Fleet response providing the Waymo Driver additional context (Image: Waymo)
Waymo is overdue to confess that its fully autonomous vehicles aren’t totally fully “humanless.” Waymo’s robotaxis get help from remote operators who see the real world through “human” eyes and answer calls from robotaxi machine drivers, or from passengers inside the vehicle. If necessary, these flesh-and-blood remote supervisors can move the car out of a pinch.
However, don’t assume that these humans are in the same city where Waymo is operating. Teleoperators assisting Waymo’s robotaxis operating in Arizona, for example, are located as far away as the Philippines.
Don’t get me wrong, though.
Having remote teleoperators is not a bad thing. It allows robotaxis to ask questions and get human help when unusual events crop up on the road.
Take, for example, a recent incident early in the morning, when a Waymo robotaxi with passengers inside drove through the scene of a police standoff in downtown Los Angeles. The video of the incident went viral on social media.
Police officers were conducting a high-risk felony arrest after a stolen vehicle pursuit. As you can see in the video below, the area was blocked by several police vehicles. Lights were flashing. Police were shouting. A man lay prostrate on the ground as the Waymo robotaxi drove blithely past him.
Was the robotaxi aware of what it was entering? And did it know how to behave in such an unusual situation?
Moreover, did the machine driver call its remote human operator? If so, when did it call? And how did the supervisor respond?
When I reached Waymo for answers for these questions, the company came back with familiar platitudes.
“Safety is our highest priority at Waymo, both for people who choose to ride with us and with whom we share the streets. When we encounter unusual events like this one, we learn from them as we continue improving road safety and operating in dynamic cities.”
Missed opportunity
The lesson from this statement is that Waymo understands, at least, that its robotaxis encounter “unusual events” that they “learn from.”
But given that it won’t be possible to train a driverless car on every single edge case—like police and bank robbers shooting at one another across a city street—this incident posed a juicy opportunity for Waymo to stress the significance of having remote human operators.
However, Waymo offered no clear answer about whether the robotaxi noticed the unusual situation that could be a threat to its passengers and called the remote operators or, if it called, when it did so.
The company told me:
“The Waymo Driver completed its actions independently of fleet response input. While the car can reach out to human fleet response agents for additional contextual information, the Waymo Driver continues using available information to execute driving tasks. Therefore, the vast majority of situations are resolved, without assistance, by the Waymo Driver.”
Phil Koopman, safety expert and professor emeritus at Carnegie Mellon University, takes issue with Waymo’s statement. He said, “When we encounter unusual events like this one, we learn from them as we continue improving road safety and operating in dynamic cities.”
He said, “That might be true in the broad sense, but we know it might take many such events for them to learn.”
In other words, learning might not happen, per the company’s claim, as a one-and-done fix with software updates.
Koopman is referring to Waymo’s illegal school bus passing issue. Since the start of the school year, “Waymo self-driving cars in Austin, Texas, have passed school buses with their flashing red lights and stop arms extended 19 times,” according to the Austin, Texas TV station KVUE.
This snafu occurred despite software updates Waymo claimed that it had implemented.
Koopman explained, “What matters for public safety in the here-and-now is not whether perhaps someday they will get high-risk events right. But rather whether they will get events right the first time they encounter them.”
In his opinion, “That is what the remote supervisors are for. The fact that Waymo says a remote supervisor was not involved with a novel high-risk situation indicates a failure of their safety strategy.”