There are hundreds of Waymo cars on Valley roadways. The company says they’ll also be returning to freeways, after a software upgrade to improve how the cars behave around construction zones.
One of the key questions surrounding the autonomous vehicles is: Are they safe? And perhaps more specifically, are they safer than human drivers? Waymo has put out data that suggests they are. And now, an independent study seems to largely confirm that.
David Kidd is vice president for vehicle research at the Insurance Institute for Highway Safety and one of the lead authors of the paper. He joins The Show to talk more.
Full conversation
MARK BRODIE: David, good morning.
DAVID KIDD: Good morning.
MARK BRODIE: So, is it a fair assessment that what you found in large part kind of mirrors what Waymo has said about the safety of its vehicles?
DAVID KIDD: It is, in large part. Our study, using a different method, found that for Waymo vehicles, they get in 68% fewer crashes than human drivers.
MARK BRODIE: And what are some of the limitations, though? I mean, I know that there are some things that the data really cannot tell us.
DAVID KIDD: So, the way that we did this study is we had to read a whole bunch, nearly a thousand crash narratives. Companies that have automated vehicles on the roadways are required to report crashes — used to be any kind of bump in the road, but now it’s things that cause at least $1,000 in damage or involve vulnerable road users like pedestrians or bicycles — to report it to the government. And so we used those narratives to be able to understand if this is something you or I would report to police so we could compare it to police-reported crashes from human drivers.
MARK BRODIE: And am I right there’s also an issue in terms of trying to figure out like how many miles an autonomous vehicle actually drives, like in terms of getting a sort of a ratio there?
DAVID KIDD: Absolutely. That is the most challenging part of all this. The government does not require these companies to report how many miles their automated vehicles travel. Waymo does it voluntarily. And that allows us to take the number of crashes that they’re involved in and divide it by the number of miles that they’ve driven to derive this rate. And we can compare that rate directly to crashes that we are involved in as humans that we report to police.
MARK BRODIE: So, you reference the 68% fewer crashes per mile that you found, and those are over a few cities, including Phoenix, Austin, LA and San Francisco. But in Phoenix, the number of decline is actually higher: 76% fewer crashes per mile. Any sense of why the data are even better in Phoenix than they are elsewhere?
DAVID KIDD: You know, the one thing that I need to stress is that so we looked at nearly a thousand of these crash narratives. Waymo, we only had 64 narratives that we used from across these different areas. So, to be able to say something definitive about “they operate better in Phoenix than they do in Austin or San Francisco,” we can’t put any strength behind a claim like that. It’s really we’re looking at the bulk of them together, what that means for safety.
MARK BRODIE: OK. What about the severity of the crashes? One of the things that Waymo talks about is the fact that even when there are crashes, the vehicles maybe are going more slowly when they hit somebody or something than a human driver. Did you find anything about the severity of the crashes?
DAVID KIDD: We did look at crashes where there was an injury or reported injury, and we found that for Waymo vehicles, that crash rate of injury crashes was 81% lower than crashes where humans were involved and where there was an injury. So, yeah, one of the things to note that these automated vehicles do if they’re programmed that way, which Waymo certainly does, is they follow speed limits.
They certainly are not impaired, they’re not distracted, and they — obviously being a computer — don’t have to worry about belt use, right? But they force their occupants to wear seatbelts. These are things that us as people, we struggle with. Certainly, we could do better, and that’s where a lot of crash deaths result from.
MARK BRODIE: Is there anything you can draw about trends that you saw in terms of the crashes that Waymos were involved in? Maybe things that the programming struggles with, or particular situations on the roadway that these vehicles struggle with?
DAVID KIDD: I think that one of the biggest things is they just struggle with interactions with human drivers. One of the most common type of crashes they were involved in were where the Waymo vehicle was struck in the rear by another driver. And another top that they were involved in often was T-bone or angle crashes, where you have people pulling out of a parking lot or people who are trying to make a turn in front of the Waymo vehicle.
And so they don’t necessarily behave like we would expect as human drivers, or the culture around a certain type of intersection or driving pathway. And those differences can violate our expectations and lead to some conflict when we interact with automated vehicles, even if they are more conservative and trying to be safer.
MARK BRODIE: That’s interesting. So, even though they are following the rules and behaving as sort of the ideal human driver would behave, humans, being humans, are not acting necessarily that way, which is leading to some number of these crashes.
DAVID KIDD: That’s right. And in some cases, you and I, we have to violate the rules in order to drive, right? There’ll be situations where there may be an obstacle in the road, and so we have to pull off the road a little bit to get around that obstacle. Those are situations that computer-driven vehicles may not be able to handle as fluidly as you and I.
MARK BRODIE: So, is it safe to say that, like, just as, you know, in a perfect world, we’d obviously love there to not be any collisions or any crashes, but, you know, humans, again, being humans, that’s probably not realistic. Safe to say that’s probably the case for autonomous vehicles as well?
DAVID KIDD: Oh, there’ll always be the case that these vehicles get into some sort of crash, right? The biggest benefit for these vehicles that these vehicles have is that because they follow the rules of the road, the base speed limits, they’re going to get people inside to buckle up before they are able to get a ride, that those crashes hopefully will be less severe, have less crash energy, and people will be safer in them.
Now, again, we’re in the early stages of all this. There’s limited deployments, mostly in urban areas. It’s typically in ideal conditions where it’s sunny out and the roads are well marked, and there’s been a ton of mapping done and a ton of testing. You and I, we drive in all sorts of conditions: nighttime, when there’s winter weather, probably conditions we shouldn’t drive in, right? But we do OK. And so there’s a long way to go with these automated vehicles, but the early signs are promising.
MARK BRODIE: Interesting. All right. That is David Kidd, vice president for vehicle research at the Insurance Institute for Highway Safety. David, thanks so much.
DAVID KIDD: Thank you.
-
As artificial intelligence has led to the demand for power hungry data centers to grow, so has public opposition to the tech facilities. Now city leaders are looking for new standards to ensure they are built more responsibly. -
With each passing day, it gets harder to figure out what’s real or a fake in our ever-evolving digital world. Take a video posted online that purports to show President Nygren sobbing outside of a Bashas grocery store. -
Investigations over Flock cameras in Arizona police departments are stacking up, and more cities and counties are pausing or changing their contracts. Sheriff Ross Teeple of Pinal County is one of them. -
Waymo is expanding its services in metro Phoenix. The autonomous ride hailing company is adding 55 square miles to its service area. -
Journalist Marin Scotten reports there may be a new option when it comes to how to power data centers.
