Earlier this month, a woman in San Francisco was hit by a car while crossing the street. Had the story ended there, it would have been just another one of the small tragedies that occur on America’s roads, where roughly 100 people die every day. But this woman’s body ricocheted into another lane of traffic. She was hit again, this time by a robotaxi from the start-up Cruise. The car braked, coming to a stop with her pinned underneath. Then it started driving again, dragging the woman along with it for an agonizing 20 more feet. The woman, whose identity has not been made public, remains in the hospital, in serious condition.
Since driverless cars from Cruise and its competitor Waymo started taking paid passengers in San Francisco this summer, they have been entangled in a series of high-profile hiccups—including a collision with a fire truck and the wrath of protesters, who have placed traffic cones on their cars’ cameras. A bad few months for America’s robotaxis has now gotten considerably worse: On Tuesday, the California Department of Motor Vehicles suspended Cruise’s license to operate its driverless cars in the state, contending that they are “not safe for the public’s operation” and that the company “misrepresented” safety information. (The DMV has accused Cruise of not showing officials the full video footage from the accident involving the woman, which Cruise has denied.) Last night, Cruise announced that it was voluntarily pausing its driverless operations nationwide.
Waymo’s robotaxis are still roaming the streets of San Francisco and now several other cities, but the path to a world in which self-driving cars are everywhere, chauffeuring us around while we nap, still feels far away. Even so, the machines may have already won. Many new cars from major manufacturers such as General Motors, Ford, and, yes, Tesla already have advanced autonomous features that can control many parts of the driving process without a human touching the wheel. And more partly autonomous cars are coming. Your next car won’t be driverless, but it might be driver-ish.
A recent ad for Ford’s BlueCruise system is a glimpse into this future. It shows a pregnant Serena Williams driving a Lincoln Navigator and, as the music swells, dramatically letting go of the steering wheel. With BlueCruise, a driver can go hands-free on the highway as the car stays in its lane and keeps a distance from other vehicles. The latest version can help the driver change lanes, with the vehicle doing so automatically after the turn signal is tapped. BlueCruise has been integrated into models such as the Ford Mustang SUV and the Ford F-150, the latter of which is among the best-selling cars in the country. General Motors also has a hands-free tool as part of its Super Cruise system; available on models such as the Chevy Volt EUV and the GM Suburban, it can likewise independently stay in a lane and change lanes on demand.
Both of these are what are called “level two” systems. Autonomous driving can be broken down into six different levels, from zero to five, according to a classification system created by the Society of Automotive Engineers International. A level-zero vehicle might have a basic feature such as automatic emergency braking, whereas a level-five vehicle can fully drive itself anywhere. It may not even have a steering wheel. The robotaxis made by Waymo and Cruise are level-four systems: No one is behind the wheel, but the cars are geo-fenced and limited to specific driving conditions.
Carmakers have long had souped-up cruise-control tech that meets the bar for level one, but now level-two and level-three systems are popping up in the cars they sell. The most well known is Tesla’s Autopilot and the more advanced “full self-driving” beta, in which the car drives completely on its own. Despite the name, it is still a level-two system, meaning a human must constantly monitor the driving. (Because anything that requires consistent human supervision is level two, this stage encompasses a huge range of potential autonomous features.)
Then there is level three, which is like the awkward middle child of autonomous systems: Everything below it requires the driver’s full attention, and everything above it doesn’t require any attention at all. At this stage, the driver must be able to drive when prompted by the system, but she can otherwise take her eyes off the road. Last month, Mercedes-Benz became, with its Drive Pilot feature, the first automaker to introduce a level-three system in the United States. Drive Pilot can be used only in certain conditions (not at night or in the rain) and only in certain areas (California and Nevada have both approved it). But when activated, it’ll be able to navigate road signs and traffic. As it does, you can lean back and play Tetris, which is included in the cars’ entertainment system.
Right now these autonomous features are primarily loaded into higher-end cars, says Paul Waatti, an industry-analysis manager at AutoPacific, a market-research company. Mercedes’s Drive Pilot system has an annual subscription fee of $2,500 a year, on top of the $100,000-plus sticker price for the all-electric EQS sedan that is equipped with it. Tesla’s full-self-driving mode can cost up to an additional $200 a month. Still, Waatti expects these features to trickle down to more automakers’ fleets in the coming years. BMW, Volvo, and Stellantis (the manufacturer of Jeep and Dodge) are also working on level-three technology.
Forget about the robotaxis for a moment, and start looking out for robo-creep: smaller but still powerful autonomous systems subtly taking over parts of the human driving experience. “We’re going to continue to see more autonomous features come into play,” Waatti told me. “But we’re not going to see fully autonomous cars for quite some time, especially that are going to be able to be sold to the public.”
These features are taking off in part because car companies are seeing dollar signs in offering them for a monthly fee, which is what Mercedes and Tesla are doing. “The secret that the car-company executives won’t tell you,” Reilly Brennan, a founding partner at the transportation-focused venture-capital fund Trucks, told me, is that compared with driverless cars, these driver-ish cars are “a much better business for them, because it allows them to sell both hardware and software.” Research from the consulting firm McKinsey found that about half of global consumers would be willing to pay up to $9,999 for these features, or that much in monthly fees.
A future in which everyone around you on the freeway is quietly playing Tetris in their car might be a ways off. Tesla’s Autopilot feature has been linked to at least 17 fatal crashes and has faced lawsuits. Drivers aren’t made to cede control to computers but still be ready to take back control at a moment’s notice. For safety reasons, GM’s Super Cruise has a camera over the dashboard that tracks head and eye movement, and the machine deactivates if it thinks a driver is distracted; other autonomous systems have similar mechanisms.
Most drivers will have to keep paying some attention for the foreseeable future. Ani Kelkar, an associate partner at McKinsey, told me that, in the best-case scenario, the company projects that 12 percent of new passenger vehicles will be level three or higher by the end of the decade. By contrast, he said, McKinsey expects level-two or higher vehicles to make up the majority of those sold, and 65 percent in the best-case scenario. Drivers will remain behind the wheel for now; they just might do less of the actual driving. The machines will start rounding off the corners, perhaps taking over in annoying stop-and-go traffic, mindless highway road trips, and the nuisance that is parking. It’ll ease the burden of driving, but you won’t be totally off the hook.
None of this is to say that Waymo or Cruise or another self-driving-car company won’t also eventually figure out how to crack the code to robotaxis. But their challenge is bigger. Level four—what the robotaxis are attempting—“is just tremendously difficult” and “humongously expensive,” Ramanarayan Vasudevan, a professor of mechanical engineering and robotics at the University of Michigan, told me. These cars are costly to build and costly to scale, and companies need to figure out how to recoup all of these investments while also charging fees that can compete with ride-hailing apps and taxis. (According to The Wall Street Journal, Cruise lost $1.9 billion from January to the end of last month.) And that’s still just level four, where the cars are stuck in certain cities and neighborhoods. “Level five is a science experiment,” Vasudevan said. He doubts it’s possible.
Of course, driver-ish cars might not be as revolutionary as driverless ones, which promise to cut into problems such as drunk driving. But level-two and level-three systems are here right now, parked on the lots at your local dealership—or maybe even in your driveway. Over time, a few autonomous features will become many autonomous features. And then you’ll have far more time to play Tetris on the highway.