Waymo Says Its Driverless Cars Are Better Than Humans At Avoiding Crashes (teslarati.com)
- Reference: 0177252083
- News link: https://tech.slashdot.org/story/25/05/02/0433206/waymo-says-its-driverless-cars-are-better-than-humans-at-avoiding-crashes
- Source link: https://www.teslarati.com/robotaxis-are-already-making-roads-safer-waymo-report-reveals/
> Waymo Driver is already [1]reducing severe crashes and enhancing the safety of vulnerable road users. As per a new [2]research paper set for publication in the Traffic Injury Prevention Journal, Waymo Driver had [3]outperformed human drivers in safety , particularly for vulnerable road users (VRUs). Over 56.7 million miles, compared to human drivers, Waymo Driver achieved a 92% reduction in pedestrian injury crashes. It also saw 82% fewer crashes with injuries with cyclists and 82% fewer crashes with injuries with motorcyclists. Waymo Driver also slashed injury-involving intersection crashes by 96%, which are a leading cause of severe road harm for human drivers. Waymo Driver saw 85% fewer crashes with suspected serious or worse injuries as well.
"It's encouraging to see real-world data showing Waymo outperforming human drivers when it comes to safety. Fewer crashes and fewer injuries -- especially for people walking and biking -- is exactly the kind of progress we want to see from autonomous vehicles," said Jonathan Adkins, Chief Executive Officer at Governors Highway Safety Association.
[1] https://waymo.com/blog/2025/05/waymo-making-streets-safer-for-vru
[2] https://storage.googleapis.com/waymo-uploads/files/documents/safety/Safety%20Impact%20Crash%20Type%20Manuscript.pdf
[3] https://www.teslarati.com/robotaxis-are-already-making-roads-safer-waymo-report-reveals/
Beating humans (Score:1)
It's a low bar.
Re: (Score:2)
Low or not, it's the bar.
exactly the kind of progress we want to see from a (Score:2)
forgot to mention facial recognition and behavioural analysis because it's part of the TOS if you enter our vehicle?
Think of the children !!
VRU's vulnerable road users. (Score:5, Insightful)
Lately, I've missing George Carlin daily. People. Just say people or pedestrians. You know all those other road users who don't drive cars. I just want to slap the soulless people who come up with euphemisms like that to avoid acknowledging other peoples humanity so it doesn't look so bad to them when they kill someone.
Re: (Score:3)
"Ran over 2 VRUs today", no, that doesn't sound so bad. Makes it sound like speed bumps.
Re: VRU's vulnerable road users. (Score:3)
I'd say that's not the only thing missing.
Re: (Score:2)
I don't mind it as a class.
It includes bicycles and motorcycles.
Things that are smaller, perhaps move different than what drivers are used to (pedestrians especially) and squishier than cars.
it makes sense to have a designation for "will be killed by car without harming driver" because those are the people that die and drivers don't look out for.
Re: (Score:2)
> Just say people or pedestrians.
Then you're leaving out deer and opossums.
How do they compare to ride-share drivers? (Score:3)
While I don't dispute the numbers, it would be interesting to see the comparison not to all drivers but to a cohort of humans that drive the same sort of routes with the same sort of regularity, which would be drivers for Uber, Lyft and other ride-share services. I don't know if those drivers have higher rates of pedestrian and cyclists collisions (because they spend their lives driving urban streets) or lower rates (because they know the streets and intersections better). Either way a more focused comparison would be informative.
Re: (Score:2)
The paper did discuss this in the intro, and said the rates were "similar in magnitude", but Uber/Lyft may be under-reported:
> [T]he Waymo crash rate (reported as part of the NHTSA SGO [National Highway Traffic Safety Administration - Standing General Order]) was found to be similar in magnitude to self-reported human transportation network company (TNC) crashes. It’s unclear what definition of a crash is used for the self-reported TNC crash data, and whether that TNC crash definition is well matched to the ADS [Automated Driving Systems] crashes reported as part of the NHTSA SGO. That is, there is an unknown amount of underreporting in the TNC crash data, while the ADS data from the SGO includes any amount of property damage with little to no underreporting. TNC drivers may have incentives to not report low severity collisions, as reported collisions may lead to deactivation from the platform.
Need I say more (Score:3)
At avoiding crashes .... where and when they OPERATE. Not in the middle of blowing snow. Not on rutted, icy roads. Not in rain. Not in fog. Not if there is a drop on the camera. Not if the area isn't fully scanned. Not if the area has changed since being folly scanned.
Re: (Score:1)
I'm not sure of the relevance of your post given that no one is claiming they are better at avoiding crashes in circumstances other than those they have been operating in it just seems like you are pointing out the obvious and redundant.
Even if Waymo can only operate reliably in a very limited area the scope to reduce injuries is still very considerable. Picking a big US city at random I think Houston has something like 1,100 pedestrians injured and 100 killed per annum. Reducing accidents to Waymo's lev
Re: Need I say more (Score:3)
Because over 2/3 more accidents happen in the conditions I mentioned. That's where we really need them.
Re: (Score:3)
"It's not 100% perfect in all conditions, so it's useless"? Some people just want to whine. These numbers are very good. Snow and ice will always be sucky, but most other issues can be worked on. I mean, do all of the Waymo cars disable themselves when there is rain in California? I doubt it.
Besides, these cars don't need to beat ice and snow. They just need to beat human drivers in ice and snow. And that's a far, far lower bar.
Re: Need I say more (Score:2)
Until one gets in an accident that seems ridiculous for a human to do. People will tend to prefer that they at least die by a mistake that they made rather than a mistake that a machine made because some developer failed to consider that situation.
Re: (Score:3)
They don't. Houston has barely any snow storms or icy roads and not much rain. And still, it has 1100 injuries of pedestrians and 100 killed pedestrians in traffic accidents each year.
Re: (Score:2)
> Because over 2/3 more accidents happen in the conditions I mentioned. That's where we really need them.
Yup, "Our cars are much safer, except under conditions where they aren't."
I still wonder about liability.
Re: (Score:2)
> I'm not sure of the relevance of your post given that no one is claiming they are better at avoiding crashes in circumstances other than those they have been operating in it just seems like you are pointing out the obvious and redundant.
It apparently isn't so obvious that those points might just be ignored. Getting from point A to point B in real life does not happen in perfect weather all the time, so you can bet if they just put out numbers that make them look like they are sooo safe is scammish.
Re: (Score:3)
They matched for road type, vehicle type and location, but not for weather. So yes, the human-driver data may include miles done in weather conditions the driverless cars would refuse to operate in. But the study used data from California, Arizona, and Texas. How much bad weather do those states have? (I genuinely have no idea).
Re: (Score:2)
But, the fact that they're better at identifying and avoiding people and bicycles is super relevant since they've presumably been driving in bicycle and people dense areas.
It's the first results that actually say anything I've seen from one of these companies in a while (much less likely to hit a person or bicycle).
Re: (Score:2)
This is the key point that they omit because this is an investment marketing piece masked as groundbreaking news.
People drive in all conditions, make changes to a route (en route), and drive in brand new places without petabytes of prepared data stating how a vehicle should operate in an area. Waymo cannot do that. Waymor does NOT adapt quickly by any stretch of the word "quickly".
And that's OK!! Waymo's taking it slow and correct. They're moving at the speed that autonomous driving NEEDS to move.
But don't
Re: (Score:2)
> At avoiding crashes .... where and when they OPERATE. Not in the middle of blowing snow. Not on rutted, icy roads. Not in rain. Not in fog. Not if there is a drop on the camera. Not if the area isn't fully scanned. Not if the area has changed since being folly scanned.
Interesting point. My Jeep has driving assist and pedestrian emergency braking. lane change warnings. anti-collision emergency braking. Anti-tailgating cruise control.
And yes, some of it gets turned off at times in rain or dust - we're in the middle of oak pollen season here.
But I'm still the one driving. I have no issues having them on my car, and yes, they are a help. I wonder if Waymo has done a study on modern vehicles with modern bells and whistles. versus their cars?
A great step in the right direction (Score:4, Insightful)
Now if we could just remove the farm and construction safety exemption for SUVs that will never be used for construction or farming we could save even more lives. For example a car or a mini van must hit the average person, cyclist or child low enough to have them fall on the hood and not take excessive damage to the vital organs. Most SUVs have hoods above my waist and I'm 5'11".
Re: (Score:3)
> Now if we could just remove the farm and construction safety exemption for SUVs that will never be used for construction or farming we could save even more lives. For example a car or a mini van must hit the average person, cyclist or child low enough to have them fall on the hood and not take excessive damage to the vital organs. Most SUVs have hoods above my waist and I'm 5'11".
I've never heard of this "farm and construction safety exemption for SUVs" and Google doesn't return one. Where did you get this from?
Re: (Score:3)
To be completely fair most jacked up trucks are absolutely flipping useless for farm / construction work. You want to ruin a pickup? Jack it up and put all that extra stress on the ball joints, axles, et al., and in general make the truck worse.
As my friend puts, there are trucks you use for work... and most of those are built like the ones from 1970's / 1980's with very little bling. Bling trucks are "sissy trucks" that are useless to get stuff done because you're too afraid to scratch something or it's
Resuming driving right after these messages (Score:4, Insightful)
World domination according to SV vulture capitalist:
Step 1: Fund development of a product with a high barrier to entry.
Step 2: Sell it at a loss, ensuring no competition.
Step 3: Destroy all alternatives, ideally by legislation.
Step 4: Enshittify to recoup Billions spent in Step 1 -3.
Self driving is now at Step 2, moving to the Step 3.
Re: (Score:2)
> Step 3: Destroy all alternatives, ideally by legislation.
In a good election cycle, I'd expect Step 3 to be the buzzsaw that stops the whole plan, in the opposite sense: legislation or legal action to ensure the market is not dominated by a monopolist.
But right now? Excuse me, I'm going to make popcorn.
Why don't we get (Score:2)
A third party to decide, instead of a company that has a vested interest in reporting numbers that make them look good.
Re: (Score:3)
Waymo is publishing this in the [1]Traffic Injury Prevention Journal. [tandfonline.com] The journal is peer-reviewed, and (I assume) publishes reader comments. As imperfect as peer-reviewed publication is, I still think it's the best method we have right now for disseminating new scientific and engineering results.
The whole point of Waymo publishing its findings is to begin a process that lets third parties "decide" what to make of Waymo's technology. Other studies will follow, including ones from third parties, as self-driving
[1] https://www.tandfonline.com/journals/gcpi20
So... (Score:1)
The other day I boarded a Waymo and asked it to take me to the airport. A short while after we departed for the ten mile route, I felt an uneasiness in my bowels. Being that it was a Waymo, I promptly dropped my trousers and proceeded to use the back seat as my own personal latrine. It took a few minutes, but soon there was shit all over the back seat of this thing, with more than a little thiamine-rich urine. Much to my surprise, the "AI" decided that this was too much and redirected my Waymo cab into a co
Not surprising. (Score:1)
A bot car that can analyze the surrounding 350m of traffic 50x per second is (waaay) better than humans at reacting to dangerous situations.
This shouldn't be surprising to anyone paying attention in the last 10 years.
Now I'd like to know when I need a special permit to drive a human controlled Motorbike rather than being forced to use a bot-car to take me from A to B. Or when manual driving will be prohibited for humans on public roads. Not sure how fast that could happen, but I expect it to be sooner than
The question is: (Score:2)
Does it repeatedly do dumb shit and cause damage and deaths for no good reason? Including repeated obstructing of emergency services.
The way I see it is if the "AI" can't handle learning on the fly then it'll never do the job of self-driving ... for a regular consumer car at least.
Re: (Score:2)
> Does it repeatedly do dumb shit and cause damage and deaths for no good reason? Including repeated obstructing of emergency services.
Yes, Humans do that all the time. That is why they are trying to fix the problem with driverless cars. Waymo has AFAIK caused only one death. The victim was a dog which run from behind a parked car to the side of Waymo, making it nearly impossible to avoid.
> The way I see it
AI is already doing the job and better than humans, so your view is already wrong.
My Son is an Honor Student, It Proclaims Like (Score:3)
When their is an emergency, Waymo will be clueless. What we need instead: a whole bunch of ordinary people--working to put billionaires--penniless.
Yea, cool. (Score:2)
I don't believe you, Waymo.
Secondary effects (Score:2)
I wonder if they indirectly cause more crashes because they drive so slug-slow and regular drivers need to execute risky maneuvers to get around them
Re: (Score:2)
No, they do not drive too slow. They're average speed around San Francisco and the Bay Area. You just don't like technology or something. Are you OK with 40,000 humans killing each other in traffic accidents in the USA every year? (Nearly one million worldwide). Self-driving is the only way forward .. even if it's slow and they're annoying .. we need the tech to develop. The tech is already safer than any humans. We need to ban humans from driving ASAP.
So when are they going to stop using humans? (Score:2)
Because about every 5-10 miles a human being has to stop and make a correction.
That's the dirty Little secret of Waymo. I guess it's still cheaper than a actual driver if they can get the hardware costs down. Although it does concern me if Waymo is ever expected to turn a profit I wouldn't be surprised to see them demanding less human interaction even if the hardware and software isn't ready for it yet.
I never did like the idea that people are forced to opt in to a live test on public roads. Never m
I have a friend blind in one eye (Score:3)
I have a friend blind in one eye.
She doesn't have lidar or radar sensors all around her face, or multiple cameras, or 5G location software to tell her where all other vehicules are ten miles all around, or GPS.
All she has is one eye. And a head that turns. And maybe a couple accelerometers in her ears. And yet she's a good driver. An excellent driver. Heck, a better driver than I'll ever be. And she can drive equally well at night, in heavy rain, during a snowstorm, on slippery dirt roads, or when her car is completely covered in mud (she drove the Demster highway all the way to Tuktoyaktuk).
And she's never done dumb shit like encaissing herself under a blue colored tank truck because her software confused the lower line of the tank with the horizon.
Food for thought.
Wait wut? (Score:2)
What the fuck are these numbers and where are they getting them? They imply that Waymo is still injuring people and getting into accidents.
"driverless" (Score:2)
Though not really relevant for operation, the army of remote controllers pushing them through their fail safe behaviour (stopping) is essential. They are less driven, not driverless. Not autonomous enough to be left alone for any significant amount of time, which is a problem in a cellphone outage.
Re: (Score:2)
Not relevant for safety I meant.
Relevant for safety (Score:2)
If I run you over and you die, I get prosecuted for something or other, probably custodial.
If a driverless car kills you, who risks incarceration?
Re: (Score:2)
> If I run you over and you die, I get prosecuted for something or other, probably custodial.
Quite possibly, but it depends. If you were drunk/distracted/careless then the case against you is strong. On the other hand, if your brakes suddenly failed, then the question pivots to who is responsible for the brake failure. You for not servicing your car? Or the car or brake manufacturer for some kind of defect? And if I was the one who was being reckless, and gave you no chance to stop, then you might not get prosecuted at all.
> If a driverless car kills you, who risks incarceration?
Just as above, it depends. If a defect in the car or its software can be dem
If a defect in the car or its software... (Score:2)
If a driverless car kills someone, this is by definition a defect unless it is deemed legally acceptable for driverless cars to kill people in certain circumstances.
No-one is in jail as a result of dieselgate.
Re: Relevant for safety (Score:2)
People at a company are very rarely shown to be responsible, very rarely so rarely in fact that it's called a corporate bail for a reason
Re: (Score:2)
> If I run you over and you die, I get prosecuted for something or other, probably custodial.
Not unless it's deliberate, no. If it's because you did something especially dumb then it might, emphasis on might, end up in court, but you're very unlikely to end up in prison.
Here's an article on the appalling low prosecution rate of people who forget, or worse, use intentionally the fact that, their vehicle is a weapon: [1]https://www.startribune.com/in... [startribune.com]
[1] https://www.startribune.com/in-crashes-that-kill-pedestrians-the-majority-of-drivers-don-t-face-charges/380345481
Re: (Score:2)
If a washing machine kills you, who do you blame?
Re: (Score:2)
> Though not really relevant for operation, the army of remote controllers pushing them through their fail safe behaviour (stopping) is essential. They are less driven, not driverless. Not autonomous enough to be left alone for any significant amount of time, which is a problem in a cellphone outage.
Is this true? I would think that a guy with a joystick and a button in a room somewhere steering a car by looking at a monitor would be a disaster waiting to happen. Even having that guy staring at the screen for hours on end for that one incident that happens in a split second seems like a bad thing. Also, if this were slightly practical, the dollar cost of that guy would be horrible. Why save on the cost of the driver just to pay for a guy in a room?