Hacker News Reader

Comment by harmmonica | Hacker News Reader

As a somewhat regular user of Waymo, these types of conversations seem like they're going to be more and more in the (sorry!) rearview mirror because we won't own the car nor the cameras that are recording the world as we "drive" around.

That's not to say that we should give up fighting for some level of privacy even when we don't own the cars, but seems more likely that legislation would be passed that forces the vehicle owners/operators (Alphabet in the Waymo case) to blur peoples' faces. Then of course the state (police/gov/etc.) will clamor for a backdoor key that will unlock the blurred faces/bodies if a crime is suspected to have occurred. Speaking of, I wonder if Waymo already does blur people when they capture them through Waymo rides? I can't seem to find mention of it online.

This commentary assumes self-driving cars are here to stay and become the de facto way we drive instead of driving ourselves. Still not sure how their adoption plays out over time because, at least in the US, people will fight against mandates to use self-driving cars because it compromises their freedom (note that the freedom crowd (no judgment) will be saying that, at first, because they will consider it their right to drive themselves, but once the privacy implications are clear there will be full-on (figurative?) wars fought over self-driving). Guessing a politician, in Texas or another red state, will sooner than later enshrine the right-to-drive-oneself into the state constitution.

View on HN

> This commentary assumes self-driving cars are here to stay and become the de facto way we drive instead of driving ourselves.

This seems like an extremely myopic "tech bubble" take to me. I'm trying to find a way to put this so that it doesn't sound like an attack, but have you been to suburban or rural areas outside of places like SF, NYC or Phoenix? Being reliant on third party transportation, on roads the are often in disrepair with poor signage, is a nonstarter for probably most of the US population.

https://www.census.gov/newsroom/press-releases/2022/urban-ru...

> urban areas, defined as densely developed residential, commercial, and other nonresidential areas, now account for 80.0% of the U.S. population

They are including suburbs in that number, so that doesn’t really counter the post you are responding to.

As I just replied, counting just the largest 100 cities covers more people than the rural population.

And as established, not the rural plus suburban population.

> is a nonstarter for probably most of the US population.

I disagree.

I grew up in Rural Australia, lived in the Yukon and have driven to many of the world's most remote and undeveloped countries.

You are saying "Self driving cars will NEVER work in <this specific case>"

When what you mean to say is "Self driving cars will first work in the easy cases, and then years later will work in more and more cases until they eventually work everywhere."

FWIW, people said exactly the same thing when the automobile came around and horses were the best transport. Of course automobiles didn't work well in places with no gas stations or very nasty horse tracks for many years. But the years will always roll on, and things will change.

While self driving cars are provided by private companies, self driving cars will first work in the profitable cases, and then years later will still only work in the profitable cases.

The Model 3 and Model Y will eventually be self driving... any anyone that wants to can simply buy one.

Obviously many other manufacturers will also sell consumer cars that can completely drive themselves.

Even a person who lives in Tok, Alaska will just be able to buy one and then sleep in the car as it drives down the Alaska Highway*

* Trust me, once you've driven it a handful of times it's a chore.

Having given this some thought recently, I don't actually think anyone will sell self driving cars for a long time, if ever.

I don't think there will be a big bang moment where self driving is suddenly here and it works everywhere. Instead I think it looks like what waymo is doing now, self driving taxi services that expand to cover specific regions and then expand more as they begin to handle more edge cases and more extreme weather.

By the time self driving is actually to the standard needed to sell it with a car a younger generation will probably have grown up using self driving taxi's and wonder what they need to own a car for.

With scale and time, the tech gets cheaper and the number profitable cases continues to increase. As long as it isn’t human labor intensive (and it’s not), they can expand profitable territory almost indefinitely.

As long as the expensive bits continue to be externalized (i.e. publicly funded), I think that there might not be any cases that aren't profitable.

The expensive bits being:

- road maintenance

- theft and vandalism policing

Surely, once Waymo covers most cities, the marginal cost of getting one extra waymo car on the road should be substantially lower than a Bus (and maybe even a Marshrutka?)

...And then, once you have the cars on the road, to avoid increasing waiting times too much (which will lead users to prefer alternative transport methods, walking, cycling, etc.) you need to have slack/extra-capacity in any area that you're serving.

This means that I don't see why Waymo might not expand even in remote mountain towns: they are not going to have a massive fleet, but keeping one or two cars there shouldn't have a different overhead than for a Waymo car serving a city.

(the only caveat might be in making sure that the overnight depot/parking lot won't be too far from the area served).

I think that would be pretty cool, if it means that people won't have to drive themselves anymore... but I'm also quite concerned about the consequences this might have for public transport: if its usage numbers falls, and thus public investments in it stops, entire small town might end up depending on extremely few (one?) self-driving taxi providers.

I don't think I really disagree with you, but I think the economic models and tech that would be required for people in more remote areas (in the US at least) to solely rely on 3rd parties for transportation is so far in the future that it's basically unknowable from our current position.

That is, for many areas, I don't think it's just a case of the tech improving bit-by-bit and taking over more and more areas. I think that will happen in some ways, but take for example a friend of mine who lives in Pennsylvania. Where he lives is a far-out suburb, but he often drives out to rural areas on the weekend to go camping, or he drives long distances to various client sites. Forget the tech, I just don't see how an economic model of a self-driving car company would make these kinds of trips feasible anytime in the foreseeable future.

> most of the US population

You do realize that "most" of the US population lives in cities, right? The "rural" population has remained almost constant since the 1960s, while the "urban" population has grown to roughly 5x the rural population's size. Even just counting the largest 100 cities is more than the rural population. Setting that aside for a moment, how does poor signage affect self-driving cars? They're not like humans expecting to take the third left but miscounting, or turning right at the Dairy Queen (which has shut down). They have GPS and full maps. Those maps might not be perfect, but they'll only ever be wrong once if they're part of a system, which is what's being proposed here.

To be clear, I'm matching your tone here. Normally I'd try to be a little more understanding and tempered.

Most of the US population is urban per the census bureau but a lot of that is spread out suburbs and exurbs. I’m “urban” between many acres of orchards and conservation land. Depending on third party transportation day to day would be utterly impractical.

In a typical suburb, providing a 2-3 minute wait would require one idling self-driving car per hundreds of houses. I think that sounds practical.

That will work well when one or two people in most of those houses are commuting, dropping kids at school, and/or running errands at about the same time. /s I fully expect that autonomous vehicles at some point will marginally expand the ranks of people who are marginal car owners and will just effectively taxi. But if you have daily use that Uber/taxis doesn't solve, regularly use a personal car to drive places, have your vehicle configured for things like care seats for kids, etc. you'll do without full autonomy in favor of presumably improved automation in your personal vehicle.

> That will work well when one or two people in most of those houses are commuting, dropping kids at school, and/or running errands at about the same time. /s

I don't really understand what you're arguing here.

At most, doing that with self-driving cars would require the same number of cars that already exist. Right?

That's good enough for me to put it in the "practical" category. What's your standard for "practical"?

But more realistically, even at peak use there's a lot of cars sitting there not moving. Make some do taxi work and you can cut numbers significantly.

> I fully expect that autonomous vehicles at some point will marginally expand the ranks of people who are marginal car owners and will just effectively taxi.

Not just them, it can displace a lot of second and third cars even when it's barely touching primary cars.

You said: "In a typical suburb, providing a 2-3 minute wait would require one idling self-driving car per hundreds of houses." with the implication that this is an alternative to car ownership and would vastly reduce the number of cars.

>it can displace a lot of second and third cars even when it's barely touching primary cars.

I agree although things like Uber and Zipcar have already reduced the need for lightly-used second cars which are what I mean by marginal cars.

All I'm saying is that if the only way to get fully autonomous vehicles is through a third party service, most people will pass and just take whatever autonomous features they can get in a vehicle they buy/lease and live with that. I certainly would.

Very few people would argue that having the option of owning a vehicle capable of safe fully autonomous operation (with manual override) is a bad thing. We're just saying that having full autonomy only available as a third-party taxi service just isn't very interesting relative to today except perhaps at the margins.

> with the implication that this is an alternative to car ownership and would vastly reduce the number of cars

The point I was trying to make is that you need very few idle cars on top of the ones in active use. A fraction of what suburbs already have.

It could be a vast reduction in the number of extra cars over the peak concurrent number.

> I agree although things like Uber and Zipcar have already reduced the need for lightly-used second cars which are what I mean by marginal cars.

Somewhat, but self-driving cars are more effective in multiple ways. They don't need to be nearly as cost-effective while idling, and they can come to your doorstep instead of making you walk to them. There's nothing like a zipcar near me.

> Very few people would argue that having the option of owning a vehicle capable of safe fully autonomous operation (with manual override) is a bad thing. We're just saying that having full autonomy only available as a third-party taxi service just isn't very interesting relative to today except perhaps at the margins.

I think it goes well beyond the margins if the price works out. Or maybe your definition of "margins" is a lot bigger than mine in which case we agree and even a change at the margins is a big deal.

But also, even if the average person wouldn't switch, that's different from it being utterly impractical to switch.

This is fun because until the last couple of decades, private transportation wasn’t very common in China at all, people simply didn’t own their own cars and taxis were (and still are) ubiquitous even in small rural towns (though they might be breadbox vans to supplement shared minibuses). Almost everyone was relying in third party transportation and even today most still are.

And really, this kind of thing was common in most countries and turned out America (along with Canada, Australia) are the odd ones out with almost ubiquitous private car ownership in most of its area.

I have no idea how self driving taxis will change the USA, but I rode in my first Waymo last week on a trip to SF and it felt very real. Having lived in a country where I took a taxi to work everyday, I can totally see that life working for me (since I already lived it anyways).

Private car ownership has not been unusual in developed countries. There is a reason so many large car manufactures are not American, after all. It's not just to sell to people in NA and Australia.

What's different about the US and those other places is that far more large cities were built after private automobiles were common. There are cars everywhere in both countryside and cities in Italy, say, but there is ALSO more walkability and transit because the cities were there first (and even there, there are cities like Florence that didn't invest in rail for decades before starting to build some more light rail again very recently).

China is something of an odd one out for development patterns because of the much higher levels of state control.

What I don't really see is places voluntarily giving up point-to-point private transit after having it as an easy option for generations. Places where even the super-wealthy have turned in their private cars and drivers and take the bus with everyone else. I think the market that self-driving cars could potentially capture over mass transit is exactly those people who would rather have a private point-to-point experience. Which can be an overlapping set with the people who would also be willing to use mass transit if it was "good enough", in addition to the people who actively dislike mass transit. The total pool of users is very large.

Nothing really replaces good mass transit. But in a rural town a taxi is often the best you can manage as an alternative. Even in cities, taxis fill in gaps that make it possible to live without cars and use mass transit more. Like a trip can involve a taxi to a station then a train ride then a taxi from destination station to your actual destination.

The reason it "works" in rural China is people largely don't drive places, at least not the way Americans do.

Lots of tech folks love to shit on "car-culture" in the US, and while I agree it has resulted in really crappy urban design for a lot of US cities, you can't ignore that a lot of people love their cars because it gives them a ton of freedom. People like being able to go off at a moment's notice, independently, and drive long distances.

America may be an "outlier", but that still doesn't mean that it's reasonable to think we'll move from where we are now to getting rid of our cars because that's how China does it.

Taxis give you most of that for inner city travel. I don’t mean American taxis, I mean Chinese taxis that are everywhere. If that was a fleet of self driving cars, then it really would be like wait one minute for a taxi if you aren’t going somewhere on peak.

My point, though, isn’t how China can show America how to do things, just that American car culture is unusual and unlikely to apply to most of the rest of the world. There are also costs of dumb decisions to consider, which is going to affect the competitiveness of our cities long term (Shanghai doesn’t have to make the same stupid decisions that Denver did, and eventually it just isn’t worth doing business in cities with crappy transit infra), and eventually some more optimal norm is going to affect even the USA.

No attack taken. Fair question. I'm in a bubble at the moment because I'm located in one of the few areas where Waymo is available (think that qualifies as a (tech) bubble any way you look at it). But I feel like experiencing this tech answers a lot of questions about what is and isn't possible (or better yet what is and what will be possible with self-driving cars).

Also, I work in a very rural environment that couldn't be more hostile to self-driving cars, at least as I thought about them before riding in Waymos (an example: where I stay when I work there I have to give people turn by turn directions because if you rely on Google or Apple Maps your car will get stuck on a road with foot-deep ruts where you'll need to be pulled out; I mean it's not driving on those crazy roads in Pakistan you see videos about (never been) but I would be willing to bet those roads, which I transit regularly, are amongst the poorer-quality roads in the US and I can see self-driving tech working there before long).

And I have spent a few evenings trying to understand how the corporate-owned-fleet economics work in rural areas (and in urban). I don't think it works today. But I do think that when costs come down, and they will come down if regulation doesn't kill self-driving cars broadly, or if Elon's right and you can do it with "cameras only," then it will only be a matter of time before the tech is adapted to crappy rural roads.

Love of driving may be a generational thing as cars were new and cool for them.

Like religion having a license is on the decline:

https://www.usatoday.com/story/graphics/2024/05/17/gen-z-les...

Generational social trends are not immutable physics.

I am hoping for an urban camouflage fashion revolution as a response to more pervasive monitoring: https://www.axios.com/2019/09/07/fooling-facial-recognition-...

Public access to object recognition models may be important.

I'd like to see that too, but time and time again we've seen that:

a) laypeople aren't usually moved by privacy violations more abstract than someone physically watching you do something.

b) most people aren't willing to don practical accessories that noticably change the perception of your face unless it emphasizes qualities considered sexy.

c) safety gear generally isn't considered sexy

I think that this stuff would be perceived like wearing a physical bike helmet for your data privacy with all the cachet of Google Glass.

> most people aren't willing to don practical accessories that noticably change the perception of your face unless it emphasizes qualities considered sexy.

Ironic related story in the news:

https://www.washingtonpost.com/travel/2024/08/29/passport-ph...

Also d) it's easier to update face recognition ML to see through the latest in camouflage fashion than to design, manufacture and sell new clothes after each update to the ML model. Especially that they need to keep fooling previous versions of the model too.

Surely object recognition models will catch up to whatever attempts to thwart it (especially if it becomes popular). As long as a person is recognizable to another person, a computer should also be able to recognize them.

Trying to camouflage seems like a losing battle.

Computers are a lot better at recognizing people than people. Simple things like your gait are enough.

Just FYI, it is illegal to mask your face in New York https://www.nbcnews.com/news/amp/rcna168590

No, that's Nassau County, New York. Not the entirety of New York. Says it right there in the article you linked. There's also a lawsuit challenging it, also linked at the end of the article.

https://www.nbcnews.com/news/us-news/federal-lawsuit-challen...

So it is illegal to wear a hijab / niqab?

To be clear, a baseball hides more of your face than a hijab. (Niqab is a different story.)

Answered in the article:

> However, it has exemptions for health and religious reasons.

I did not see that. I double checked just now and there are still zero results for "religious", not sure why.

I think it's safe to assume that 0.0000001% of the population will bother with that.

Muslim women know. how is this controlling Gilde called in Dune again?

Does gait recognition (and body tics / unique movement style) not make this moot?

My sense is that facial recognition is a stop-gap and soon to be superseded, because the tech is there for more holistic 'reads' of a person - And that those subtle things that we humans can't see are actually plain as day and as clear as a fingerprint.

If we cover our face, then the data collected on gait etc. will be more than enough. If we adopt a different gait, then the data on other foibles and styles will then give us away. Etc. (we can't hope to disguise all of these at once)

A couple of years ago there were news articles that the pentagon has a "lasers that can identify people in a crowd from 200m away based on their heart rate signatures".

No idea if that's true or overblown, but it doesnt seem unlikely that such technology becomes possible in the future.

From where did they obtain your or my "heart signature"? What about gait, etc.?

you can pull a heart rate from color amplification, same basic idea as motion amplification.

surely there are simpler ways of course if you've dealt with most hoodish ruffians, as the names tells us, the people of the hood love to wear hoodies - which generally work quite well to hide from cameras, especially those above, if the hood sags forward and over the face somewhat or these: https://en.wikipedia.org/wiki/Groucho_glasses

or outside or alongside hoodies, these can be seen as socially acceptable (mostly) ; https://slate.com/human-interest/2020/07/mask-of-your-own-fa...

> these types of conversations seem like they're going to be more and more in the (sorry!) rearview mirror because we won't own the car nor the cameras that are recording the world as we "drive" around.

Perhaps in the urban setting but the majority of this country is not contained within cities. Even then are you planning on banning motorcycles and RVs?

> because they will consider it their right to drive themselves

Until a law is passed otherwise they are absolutely correct.

> the right-to-drive-oneself into the state constitution.

I doubt it. The real fight is likely to be whether we continue using mixed vehicle and pedestrian infrastructure or if we force pedestrians off the roadway entirely. Then we'll have a "right to walk" constitutional crisis.

> Perhaps in the urban setting but the majority of this country is not contained within cities.

83% of USA population live in urban areas, and that proportion is still steadily growing. The same trends apply everywhere else in the world as well.

According to the 2020 US Census, about 80% live in an urban area. However, the definition is not exactly what people think of when you say “urban”.

In 2020, the census lists 2,611 urban areas, including areas with a few thousand people.

https://www.census.gov/programs-surveys/geography/guidance/g...

Shouldn’t be too hard to break this down with a more colloquial definition (say, areas over 500k or 750k people). I’m just not at a real computer :)

Calumet park is a village right next to Chicago with under 7000 people. Northfield village has under 6000 people and it's located within about 15 miles of Chicago. It wasn't until I was about 18 that I realized what I thought was Chicago were actually small towns/villages. There's a whole slew of small villages next to or "in" larger cities.

Yes; the ordinary thing to compare is MSAs, which take this into account; Calumet Park (and Blue Island and Oak Lawn) are all part of the Chicago MSA.

> 83% of USA population live in urban areas

If you combine dense urban areas with suburbs. It's about 33% in dense areas and 55% in suburban areas. Which actually doesn't improve the driving situation.

> and that proportion is still steadily growing

Which is why I specifically mentioned motorcycles. In areas of the world with even greater urban density than the USA there are a lot of these on the road.

> The same trends apply everywhere else in the world as well.

These trends are influenced by economic policy and socioeconomic mobility of the population, which are not similar everywhere, so expectations do need to be tailored to them.

Editing to add, I actually think we'll see a new class of Drivers License, one that allows you to operate semi autonomous vehicles, and one that allows you to operate fully manual vehicles with a higher level of continuous written and on the road testing required to hold it. Which is a reasonable and non discriminatory solution to the problem.

> It's about 33% in dense areas and 55% in suburban areas. Which actually doesn't improve the driving situation.

Are the suburbs not one of the easier places for autonomous vehicles? I’d think the lower traffic, larger roads, and reliance on cars to get around (due to low density and lack of transit) would make them the ideal place for self-driving cars to succeed.

I don’t agree. The problem is suburbs have a lower cost-per-mile for trips than urban areas.

It’s way more common to take a taxi in the city than the ‘burbs. Behavior for car ownership and expectations around waiting on rides is different. Self driving taxis are an easy transition in cities. In the suburbs, you need to sell people on expensive vehicles that cost a lot per trip (whether owned or hailed).

Ultimately I think it’s the same city/rural (really dense vs less dense) divide between a lot of things.

In a suburban area, it could take 15 minutes for a taxi to get to you. In a rural area, 30 minutes to an hour. Inconvenient, especially since you could hop in the car you already have because of this situation, and probably already be where you want to go by the time they arrive.

In an urban area (especially a super dense city like Manhattan, Tokyo, Mumbai, etc.), you probably spend more time figuring out if you need a taxi than actually getting one (literally seconds in most cases), and god help you if you’re trying to park. It will not go well.

    > suburbs have a lower cost-per-mile for trips than urban areas.

Can you explain why?

Two ways to get these numbers. Consider the total miles driven, divided by cost of the car. Or consider the cost of a taxi if you don’t own a car.

I believe uber says their average cost per mile is roughly $1. So maybe $2 in urban areas. Waymo is $3 they said.

I saw some statistic that said a new car costs $800 a month now. Since we’re talking about selling new manual vs self driving cars, we can ignore people buying used cars or particularly cheap cars.

If you own a car in a city, you might drive to get groceries once a week and you may drive to a furniture store once every few years, and you take a couple trips to the airport every year. Cost city dwellers walk or take transit. Cities are dense, so the grocery store may be 2mi a way, so roughly 200mi a year, and then maybe 200mi a year for everything else. That’s 400mi a year (or 8mi/wk) with a car that statistically costs $800/mo in America - or 200/wk, so it actually costs $25/mi.

In the suburbs, you may drive 20mi round trip to the grocery store. Then 20mi a day round trip to commute, then 5mi a trip to a restaurant… it adds up to a lot more miles total. I googled it and the average American drives 1200mi/mo. That’s $1.5/mi assuming the same average $800/mo cost of a new car.

That means it’s cheaper for an urban dweller to take uber or Waymo instead of buying a new car. It’s almost but not quite cheaper for a suburbanite to take an uber but definitely not a Waymo.

> Editing to add, I actually think we'll see a new class of Drivers License, one that allows you to operate semi autonomous vehicles, and one that allows you to operate fully manual vehicles with a higher level of continuous written and on the road testing required to hold it. Which is a reasonable and non discriminatory solution to the problem.

I hard disagree. I think it’ll follow a path more like gun ownership (not trying to wade into that here though). In rural and low density areas, people believe guns provide safety, while in dense urban areas, people believe guns add risk. In low density areas, people will need to drive themselves (I doubt as many people would buy self driving cars vs use as a taxi, roads will be less well mapped, etc ), while urban areas with increased risk of driving accidents will want to restrict access of roads from humans.

This urban/rural divide doesn’t make for good licensure policy. People who depend on driving themselves are less “sophisticated” - they won’t want to spend more time getting a license, because they live farther away, they’re less likely to take drivers classes, etc. We already see states with smaller urban population have easier driver’s license standards and age requirements. Pride in vehicle ownership and car culture is already geographic simply because urban residents are less likely to own a car. So rural-leaning elected officials will want to keep human-driving easy to access.

> This urban/rural divide doesn’t make for good licensure policy.

Have you spent any time living in rural areas?

> People who depend on driving themselves are less “sophisticated” - they won’t want to spend more time getting a license, because they live farther away, they’re less likely to take drivers classes, etc

You do realize a lot of these people have class A license already because there are a lot of those jobs out there and farmers often get one to move their own product? You couldn't be _more_ wrong.

> We already see states with smaller urban population have easier driver’s license standards

They also have wildly different politics. It turns out density has more than one impact. The largest one is suicide rates. Lowest in New York highest in Alaska. You can accidentally measure population density in all kinds of ways.

> So rural-leaning elected officials will want to keep human-driving easy to access.

It's going to come down to who controls access to the freeways. I'm actually on the rural peoples side, but the interstates as a whole are a little bit out of their typical zone of influence. Given that rural life is already very different, much more so than some people can even imagine, I would expect to be the most likely point of negotiation and the most likely outcome given the parties involved.

Reasonable people could differ I guess.

> You do realize a lot of these people have class A license already

Admittedly, I didn’t know that.

> They also have wildly different politics.

Yes, I’m just assuming self driving will be a topic broken down by politics. Do you disagree?

My core thesis is that rural people won’t want self driving because it’s less compatible with their existing life, and the “saves lives argument is stronger in urban areas. I think, like guns and many other political topics, it’ll be polarizing, and the rural voters will get an outsized influence to fight it. Highways are an important part of the road system. I can’t imagine rural people being locked out nor forced to have two vehicles.

Some people hunt for sustainance, and we protect that right despite being irrelevant for 99.9% of people. Many more people use trucks or heavy equipment on their farm/homestead for work (or pleasure eg off-roading). I assume we’ll end up protecting that the same way.

I don't think you mean "you" as in me specifically, but in case that is what you meant I'm of course not banning anything and I'm not even advocating for that. I was predicting what I think will happen with self-driving cars and the privacy implications if that does to come to pass (still a big if).

As for urban vs. rural, I feel like rural will benefit just as much from self-driving as urban. I won't detail why, but it's pretty much the same reasons as urban. Economics will have to be better if it's going to be corporate-controlled cars, but, really, if Elon's right (huge if) and you can successfully have autonomy with a Model 3-level of hardware, then rural America may very well have widespread autonomous car access in the next couple of years.

Motorcyles? Great point. Not sure how that will shake out. RV's? Seems like a fantastic opportunity for autonomy. Sit in the captain's seat, beer in hand, actually watch the scenery as you drive on by. In fact I think the market for RV's will grow if folks don't have to drive them themselves because there are likely many people who wouldn't be comfortable driving an RV due to its size, but if it drives itself it could open up a whole new audience.

As for pedestrians, do you think pedestrians are being restricted more and more as the years go on? I see the opposite in both urban and rural areas in the US. Genuinely curious how your experience is different and where. I tend to think autonomous cars will make walking more pleasant. No more worrying about a car clipping you making a right turn, or a car driving unnecessarily fast and losing control, or a drunk driver losing control. All of those things may go away with autonomy (I say "may" because we're millions if not billions of miles away from anyone saying definitively that self-driving is safer/better/etc. In my limited experience riding in Waymos, though, I am incredibly optimistic about the technology. And I really look forward to us figuring out the privacy implications and other negatives that could come along with it because I think the benefits are so enormous that it'll be a massive shame if the tech does not work out long term).

I think this plays out where it becomes a luxury to drive oneself. Over the next 10-15 years it looks like self driving will continue to advance and will likely become safer than the alternative.

Once that happens insurance companies will start charging more for people that drive themselves compared to people that let the car drive.

I can see some states outlawing that practice. Then it’s left to see who is still underwriting insurance in those states.

Won't happen unless tradesmen, engineers, etc can use self-driving cars, which will e awkward when you need to park up onto a curb or something to inspect a downed power line.

That is what .5% of all drivers that need to drive on a curb to inspect a power line? I’m not sure what your point even is here?

Fourth of July, kids sports that take place outside, etc. are all regular occurrences where people need to be able to easily go up onto a curb and park in the grass.

Seems like you’re grasping at straws to find weird edge cases that don’t at all preclude a mostly self driven future. I don’t think these cars will be made without steering wheels, or a mechanism to control the vehicle directly.

But I bet they get a report of how often you manually drive and if you cross a specified threshold your rate increases.

Edit: Also why would you need to park on grass! Hop out of the car and it will park itself wherever in a nearby lot, or just drives around until you need it.

How do you call this grasping at straws exactly ?

Your suggestion that all the cars should just e driving around waiting for you is fanciful, imagine something like a live music event with 100k people? Will there just be 100k cars circling the venue waiting for pick everyone up ?

I agree with you that driving manually will become a luxury but it's important to recognise that it will manifest as a discount on self driving, not a surcharge on manual.

The only way in which I can see a surcharge on manual happening is if it becomes so incredibly rare that it becomes a niche product, or if there ends up being a bias e.g. it turns out that the pool of manual drivers is now biased towards people who like to drive in a risky manner.

If anything, in a competitive market that is able to price individual risk appropriately, the cost of manual insurance for you or I should be lower in the self driving world, because most other drivers are now "superhuman" and thus we should get into fewer accidents.

And historically has this competitive market manifested itself? Or have insurance companies instead vastly changed from their 'distributed risk' origins and instead act more as corporate entities with profits at the forefront, where the moment you actually use them the cost of being insured rises?

I don't think insurance companies need to distribute risk from auto insurance. They do, however, need it for property insurance. Floods, earthquakes, and fires and level thousands of homes quickly. There isn't any risk like that for driving.

> I think this plays out where it becomes a luxury to drive oneself. Over the next 10-15 years it looks like self driving will continue to advance and will likely become safer than the alternative.

I think this is a long way away and will vary geographically.

For a long time, self driving cars will be more expensive because they’ll have expensive sensors on it. Not many people will want a $100K self driving car instead of a $30k Camry. This means the cost-per-mile goes up unless utilization rate goes up. The most effective way to do that is make it a Taxi.

The natural result of self-driving-taxis is that the people least likely to take a taxi but most dependent on a car (rural Americans) will drive themselves still and those cars will be cheaper because they’re sensor free. That will never be a luxury product.

In urban environments though, the poorest people will continue to own cars but be slowly priced out by insurance. But maybe insurance won’t go up for manual cars, but down for self driving cars. They’ve already priced the cost of manual driving, which won’t get more dangerous as less cars are human. States might try to protect them, but I think politicians and citizens will be persuaded by “safety” over “poor people need to afford transportation”.

I wouldn’t be surprised if safety requirements will gradually tighten until every car will need most sensors. Manufacturers have an interest in selling expensive difficult to manufacture cars and politicians like to reduce traffic fatalities.

The average age of a car in the US is 12.6 years old right now. History shows that we’ve not forced safety items into existing cars. (I have one car where I would be permitted to pass safety inspection without seat belts (any) because it wasn’t originally equipped. [I have chosen to add them.])

12.6 and rising.

After the 15 year mark, give or take, it becomes a luxury to maintain the car. There will hopefully be a point in the nearer future where ICE will become a luxury.

If I understand your point about luxury correctly, I agree that there’s a crossover point, but I think it’s more like 35 years (1990 model year or earlier) rather than 15 (2010 MY or earlier).

I don’t think many people driving a 20 model years-old 2005 (including myself) are treating maintaining that car like it’s a classic car hobby.

> Not many people will want a $100K self driving car instead of a $30k Camry

What you said is true, but does not reflect reality.

Today The Model 3 is ~$40k, and once self driving is "solved", there is no reason a ~$40k car won't be capable of it.

Sure, but that’s assuming the technology on a Tesla Model 3 actually is capable of safely self driving. Today, Waymo is the clear winner, and the equipment costs a lot more than a Tesla.

I know Elon/Tesla can be a charged topic, but teslas don’t self drive today. It might be true, but I don’t think we can assume they’re capable with just a software update based on the info we have today.

> Waymo is the clear winner, and the equipment costs a lot more than a Tesla.

I don’t think that is true. What costs a lot more? Lidar?

Call it $10k more on the very high end. So you have fully self driving cars for $50k. A very long way from your $100k

The LIDAR sensors are significantly more expensive than 10k. There are a lot of them on it too. Additionally, I expect that the GPUs or other compute requirements are an extra 10+k at least.

While I'm sure the prices can come down with mass production, the estimated BOM for a Waymo are speculated to be >200k per vehicle. Maybe we can that down to 100K, but I'd be very suspicious of a 50k vehicle anytime soon.

Tesla Model 3 is right below that $30,000 mark, Tesla FSD does the bulk of my driving today. I think your core premise is flawed, it will be cheaper and sooner than you anticipate which will alter the conclusions you’ve come to.

Teslas 'FSD' is not the commonly understood HOOTL self driving.

And? Just a few years ago waymo had safety drivers. I really only intercede when I want to take a route different than the Tesla routes me on, or I want to be a bit more aggressive in highway traffic.

Seems like 10 years isn’t too out there for affordable self driving vehicles considering we have one working today at the high end of cost and another getting close at the lower end of cost.

> freedom crowd might fight it

Look at the 2020 covid vaccines. The freedom crowd said this was going to have massive privacy implications, there was a propaganda machine pushing those people as being crazy antivaxers (maybe many were, let's talk about the large subset who just had privacy concerns), and the net result is that most US citizens have their names, addresses, preferred vaccination locations, preferred vaccination times, propensity for following local regulations, ..., recorded in a database so broken it's basically public.

The freedom crowd didn't have a lot of power against a propaganda machine turning their neighbors against them. Tack in a few dozens of billions from Tesla or Google claiming their cars are safer than the average driver (in well-studied, dry, daylight, slow streets) and using that to push anyone unwilling to roll that tech out globally as a road-raged Luddite, and I'm not sure the freedom crowd are going to be able to do much to slow down our corporate overlords.

I'm not going to respond to your Covid commentary in an effort to avoid going down a rabbit hole, but I should say that when I wrote "freedom crowd" it's not some monolith. Specifically, I think there are plenty of freedom-minded folks across the political spectrum, and I do believe that those folks, again, across the spectrum, are going to have a hard time accepting self-driving cars en masse because of the privacy implications. I do think it will weigh more to the conservative side of the fence, but if you haven't hung around with extremely progressive people there is a huge contingent on that side that is very wary of government mandates and, I feel dumb even writing something this obvious, far more wary of corporate America's agenda than any other population in the country. Maybe you were talking about the left in the first place, but I think you were referring to the right based on bringing up Covid and the other language you used.

That isn't a political statement. The flow from "we won't shove you in a database" to "the non-existent database was leaked" is something that's independently verifiable.

> left/right

Yes, many people care about freedom

> going to have a hard time accepting self-driving cars

I'm usually optimistic, but I don't think public sentiment matters here. To the extent it does, it'll be some sort of Faustian bargain, where the law says your car must have a backup camera (arguably useful) and as a byproduct brings in unwanted, undisclosed tracking and also remote takeover capabilities through radio bugs (buffer overflows plus poorly designed canbus access). However self-driving inflicts itself on the masses, it'll be in a series of small enough steps that the freedom crowd can't appropriately fight back.

I'll certainly accept the possibility of bugs allowing remote takeover of vehicle cameras, but in general I don't think rear cameras are much of a risk to privacy. They're not always-on (generally they only turn on when the car is in reverse gear), and on many car models they're even physically covered when they're not active. (E.g. on my car, the manufacturer logo on the rear of the car flips up to allow the rear camera to "see" when I'm in reverse gear.)

Regarding bugs and remote takeover, the one thing that does feel like a mitigating factor is I expect for many car models it might be tough to find an exploit that enables the rear camera, but doesn't let the driver know that it's active. Usually when the rear camera is active, the infotainment display automatically switches to the rear view, and I wouldn't be surprised that there's no mode for "turn on rear camera but don't display the video feed to the driver" on most cars. Not as a security/privacy feature, but just because it's easier to write the code that links the two all the time rather than having it be conditional on something.

But yeah, as autonomous driving becomes more pervasive, there will be more cars with cameras recording everything around them at all times. I expect the manufacturers/operators of these cars to store that video for some time after capturing it, even just for quality control, bug tracking, and dispute resolution if there's a crash. And certainly law enforcement already has legal tools to compel the release of this kind of video.

I suppose this is just the next step of what's already the status quo, though. Many public places already have (stationary) cameras, either operated by the police or by private individuals, and police generally have the ability to access the latter, even. Cameras on autonomous vehicles just mean many more cameras over time, cameras that move around.

I don't really love this situation; I would like to see legislation that makes all of this data protected, and requires companies that store it to delete it after a relatively short amount of time (unless subject to a legal hold). But like most things, legal protections tend to lag behind technological progress.

My complaint (from above, not the only problem with cameras) about the rear camera is that the cheapest way for them to be installed nowadays, given that other people want infotainment systems, is for them to go hand-in-hand with an infotainment system. Whether the camera is on or not, the car is now vulnerable.

The sin being committed is having too much code, especially too much untrusted, untested code (it's "just a radio") wired into the canbus. Toss in a little auto-update functionality and some always-on antennas, and you have all the makings of a nasty exploit. Last I checked, none of those have been known to be exploited in the wild, but they've also not been patched in the last decade. It's a fundamental design flaw that saves a few dollars, so it persists.

Can’t I just buy a car that lets me drive it but also can drive itself?

I can already today add functionality to cars with after-market hardware: https://comma.ai/?

What is the legality of comma.ai kit? I find it hard to believe that states would want to allow it. It is surely much worse than Waymo.

Reply on news.ycombinator.com