After 50 million miles, Waymos crash a lot less than human drivers

Waymo has been in dozens of crashes. Most were not Waymo's fault.

See full article...

 

[ColdWetDog]

That's it! Everybody out of the driver's seat! Trust the computer!

(Thanks for the data Mr. Lee, it's something I've really wanted to see for some time.)

 

[deus01]

Having recently taken waymo in LA this is not surprising. It seems a bit more cautious than a typical cab/Uber driver and also isn't on the phone of watching a TV show on their phone (which one of my Uber drivers in LA was doing)

 

[mschira]

We all know Tesla's autopilot is WAY WAY better!
Fantastic cars, fantastic.
/s

 

[ItchyPoo]

Glad to see the data and certainly looks promising. I had questions on some of the incidents as I’m sure Lee did, but overall I would love to see Waymo come to Denver. I don’t drive or own a car, so hailin a Waymo if less costly (hopefully at some point at least) than an Uber would be great.

one question… does Waymo mostly/only do city driving as opposed to highway? Wondering how well comparisons are in that regard.

Edit: also curious on the progress in snow? I think I read they might do Washington DC soon, but so far seem to be in areas without snow.

 

[mn_pianophile]

They should try a cold weather city next, like, say Minneapolis. #minneapolitan

 

[Daniel]

As someone who doesn't exactly have a passion for driving, I'm glad to see this kind of data and would welcome them coming to my town. I've had some wildly dangerous trips using Uber so it at least sounds a lot safer than that.

 

[WDoyle]

Are they still having the occasional gaggle of Waymos clog up a street and need Waymo employees to come out and unclog their vehicles manually?

 

[Hopefully Smarter]

What cars? Oh, you mean Tesler!

Got mine at Dollar Bill's Ezy Autos. Comes with a built in lighter.

 

[Tim Lee]

Glad to see the data and certainly looks promising. I had questions on some of the incidents as I’m sure Lee did, but overall I would love to see Waymo come to Denver. I don’t drive or own a car, so hailin a Waymo if less costly (hopefully at some point at least) than an Uber would be great.

one question… does Waymo mostly/only do city driving as opposed to highway? Wondering how well comparisons are in that regard.

Edit: also curious on the progress in snow? I think I read they might do Washington DC soon, but so far seem to be in areas without snow.

They have been testing driverlessly on freeways for about a year now but the commercial service doesn't operate on freeways yet. I expect that will start to happen this year.

I think (but I'm not sure) that the human benchmark from Waymo's safety hub exclude freeway miles. The insurance data includes freeway miles in its human benchmark, but they argue that this actually makes the benchmark conservative since the rate of crashes per mile tends to be lower on freeways.

Waymo has not demonstrated snow driving yet. My guess is that for the first couple of winters they'll just suspend service when there is snow on the roads. This only happens 5-10 days a year in DC.

 

[TracklessDeep]

Waymos operate under a wildly unnatural set of driving restrictions regarding speed, conditions, and locations.

I’m curious what you mean — the cars are driving in real-world cities. They compared safety records to actual humans in those cities.

If you mean it doesn’t necessarily translate to the same performance in locations with worse conditions — poorer roads or weather — then I agree.

 

[schteeve10]

It's good to see that Waymos aren't clearly worse than human drivers, but I don't think it's so clear-cut that they're clearly better. Waymos operate under a wildly unnatural set of driving restrictions regarding speed, conditions, and locations. Waymo is half research project, half PR campaign orchestrated by a trillion dollar company. When someone with a trillion dollars shows you something and says that reality exactly lines up with their corporate interests, you should be more than a little suspicious.

I get where you're coming from regarding corporate interests, and we should certainly always take promising data with a grain of salt, but I think you're off base about the driving restrictions. It's operating on the same roads under the same conditions as thousands of other human drivers on a daily basis, and the only reason it's racked up dozens of accidents is because other HUMAN drivers keep hitting it. Here are some things that autonomous cars don't do (some of which others have hinted at already): glance at someone walking down the street for too long, text while driving, fiddle with the AC/radio while driving, fall asleep, recklessly tailgate or swerve around in traffic because they're impatient... I could go on. So while Waymo's cars may not be quite up to par with a GOOD human driver in ALL situations, I'd feel much safer driving around a bunch of Waymo's than most human drivers these days.

 

[daemonios]

The hard data are very interesting and make me a tiny bit less suspicious of autonomous driving. Still, something in me resists becoming a cheerleader for the tech. For instance, with digital infrastructure and services becoming the new battlefield and with the world much more belligerent, what's to stop catastrophic interference or takeover of Waymo or other autonomous vehicles? The incentive to do so will only increase, proportionally to the adoption of the tech. Maybe these fears have already been considered and the systems don't pose any single point of failure. I'd love to see a feature on the safety and security measures in driverless cars.

 

[Cthel]

What, no breakdown of how many trompe l'oeil tunnels Waymo tried to drive though?

 

[Erbium68]

What cars? Oh, you mean Tesler!

It's all computers. But one. And don't sell your shares, helots.

Pity the BYD one is apparently just as good as the South African one, and doesn't come with a price premium.

Incidentally Nissan's development vehicle is now achieving up to 60mph on British B roads and if you know what that means you will be impressed. But the LIDAR cost needs to come way down.

 

[Erbium68]

If only we could run a city wide experiment, like hard limit of let say 50mph and then see in a week what would be results compared to normal driving week in terms of accidents and victims.

maybe autonomous cars are not final solution.

With all new systems in cars, i would not be sure we are not so many years away to geofencing and enforced throttling when entering an designated area. Cars talking between themselves, with traffic lights etc. Brave new world.

The experiment is being done in London where my speed limiter is on 20 most of the time.
Accident rates are down but you know enough autonomous cars should actually allow speed limits to increase if the Waymo data proves reliable. It's like you could never maintain current levels of air traffic without modern navigation aids, too many pilot errors.

 

[Erbium68]

What, no breakdown of how many trompe l'oeil tunnels Waymo tried to drive though?

This is why you need something in the vehicle which sees what's really there, like radar or LIDAR.
Remember Musk demanded radar be removed from Teslas? And now he's admitting they cannot get to FSD with the current technology?

 

[robert.walter]

Kudos to the team John Krafcik built !

 

[MGAubrey]

But I just want trains and better public transit.

 

[tratclif]

I'd just recently run across Sophia Wang's videos of the antics at the Waymo parking lot next to her apartment building. I think there's some devils in the details.

 

[Erbium68]

OK, but 20mph seems more like a punitive number, not realistic attempt to limit higher ranges that are far more dangerous in terms of damage and injury.
I have no beef with Waymo or autonomous vehicles, just curious to see if reasonable speed limit would have disproportionate effect in regards to accident number and effects.

That comment sounds like our far-right Reform party in action. They have this weird obsession that London's Mayor is trying to persecute them.
In fact average speed in London is around 13mph due to traffic, so a 20mph limit has almost no effect on traffic speeds, but it does make it possible to catch and prosecute dangerous drivers, and it makes bicycles and pedestrians safer at junctions.

 

[SnakeJG]

So even if both of the pending claims against Waymo succeed, two injuries represent a more than 90 percent reduction in successful injury claims relative to typical human drivers.

Waymo and Swiss Re estimate that human drivers in the same geographic areas would have generated 78 property damage claims. So Waymo generated 88 percent fewer property damage claims than typical human drivers.

The term "typical human drivers" is doing a lot of heavy lifting. From this (possibly biased site about teen driving: https://www.safetyinsurance.com/driversafety/tips_statistics.html )

• Nationwide, 43 percent of first-year drivers and 37 percent of second-year drivers are involved in car crashes.
• Individuals 15 to 20 years old make up 6.7 percent of the total driving population, but are involved in 20 percent of all crashes and 14 percent of motor vehicle deaths.

So, obviously new drivers are involved in more than their fair share of crashes. My google skills are failing me at the moment, but I suspect that it is also very likely that accidents are predominately caused by a population of "bad" drivers and if you take out those outliers, the "typical" human will do a lot better.

Counter point, I bet a lot of those outliers are driving when tired or drunk, so getting them into a Waymo instead would be best for everyone.

 

[sethrj]

missed opportunity for title: “human drivers crash way mo than Waymo”

 

[Ajar]

So all of the data cited in the article either came from Waymo or from research Waymo sponsored?

Waymo's estimate of how many accidents human drivers on similar roads would get into are useful, but I'd definitely want to see independent estimates before taking them at face value.

Waymo does seem to be the company taking the "slow and steady" approach to this, at least, which I think makes them more likely to succeed in the long run than other companies that underestimate just how hard this problem is.

 

[Danathar]

I’ve always thought that the Rubicon for self-driving vehicles will be when your insurance company asks whether you plan to drive or let the computer drive, and charges you more per month to choose manual driving over automated driving.

 

[solomonrex]

As someone who doesn't exactly have a passion for driving, I'm glad to see this kind of data and would welcome them coming to my town. I've had some wildly dangerous trips using Uber so it at least sounds a lot safer than that.

I'd love to see them on restricted routes, aka, replace bus drivers first, where the standards have plummetted since COVID. Overly cautious driving is perfect for buses/shuttles, and you can still have an attendant staring at their phone for rider management and safety concerns. And it's also natural to have it monitored over video/GPS from centralized locations.

 

[TheMolesRevenge]

Incidentally Nissan's development vehicle is now achieving up to 60mph on British B roads and if you know what that means you will be impressed. But the LIDAR cost needs to come way down.

As someone who drives on British B roads on a daily basis, and in an area where getting up to 60mph on half of the A roads is pushing your luck, that is both impressive and frightening!

 

[redleader]

It's good to see that Waymos aren't clearly worse than human drivers, but I don't think it's so clear-cut that they're clearly better. Waymos operate under a wildly unnatural set of driving restrictions regarding speed, conditions, and locations.

Having ridden in them a few times, they're basically like a moderately confident human driver that actually follows traffic rules.

That makes them both wildy unnatural and safer than the average driver. If driving cautiously makes them "better" is a more subjective question that's harder to answer from data alone.

 

[fenris_uy]

Three crashes occurred while Waymo was dropping a passenger off. The passenger opened the door and hit a passing car or bicycle. Waymo has a “Safe Exit” program to alert passengers and prevent this kind of crash, but it’s not foolproof.

Time to change the doors from normal doors to sliding doors for all exits.

In another incident, a pedestrian walked in front of a stopped Waymo. The Waymo began moving after the pedestrian had passed, but then the pedestrian “turned around and approached the Waymo AV.” According to Waymo, the pedestrian “may have made contact with the driver side of the Waymo AV” and “later claimed to have a minor injury.” Waymo’s report stops just short of calling this pedestrian a liar.

Doesn't Waymo have cameras all over their car?

 

[Cthel]

Time to change the doors from normal doors to sliding doors for all exits.

Unfortunately, that would probably just shift the damage from the vehicle door to the just-disembarked customer.

Having an interior layout that allows every seat to exit from a curb-side door seems safer.

 

[JanneM]

But I just want trains and better public transit.

I can definitely see how self-driving buses and cars can be a good part of a public transport system, covering the large catchment areas where there isn't enough passenger density to handle with regular buses or rail.

 

[fenris_uy]

Unfortunately, that would probably just shift the damage from the vehicle door to the just-disembarked customer.

Having an interior layout that allows every seat to exit from a curb-side door seems safer.

AI can't beat Darwin.

I can of understand not looking while opening a door, I have seen it several times in cars with human drivers, specially if it's a curb side door into a bike lane.

But not looking while your body leaves a car, that's harder to justify.

 

[Bongle]

Doesn't Waymo have cameras all over their car?

Yes, doing crimes or attempting fraud next to an AV seems like a really stupid idea. "Look at me diving like a soccer player in 10 different wavelengths and from 20 different angles!"

 

[yumegaze]

it will take a while (think a decade or so) until a waymo-like service reaches where i live, but i'm curious about how well they handle narrow, busy streets. because if it's good to go, i would 100% take a waymo instead of an uber; uber drivers are insane over here. if i'm in a choosing mood though, quality public transport would be even better but... a guy can dream!