> Plus they have a fleet of support folks and standby remote and dispatchable operators which cannot be cheap, we get no insight into the financials except that they're losing several billion dollars a year. If at the end you end up paying more than for a Uber for the service to be sustainable then what's the point.

I think they are banking on the technology continuing to improve, the number of manual interventions decreasing, and costs coming down.

Great to hear someone is still pursuing the Uber model long after Uber themselves gave up on it!

I thought the problem with Uber’s self-driving initiative was that they didn’t know what they were doing, except for the main guy who stole the IP from Google.

I think the Uber model they mean is the more general "sell your service at a loss today to increase market share, hoping it will become cheaper for you tomorrow".

If we're talking about Uber without self-driving, then the vast majority of costs are outside of Uber's hands. Even if they get their servers, payment processing, and customer support down to $0, it only helps the business model a mild amount.

On the other hand Waymo is paying enormous amounts to design systems and outfit their cars, and when that gets significantly cheaper the entire service gets significantly cheaper to provide. If they get those costs down then they're simply buying and running cars, which is much better than needing cars and drivers.

But probably not better than just needing drivers who you can trick into providing cars.

If you start driving for Uber with some vehicle that’s extremely expensive to operate, like a Land Rover or something, maybe you’d lose money, but that’s not the norm. It’s common for Uber drivers to finance their vehicles by driving for Uber. This obviously only works if there is a surplus of value created by driving that vehicle for Uber. It’s not like independently wealthy people are subsidizing Uber by donating the equity of their cars. The cost of cars used by drivers is ultimately paid by customers, there is not some magic trick here.

People heard the news Uber is profitable these days, right? The plan worked.

Turns out, the fixed cost of technology has less of an impact on your margin when amortized across a larger business. So growing by losing money can make you more profitable later.

I don't know the details, but I heard Uber Eats is what tipped them into profitability.

Luckily for all of us, a second order effect of this was that the experience of getting a taxi has been jolted forward 20+ years.

Part of being a safe driver is knowing when it's not safe to drive and resisting pressure to drive anyway.

Maybe, but I do drive on ice and snow all the time, I just slow down.

So do I, but other drivers don't.

If you live in an environment with inclement weather and awkward roads, everybody can't just choose to not drive there, but wayno can.

It's a question of their dataset not being comparable, not their individual driving decisions being better on a day to day basis

> If you live in an environment with inclement weather and awkward roads, everybody can't just choose to not drive there, but wayno can.

they accounted for that with zip code data.

but if waimo is opting out of difficult days, leaving the human to pick up that part, it will severely bias the safety comparisons because then of course humans will have more crash per mile

then any 'same period comparison' are destined to just measure this first-pick bias. they are going to need to compare with the driver data from the period before waymo introduction (while accounting for other constraints, so excluding rides on freeway, etc)

Sure.

But honestly, you have to have enough money to be able to lose your job to do that. It isn't always possible for you to miss work. All it takes to have bad attendance is a strict policy and a child that needs numerous doctor appointments. Or low pay and no paid time off when you can't get to work.

It is easy to resist pressure when you are talking a fun night out, but not so easy when the outcome could be utilities shutting off, less food, and so on.

I think a lot of people my drive because of bad weather specifically, e.g. it is raining.

It’s a comparison between real-world transportation systems, not a sporting event. The systems are different, and that includes operating under different safety rules.

It’s true that they haven’t reached sustainability yet.

>It’s a comparison between real-world transportation systems

...which have very different comparable characteristics.

Yes, but this is also true of a question like “is it safer to fly or drive.” Doing the comparison makes the most sense when there’s a flight you could take and also a way to drive there.

In this case, a practical comparison would be between taking a taxi versus a driverless car service in the same region. Maybe the driverless car avoids highways and takes longer, but you can still ask which one is safer.

I’d like to see AVs compared to the average good driver. I’d bet it’s a lot closer in that case.

The average "good" driver is likely the p90 though.

60% of US drivers survey to using a hand held cell phone while driving.

an intersection, but not exclusively 30% of US drivers survey to driving under the influence of alcohol in the last year; and that's generally excluding "buzzed"/under the legal limit - which may be lawful - but likely take one out of the "good" category.

Then you have medical conditions that go unchecked for driving in the US, peripheral vision, far-sightedness checked once a decade, reaction time, physical disabilities.

Then you have following too closely, speeding, accelerating too quickly, braking too quickly, taking turns too quickly, vehicles lacking maintenance (tires with little tread, inappropriate tread for roads, brakes not bled every ~2 years, lights out, suspension unchecked, carolina raised trucks)

Finally, you can add in US infrastructure, which traditionally maximizes speed and minimizes safety -- and generally gets a C-.

I'm not bullish on self driving cars, but that's a bit unfair. If their car performs better than 50% of the population, then that sounds like a success.

The thing is that 50% (even more) of the population should not be allowed to drive to begin with and we shouldn't wait for any technology to be ready to remove them from the road. That is the real elephant in the room that nobody wants to mention, especially political representatives who would be afraid to lose votes.

You need to set up alternatives or you're destroying lives and they'll drive anyway.

This is knee jerk reaction. Why the states and cities would have their role to increase public transit, other alternatives would set them up naturally in fact from private sector or people adopting other means of transportation.

[1] which would become immediately more profitable or less unprofitable depending on the area.

Getting those people into self-driving cars would be fantastic. Pretending there's a larger issue at play that overshadows that fantasticness is just pessimism. Sometimes it's easier to spend billions and grapple ingeniously with reality to create something new than it is to change the mind of a public official.

In the areas where it's not profitable enough, it won't get set up. And we can guarantee that states and cities will leave huge gaps.

And people would share drive or whatever it takes to make up for that.

Lots of people have no one else going near their workplace at nearly the same time. Shared drives can't get anywhere near half of people to work.

So you have to fall back on paid trips, and that is too much money for a random shift making anything in the realm of minimum wage. It just doesn't work economically.

Many many areas without sufficient transit will suffer badly.

If people are not willing to drive correctly without being distracted or under influence, that is their fault. We shouldn't have to trade security of everybody for the conveniency of the reckless ones. That whole sense of entitlement is baffling.

Besides cars are more expensive than riding a bike to a place where public transit is available.

People need cars. You can't take away half of people's cars without massively restructuring transit.

In aggregate, people are correct to feel entitled to get around town without going bankrupt trying. Why would that baffle you?

You're not talking about exceptionally bad drivers, you're talking about half of them. That makes a big difference.

If half of them are bad and dangerous, I am talking about bad drivers, regardless of the number.

People don't need cars. This is a luxury they based their life one. They can adapt the other way too.

> I am talking about bad drivers, regardless of the number.

Th number matters for whether it's feasible to implement.

In a large sense everyone is a bad driver.

> People don't need cars. This is a luxury they based their life one. They can adapt the other way too.

People are stuck in situations that need cars. Adaptation without external help would take decades. Tens of millions would suffer badly.

With enough change cars could be turned back into a luxury, but right now they aren't a luxury for most people in the US.

People are much more adaptable than you pretend.

People can't make cost-effective options appear by sheer willpower.

They can bike short distances, but longer trips would depend on access to vehicles they don't own, whether transit or other people's cars.

If you give them enough warning they can adapt. But for a change like this the only solution for many people is "don't live there, good luck moving".

Let's not look a gift horse in the mouth though?

Petty little things like political ramifications have a funny way of not being so petty,by definition 51% of people in a democracy can decide they will not be stopped from driving.

> Petty little things like political ramifications have a funny way of not being so petty,by definition 51% of people in a democracy can decide they will not be stopped from driving.

I am pretty sure among those 51% of people only a fraction is made up of total assholes that wouldn't change the way they drive if rules were more strict while another fraction could accept more and more stringent laws, especially if they come gradually and would increase their standards of driving to keep that privilege.

> by definition 51% of people in a democracy can decide they will not be stopped from driving.

This is an oversimplification. First of all, some percentage of legal drivers are not allowed to vote (non-citizens, 15-18 year olds, felons). Second, this assumes a "pure democracy," which the USA is not for several reasons, first of which is the fact that it's representative, second of which is that its representative rations haven't been updated in like a hundred years (if they were, there'd be thousands of members of congress, and California would have over a tenth of them), and finally, because of disenfranchisement through voter roll purges, gerrymandering, and political arrests (such as when ex-con felons in Florida were directly lied to and told they were allowed to vote so as to entrap them into committing the felony of voting as a felon).

Anyway we shouldn't assume people wouldn't vote against taking away the right to drive, if it came with the building out of good public transit infrastructure in return, which is what the USA desperately needs anyway. Not that I'm naive enough to think there's political will to make this happen, but the whole self driving car thing would be pretty moot if it was easy to get around San Francisco as it is to get around Paris, NYC, Tokyo, Taipei, etc.

    remove them from the road

What if that was you? Or is everyone on HN a "good driver"? I am terrible driver. I am so easily distracted by construction sites or nice nature.

Stand in a room with an actual representation of the population and you will feel very differently about what it takes to get to 51%.

> Aren't they operating under a lot of constraints? The comparison should be with human drivers operating under similar conditions and on similar roads, similar cars(eg. new ones), weather conditions etc.

The insurance data for the comparison is from the zip codes Waymo currently operates in, so all of that is accounted for. Here is the full study: https://arxiv.org/pdf/2309.01206.pdf

"All of that" is not encompassed in the human driver's home zip code. OP is talking about more than just the zip code, they're also talking about restrictions on the times of day Waymo operates, the types of vehicles they use, whether Waymo shuts down in adverse weather, etc. I'm not sure what all the current and past restrictions are, but those have to be controlled for.

The study you linked only discusses one uncontrolled variable (which is a bad sign given how many others there are that they don't talk about) but it's one I hadn't thought of:

> A limitation of the selected human baseline is that the location of crashes that generate claims is not known, which limits the ability to filter claims based on Waymo's Operational Design Domain (ODD). As a result, whereas the Waymo ODD largely does not include freeway driving, the human database includes miles driven and claims which occur on freeways. Due to variations in collision frequency per million miles between freeways and non-freeways, this may have led to a baseline which may be more conservative than a roadway- matched baseline.

Fair enough. Maybe not “all of it”, but most of it is covered.

Waymo has always operated 24/7 since the beginning and had 99.4% fleet uptime [1] during the past winter season (SF had record rains). They used to shut down during inclement weather in Phoenix years ago, but their SF rain operations have scaled up with the mileage. Waymo vehicles use its own set of sensors and don’t rely on the vehicle’s safety features, so I’m not sure how types of vehicles is relevant for the comparison.

This is pretty good as far as “apples-to-apples” comparisons go.

[1] https://waymo.com/blog/2023/08/the-waymo-drivers-rapid-learn...

I don't live in SF, but I was under the impression that Waymo's 24/7 operation is a relatively new thing [0].

EDIT: On further research, it looks like for Waymo this change only applied to when they could charge fares, so OP is correct that they were authorized to operate 24/7 [1].

[0] Article from August 10 of this year: https://www.theverge.com/2023/8/10/23827790/waymo-cruise-cpu...

[1] https://www.cpuc.ca.gov/news-and-updates/all-news/cpuc-appro...

Nope. It was Cruise who gave rides only at night. Waymo has operated 24/7 since day 1. The only restriction has been who is approved to ride (they have a long waitlist).

>so I’m not sure how types of vehicles is relevant for the comparison.

Surely "car with ABS and good tires" vs "car without ABS and with bald tires" makes a difference. And presumably waymo operates exclusively the former while some people drive the latter.

Sure, but that's driver discretion, so it's a factor that we should be considering.

That depends on exactly what you're trying to measure.

Abs has been mandatory in the us since 2012, and was very common long before that (my base model truck from 1999 has it) we can consider cars without abs statistical noise by now.

>Abs has been mandatory in the us since 2012

The average US vehicle's age is more than 12 years.

>and was very common long before that (my base model truck from 1999 has it)

My non-base-model car from 1995 didn't.

>we can consider cars without abs statistical noise by now.

I'm not so sure that's true. But even if it is, that doesn't mean that there are no meaningful differences in safety between an exclusively new (or at least very recent) and maintained vehicle fleet and the general population.

And for cars that are geofenced. Not accurate comparison.

If you're comparing to other accidents in the same (geofenced) zipcode, then why is it not an accurate comparison?

The linked study even calls this out: the zip codes they controlled for are the residence zip code, which means accidents that happen on the freeway (where Waymo typically doesn't drive) or even in other zip codes are included in the human baseline.

Precisely.

In Phoenix, the waymo drive on the freeway around the city, and given the number of crashes I see in that city I have few doubt that Waymo cars are safer.

I’ve taken more than two dozen rides and there haven’t been any manual interventions so far.

I believe for the insurance purposes, this difference is irrelevant. The point is under the current condition, Waymo driverless cars are safer than human controlled cars, thus the question of liability and permit to operate should be cleared.

The question whether humans are truly inferior to robots in term of controlling cars is rather a question of ego than a practical one.

If you had insurance for a human whose vehicle could only operate between 9pm to 6am, would it still be more expensive than Waymo's?

Probably, because Waymo runs 24/7 in San Francisco and has been for a while.

Fatalities are significantly more likely if you're driving at night than during the day, so probably yes.

Not sure that can be concluded easily as you would need to synthetically create a human driver profile with the same restrictions as Waymo. It does not say that in the linked article.

If the accident data for humans contains information about weather, location, additional complications, then it should be possible to only filter out accidents that are within Waymo's restriction. You'll similarly need to filter the human-driven miles, but a statistical approach doesn't seem to hard.

So ... possible, but TFA doesn't say, of course.

Of course insurance data doesn't contain that.

Have you worked in the real world?

You've unfortunately gotten back in the habit of breaking the site guidelines regularly - not just in this comment but in many others. Examples:

https://news.ycombinator.com/item?id=37110814

https://news.ycombinator.com/item?id=36874717

https://news.ycombinator.com/item?id=36633867

https://news.ycombinator.com/item?id=36624768

I had the impression that you'd largely fixed this for quite a long time, which was great, but this is not acceptable. We have to rate-limit and eventually ban accounts that post like this. I don't want to ban you so if you'd please review https://news.ycombinator.com/newsguidelines.html and stick properly to the rules, we'd appreciate it.

You need the driving data to contain that, too, not just the accident data to create similar driving profiles. Might not be so easy.

> I believe for the insurance purposes, this difference is irrelevant.

Can you explain this some more?

My understanding is that an insurance policy is valid in a certain set of circumstances.

They can argue that if you drive in in unintended conditions, it's your fault and they won't pay?

(But if it's true, it just make me want a self-drive car even less)

My insurance covers me driving any road in the US. I can't drive in Mexico but if it is a legal road in the US I can drive it any time. They assume i'm mostly driving near my house, but I can drive elsewhere.

"The comparison should be with human drivers operating under similar conditions and on similar roads, similar cars(eg. new ones), weather conditions etc."

It is my guess only, but a significant difference would be that robots cannot get drunk, don't drive in a sleep-deprived state and don't get distracted by texting at the wheel.

Any they don't get road rage.

The question of whether it’s financially viable is orthogonal to whether they are cheaper to insure.

That said, actuaries aren’t stupid. They know whether it’s better to have an autonomous vehicle driving in the idealized subset of conditions that Waymo operates in or whether humans are better.