California suspension is an existential threat to Cruise. By Timothy B Lee

www.understandingai.org

8 - 10 minutes

On Tuesday, the California Department of Motor Vehicles ordered Cruise, the self-driving subsidiary of General Motors, to suspend all driverless operations in the state. It’s a major setback for Cruise because until this week the company had far more vehicles active in San Francisco than any other metro area.

The announcement follows a grisly October 2 episode involving a pedestrian and a Cruise vehicle. The pedestrian was struck by a human-driven vehicle and thrown in front of the driverless Cruise taxi. The Cruise slammed on its brakes but was unable to avoid hitting her.

The California DMV reports that after coming to a complete stop, the Cruise autonomous vehicle “attempted to perform a pullover maneuver while the pedestrian was underneath the vehicle. The AV traveled about 20 feet and reached a speed of 7 mph before coming to a subsequent and final stop.”

She wound up trapped under the rear wheels of the Cruise. The San Francisco Fire Department eventually used “jaws of life” to lift the car and free the woman. She remains hospitalized with serious injuries, according to the San Francisco Chronicle.

Cruise’s initial tweets the day after the crash didn’t mention its vehicle dragging the woman as it pulled over. Indeed, the DMV says that when it met with Cruise the day after the incident, the company only showed footage from the initial crash, leaving out the part where the victim got dragged under the vehicle as it pulled over.

Cruise disputes the DMV’s account of the meeting, telling Vice that it played the full video “multiple times” during the company’s October 3 meeting with the agency. But the company now admits to its vehicle “pulling the pedestrian forward.”

The DMV’s order suspending Cruise’s driverless operations in San Francisco doesn’t just cite the Cruise vehicle’s poor handling of the crash. It also cites the company’s subsequent duplicity. Under California law, the DMV can suspend an autonomous vehicle maker if it “has misrepresented any information” about the safety of its technology.

Bryant Walker Smith, a legal scholar at the University of South Carolina, argues the incident raises questions about the honesty of Cruise executives. If they deliberately misled California officials about the crash, he told me by email, that “should be disqualifying.”

“Automated vehicles are only as safe as the companies behind them,” he added. “If we can't trust Cruise about this crash, then we shouldn't trust Cruise about anything. And an AV company that isn't trustworthy has no business being on our roads.”

Cruise faces an uphill battle to regain the trust of California regulators. (Image courtesy of Cruise)

While Cruise started out in San Francisco, the company has much broader ambitions. The company has announced plans to operate in more than a dozen cities, including Atlanta, Miami, Las Vegas, and Washington DC. Cruise needs to convince government officials and the general public in all of these metro areas that its technology is safe. A good way for any company to build trust is by being transparent. But Cruise hasn’t done that—either in this incident or previous ones.

When Cruise reported the October 2 crash to federal regulators, it wrote that its vehicle “braked aggressively but, shortly thereafter, made contact with the pedestrian.” The regulatory filing doesn’t mention that after coming to a stop, the vehicle pulled forward another 20 feet with the victim still underneath.

And it’s hard to believe Cruise didn’t know about this because Cruise vehicles are festooned with cameras and other sensors. If Cruise wanted to it could release video footage of incidents like this to the general public. Instead, Cruise’s practice has been to show video footage to a hand-picked group of reporters over Zoom—and to prohibit those reporters from recording the video or sharing it with others.

This gives Cruise an opportunity to present the footage in a light favorable to Cruise. For example, reporter Peter Bigelow tweeted that Cruise showed him footage that “ended at the initial strike. It did not show the additional movement” as the vehicle dragged the injured victim to the curb. Reporters from the San Francisco Chronicle reported the same experience.

This doesn’t surprise me. Last month, I was treated to a similar viewing opportunity while writing about an accusation from the San Francisco Fire Department that a Cruise vehicle blocked the path of an ambulance trying to take an injured patient to the hospital (the patient died soon after arrival). Rather than sending me video from the incident, a Cruise representative played it for me using a Zoom screen share—and demanded that I not try to capture the video.

The video did contradict some of the specific claims made by a SFFD memo about the blocked ambulance. At the same time, I found it striking that Cruise seemed unwilling to clearly acknowledge the obvious: that its vehicles had malfunctioned and blocked traffic at a crash scene for several minutes. While I don’t think this led to the death of this particular victim, it’s easy to imagine similar malfunctions having deadly consequences in the future. Yet I couldn’t get Cruise to tell me anything about why the vehicle was stuck for so long or how—if at all—Cruise planned to prevent similar malfunctions in the future.

Back in April, Cruise published a report analyzing the company’s first million miles on the road. Last month I asked Phil Koopman, a transportation safety expert at Carnegie Mellon, about a chart from that report purportedly showing that Cruise vehicles crash significantly less than human drivers.

“These metrics are so cherry-picked and the baseline is so opaque, that bar chart means nothing,” Koopman told me. “What we see from Cruise is a conspicuous display of opacity.”

Losing access to the San Francisco market wouldn’t be a disaster for Cruise on its own since Cruise is planning to offer service in so many other markets. Many of these cities are in right-leaning states like Texas, Florida, Georgia, and Tennessee, which means they may not face the same level of regulatory scrutiny Cruise has faced in liberal California.

But while Republicans may be willing to cut Cruise more slack than Democrats, their patience won’t be unlimited. Back in 2018, Uber’s self-driving program was effectively destroyed after one of its vehicles struck and killed pedestrian Elaine Herzberg in Tempe, Arizona. Uber was forced to drastically scale back its testing operations in Arizona and elsewhere, and Uber’s self-driving effort never really recovered.

Cruise’s main rival, Waymo, is also operating commercial taxi services in Phoenix and San Francisco. Waymo’s technology isn’t perfect, but it has suffered fewer high-profile safety incidents in recent months. Waymo has been more transparent about its safety record and more proactive about explaining that record to journalists.

Waymo also seems to be expanding more cautiously. Whereas Cruise has announced plans to operate in more than a dozen cities, Waymo has only named four: San Francisco, Los Angeles, Austin, and Phoenix. Waymo has expanded relatively slowly despite the fact that it has logged more than 35 million miles of testing with safety drivers behind the wheel—several times more than Cruise.

And that’s why the decision of the California DMV could be so damaging to Cruise. It’s not just that Cruise will lose access to the California market. It’s that the crash—and allegations of a subsequent cover-up—could cement Cruise’s reputation as a self-driving company that doesn’t take safety seriously. Such a reputation could be a huge handicap in the coming years as Cruise tries to convince policymakers to allow it to operate in cities across the country—and eventually around the world. The challenge for Waymo is to convince those same officials that its own technology doesn’t suffer from Cruise’s flaws.