Things just keep getting worse for Cruise, the troubled robotaxi company that once dreamed of being a leader in the autonomous driving industry. Only a month after a violent collision forced the company to ground all of its fleets nationwide, multiple news outlets have reported that the company’s “self driving” cars are…well…not actually driving themselves, all the time. Instead, the vehicles appear to be aided by remote human assistants, frequently as often as every four to five miles.
Cops Pull Over Self-Driving Car
Over the weekend, the New York Times dropped a story that alleged that Cruise’s vehicles were supported by a “vast operations staff” and that, prior to the company’s “pause” of operations, this staff frequently had to intervene to “do something to remotely control a car after receiving a cellular signal that it was having problems.” The Times report, while intriguing, didn’t provide a ton of detail about what that meant. Following the story’s publication, however, Cruise’s CEO, Kyle Vogt, slipped into the comment section at Hacker News and largely confirmed many of the report’s details. Vogt said:
Cruise AVs are being remotely assisted (RA) 2-4% of the time on average, in complex urban environments. This is low enough already that there isn’t a huge cost benefit to optimizing much further, especially given how useful it is to have humans review things in certain situations.
While 2-4 percent of the time may not sound like a lot, the company later provided additional details to CNBC. A Cruise spokesperson, Tiffany Testo, wrote in an email to the news site, that a “remote assistance” session typically occurs every four to five miles for the company’s vehicles. Testo continued:
“Often times the AV proactively initiates these before it is certain it will need help such as when the AV’s intended path is obstructed (e.g construction blockages or detours) or if it needs help identifying an object,” she wrote. “Remote assistance is in session about 2-4% of the time the AV is on the road, which is minimal, and in those cases the RA advisor is providing wayfinding intel to the AV, not controlling it remotely.”
Of the remote assistance advisors, Testo said that there was typically one remote assistant “for every 15-20 driverless AVs.” She added:
“RA advisors undergo a background check and driving record check and must complete two weeks of comprehensive training prior to starting, consisting of classroom training, scenario-based exercises, live shadowing and knowledge-based assessments. Advisors also receive ongoing training and undergo supplemental training whenever there is a new feature or update. Regular reviews, refreshers and audits are conducted to ensure high performance.”
While the company’s transparency here is admirable, the very existence of this operations center inspires so many questions. How, exactly, are these staffers intervening in the vehicle’s trips? What kinds of control does the remote assistant have over the car? What kind of digital security precautions has Cruise implemented (or not implemented) around the remote access software that allows for this to happen? How big is the remote access team? Gizmodo reached out to Cruise for additional information and will update this story if they respond.
At its most basic level, the revelation about Cruise’s remote operations center would appear to reveal more evidence that AI still doesn’t really function all on its own. Instead, largely invisible human workforces toil away in the background, doing indispensable if—in many cases—undervalued labor. While a lot of the details about Cruise’s remote operations team are unclear, it’s another reminder that “autonomous” machines still require adult (human) supervision.