked into the path of a Cruise "robotaxi." The taxi attempted a "pullover maneuver" and dragged the pedestrian for 20 feet. Soon after, California revoked Cruise's permit to operate fully driverless vehicles on the state's roads and Cruise CEO Kyle Vogt — who a month earlier had claimed concerns over the technology were being "sensationalized" — resigned. 

Have there been other accidents? 
Since 2019, at least 17 deaths and 736 crashes have been linked to Tesla's Autopilot — driver-assistance technology that falls far short of full automation — according to National Highway Traffic Safety Administration data. Drivers are told to keep their hands on the wheel and stay observant when Autopilot is engaged, but many don't. Tesla's Full Self-Driving (FSD) system, which is supposed to guide the car almost anywhere, has also been connected to at least one death. In 2022, Tesla recruiter Hans von Ohain was killed when his Model 3 suddenly veered off a Colorado mountain road, smashed into a tree, and burst into flames. A passenger who survived the crash said FSD had been engaged; damage to the vehicle meant authorities couldn't confirm that claim. 

What comes next? 
Driverless "robotaxis" are now available for hire in cities such as San Francisco, Los Angeles, Phoenix, and Wuhan, China. But Waymo and Cruise are thought to be scaling back expansion plans after racking up huge losses. "You'd be hard-pressed to find another industry that's invested so many dollars in R&D and that has delivered so little," said engineer Anthony Levandowski, an early pioneer of self-driving tech. Still, the consultancy McKinsey estimates the market for autonomous-driving systems could be worth $400 billion by 2035, as many companies remain eager to replace their human drivers — who can get sick and need to be paid wages — with machines. For now, public distrust remains a major barrier to the technology's spread and success. Journalist Cade Metz likened his ride last year in a Cruise robotaxi to being in the car with his 16-year-old daughter when she was learning to drive. "But more unnerving," he said, "because my daughter could at least respond to my moments of panic."

Backlash in self-driving city 
For the past two years, San Francisco has served as a testing ground for hundreds of driverless taxis operated by firms such as Cruise and Waymo. Not everyone there is happy. In the space of six months last year, the S.F. Fire Department recorded 55 incidents of self-driving cars interfering with rescue operations, from driving through yellow tape to parking between a fire truck and a burning car to stopping on top of a fire hose — as  firefighters battled a blaze. Residents have also fumed about how the cars block city buses and occasionally stop dead in the middle of intersections. Some anti-driverless car activists have started placing traffic cones on the cars' hoods, which puts the vehicle into panic mode, shutting it down until a human employee can remove the cone. "I know the city is not likely to ban robot cars," one activist told The Guardian. "But what we are trying to do is move the needle in opposition to these vehicles. And it's working."

"The horror of it is beyond my capacity to describe," said Moira Donegan in The Guardian. On a recent Sunday, Aaron Bushnell, an active-duty member of the U.S. Air Force, approached the Israeli embassy in Washington, D.C., and propped his phone on the ground. As it sent a feed to a live-streaming site, Bushnell, 25, said he was "about to engage in an extreme act of protest" against Palestinian "genocide" committed by "colonizers." He then doused himself in flammable liquid and set himself on fire, screaming "Free Palestine." He died hours later. Bushnell's shocking act "opened a rupture in American political discourse," dividing even those who share his anger at the U.S. backing of Israel's war in Gaza. Some were horrified and questioned his sanity. But others hailed his sacrifice. Progressive presidential candidate Cornel West praised his "extraordinary courage and commitment"; Egyptian-American social activist Aya Hijazi called him "a hero and a martyr." 

To celebrate Bushnell's horrific display is "deeply sick," said Graeme Wood in The Atlantic. Self-immolation is a violent act, and "if you dislike violence, then you should abhor it no matter your view on the war." Bushnell's self-sacrifice is just one more act of "fanaticism" in a conflict overrun with it. And ironically, he shared no empathy for the Israelis who met the same end as he did, when they were burned alive by Hamas terrorists on Oct. 7. We should feel compassion for a man who was "clearly deeply unwell," said Brad Polumbo in the New York Post. But influential progressives glorifying him may inspire other misguided young people to give their lives "in this tragic, and ultimately futile, form of protest." 

"How do we judge what is reasonable and what isn't?" asked Shadi Hamid in The Washington Post. Nobody questioned the sanity of Mohamed Bouazizi — a Tunisian street vendor enraged by corruption and injustice — when he self-immolated in 2010, sparking the Arab Spring uprisings. In fact, then-President Barack Obama compared Bouazizi to civil rights icon Rosa Parks. Bushnell was engulfed by despair and a sense of complicity as he watched his country arm and support Israel in a war that's killed some 30,000 Palestinians, many of them women and children. You could argue that what he did was "unreasonable or even crazy." But it's not unreasonable that Bushnell felt so tortured by the bloody havoc in Gaza. "That, too, is unreasonable. It is worse than unreasonable." 

Lou Reed has been dead for a decade, said Walter Marsh in The Guardian (U.K.), but Laurie Anderson still spends a lot of time talking with her late husband. Or rather, with an artificial intelligence version of the alternative rock icon. An avant-garde artist, musician, and filmmaker, Anderson in 2020 spent a week at the Australian Institute for Machine Learning, where researchers fed a trove of Reed's writing, songs, and interviews into a computer. Anderson was then able to type prompts and the algorithmic version of Reed would riff back written responses.

"I'm totally 100 percent, sadly addicted to this," says Anderson, 76. "I can't stop doing it." Her friends "just can't stand it. I mean, I really do not think I'm talking to my dead husband — I really don't. But people have styles and they can be replicated." Anderson sometimes writes songs with the AI Reed. "Three-quarters of it is completely idiotic. Maybe 15 percent is like, 'Oh?' And then the rest is pretty interesting. That's a pretty good ratio for writing, I think." She says she wouldn't mind if someone used an AI to make Laurie Anderson art after her own death. "When people die and you hear their words or their music, it's like they're alive. You can say that about Dostoevsky, Plato, the Buddha. So, I feel fine about that. I mean, we're just mayflies, really."