I really wish some of these rah-rah self-driving cars articles would acknowledge that while self-driving cars are great WHEN THEY WORK, there are still huge obstacles to making them work outside of areas like San Francisco that have easy driving weather and have been mapped and ring-fenced to within an inch of their lives. We're still a long way from when you can get in a self-driving car in Grand Rapids, Michigan in January and ask it to take you up to the Ferris State University campus in Big Rapids. These things are currently best thought of as a new form of public transit in certain large cities, not as a genuine replacement for the uses for which the vast majority of American drivers need a car.
I think this probably understates how quickly things can improve. It seems entirely likely that by 2035-2040, a majority of vehicle miles driven will be fully autonomous. A decade isn't a long time.
I don't understand why you think an AV wouldn't be able to drive you down the freeway, even if it is a longer trip? It can't possibly be more complicated than downtown San Francisco, with bike lanes, buses, pedestrians everywhere, huge amounts of traffic, cars blocking intersections, emergency vehicles, etc. And (for what it's worth) mapping is a relatively minor consideration. We already have all of the information these AVs typically need, including route information and speed limits.
They can't see the road, or other traffic, in snow. They struggle to deal with streets that have worn lane markers, or are undergoing construction. The sensor technology and on-board processing needed to deal with suboptimal driving conditions can't yet be deployed in anything like an economical matter.
I do think you're likely underestimating the current capabilities of these AVs. LA and San Francisco have plenty of extremely worn lane markers and plenty of constructions sites (in fact, I'd argue those cities have a significantly greater density of construction sites - including sites that are blocking lanes - than suburban Michigan). And yet, they're vastly safer than drivers in those cities.
It's true they haven't yet been tested in snow, but I would gently disagree that it would pose as much of a problem as you believe.
I've lived in both California and Michigan. Road quality in California is much, MUCH better, simply because of the environment. Michigan roads are subjected to frequent freeze-thaw cycles, California roads generally are not (and certainly not in LA or San Francisco). and as a result Michigan roads break down much more frequently and severely than they do n in California. They require much more frequent maintenance and repair than California roads. The amount of road closures and diversions in California is nothing like what it is in Michigan or states with similar weather. AVs really struggle with it. And as for snow: it's not that AVs haven't been tested in it. It's that no sensor/compute solution that could be deployed economically has solved it yet. Humans can drive on a road that's been completely obscured by snow because they can tell where the road and lanes should be; AVs can't, at least not consistently or safely.
I was actually working in the auto industry back in 2017, when everyone was pouring huge money into AV development and the industry was very confident that Level 5 AV technology would be a solved problem by 2022. It's not, which is why the industry has been retreating back to Level 2 and 3 systems in their commercial vehicles. I've been watching companies like Waymo and May Mobility drive their test cars around the Detroit area for almost a decade now; they're no closer to going live with an actual commercial service than they've ever been. The technology is at a place where AVs can serve the same function as a bus route or subway; but in most places, the economics don't support that.
The point is that Boston is absolutely a place where it will make sense to launch soon (if they can handle Austin and Atlanta weather, they’ll have proof they can work in some snow days on top of gigantic thunderstorms), so Boston shouldn’t ban them!
I’m surprised old people are so opposed. Many of the people I know in their early 70s are scared to drive but feel like they have to in order to live their lives. I told some of them that in a few years they will be able to push a button and a self-driving car will affordably take them wherever they want and they were all very in favor.
Absolutely no argument whatsoever on the safety aspects here. Just want to gently push back on the notion that ending the ability of people driving their own cars doesn't require a ban. It will emerge naturally via insurance markets. As the share of self-driving vehicles increases the risk differential between human and self-driving vehicles will become more and more stark and the cost to insure a non-self driving vehicle as a daily will, at some point, become prohibitive.
The risk of driving is what it is and premiums will be priced accordingly - if self driving cars are safer then premiums will go down for self driving cars, not up for everybody else because insurance is priced on risk x damages, not differential. It's possible that legislatures could change this (say demanding $1m limits if you deign to drive) but why would they?
And this is kind of rhetoric, already fairly widespread in tech circles, is exactly why those of us who deeply value human driving are unwilling to trust that AV rollout isn't the thin end of a wedge for a ban, despite the protestations of Kelsey and others.
I've been hearing for my entire life that the government is coming to take [x] because [x] is unsafe. Guns, alcohol, gas stoves - all of which are still available, and probably more available than most times in the past. (Drugs are an edge case, but here the tendency has been toward liberalization too. Same with gambling.) And then I've been told that the government is going to mandate more conduct in the name of safety. Eating broccoli, famously. Which also has not happened.
I find it extremely unlikely that in a democratic society, we are going to end up with a ban on cars in the name of safety, when a lot of people really like driving cars and will probably not vote for the people who are like "hey let's ban driving cars."
FWIW, my above comment was meant to reply to Casey, not you, but I'll leave it up.
Anyway, those things remain legal in large part due to very aggressive political efforts against those who have sought to ban them (and at times, partially, did). Many west coast cities, for instance, *are* banning new installs of gas stoves. And more broadly, Rich Trumka, head of the US Consumer Safety Commission at the time, *did* explicitly say a blanket ban on gas stoves was on the table. This produced an enormous backlash which ensured it would not happen. The backlash (and preemptive action) matter. The gun rights movement is another example - gun laws are where they are today because many people fought very hard for it, not because it just magically fell into place.
Furthermore, if truly nobody is coming to take our cars, then a law enshrining this should be zero-cost and obvious common sense. The reality though, as we see in these very comments as well as the proclamations of various technologists, is that many *do* seek to ban human driving, which is exactly why such a law is needed.
Fair, I think. The thing we need, maybe, is more democracy and less government-by-the-regulators, because the gas-stoves-ban-discussion is one of those things that was floated by a regulator and, to your point, was quashed through the opposition of the people. These sort of paternalistic banning beef because of methane suggestions (which we're currently seeing on the right in the form of vaccine availability; I'm not meaning to suggest it only goes one way) seem much more often to be done through the administrative state than the elected branches.
Ultimately, I would have as much issue with a law that says "you get to drive a car" as I do with a law that says "you get to breathe air." I don't think either is singularly necessary, but since that's the case, there's no harm in it. If people change their mind, then the law can change with it.
But all this is well downstream of the underlying issue: driverless cars appear to be safer and better than human drivers, and their acceptance is a good thing. I noted that Kelsey Piper said in a separate column today that if X deaths are a consequence of the first amendment, then she'll take that trade, because the first amendment is that important. I personally don't get that worked up about driving, but I think it's a reasonable position to think that driving is so important that we have to be willing to make sacrifices to accept it. We do with drugs and alcohol and guns, after all. Kelsey's answer to the first question ("We should allow driverless cars, which are good") seems (a) true, and (b) a real stretch from the second question ("Should we stop humans from driving?")
I do think that on some time scale a ban is reasonable. The people who thought cars shouldn’t be allowed because some day we would ban horses in city streets were factually right, but morally wrong.
“they are scared that as soon as Waymos are safer than human drivers, we will take their cars away.”
This is a common fear that leads people to resist technological and social change, and arguably liberalism more broadly. Some other versions of this fear I’ve seen:
- Zoning reform means banning single family homes.
- Incentivizing electric cars means taking away your ICE car.
- Accepting gay and trans people means everyone will be pressured to identify as queer or nonbinary in some way, and to reject heterosexuality and traditional families.
It’s basically “allowing a new thing will mean the old thing will be banned or heavily socially sanctioned”. I won’t say this fear is never rational, but most cases I’ve seen have been.
It is probably worth acknowledging that--once AVs are widespread enough to command large market-share--the rational thing to do *is* to ban human driving, at least in high-traffic areas (if anyone even wants to drive?) Whether or not that becomes a culture war talking point remains to be seen...
Just like the point of banning fake meat (in Florida) and banning the word “oat milk” (in Europe) is that people are worried that some day we will ban real meat and real milk, and they don’t like the idea that they themselves will be happy with that when the time comes. They want to lock in their values now.
For years, I've wanted a self-driving car. On longish trips, I'd love to do something else instead of staring out the windshield at six lanes of traffic bounded by endless billboards. I live in a small town in Florida and -having- a car is useful but -driving- my car is often a time-waster.
Open up your vision, having self-driving cars isn't just a bus-replacement.
It’s also a bus enabler! The city has enough buses to run the rush hour schedule 24/7 - it’s just the driver costs that make it not economical to do so.
The US is a -very- large country where long distance travel is a norm. Even in extended cities like LA, I've known people who made 150 mile daily commutes. Mostly freeway travel, self driving, privately owned cars would be a god send.
Americans make long trips by car as a norm. I live in Florida and airline service goes from hub to hub, Atlanta to Orlando, Atlanta to Miami, there's very little schedule air service between e.g., Pensacola and Miami, a distance of 670 miles. Most states outside the Northeast have long, skinny routes. Tennessee for example is the longest state east to west and lets not think about the Great State, Texas.
All of these are prime markets for self driving privately owned vehicles.
It doesn’t really matter how big a state is. No one chooses their destination based on whether they are crossing state lines. They choose based on whether there is a closer destination where they can do the thing they want to do.
The vast majority of driving is in town, and the vast majority of town driving is on trips that are under 15 miles each way. In any case, self driving helps buses far more than it helps cars (because it’s a predictable route and it is cost-efficient to run on a constant schedule all day), even if buses and cars serve different types of travel markets.
At bare minimum, then, an explicit law protecting the right of drivers to drive their own vehicles (including motorcycles, too) on all public roads, *in perpetuity*, with robust enforcement mechanisms, ought to be a precondition to the mass rollout of self-driving cars. Too many times we have seen the safetyists turn the novel and optional into the mandatory. Hundreds of millions of us simply will not stop driving the old fashioned way, nor give up our vehicles, and we could easily throw a big fat wrench into the works should we choose. Work with us or fight us: those are the options. For the rollout to go smoothly, people like me need to trust that this isn't the thin end of the wedge to banning human driving (which a fair number of technologists are already excitedly blabbing about), so it ought to be very much in the interest of AV proponents too.
The law will of course need to also cover punitive differential tolling and registration fees designed to keep human drivers off the road via the backdoor.
Earnest question--why do you feel so emotionally strongly about this? Why is it important to be able to drive, especially in big, congested metros? It just seems kind of low stakes to me.
Frankly I'm not interested in litigating this question. It doesn't and shouldn't matter.
I'll just say that:
(1) if you've never taken a spontaneous self-directed road trip for days or weeks on end (or experienced the thrill of the open roads from the seat of a motorcycle), we're not arguing from remotely the same plane and cannot and will not see eye to eye. This is not purely a technocratic calculation.
(2) even beyond that and into the mundane, autonomy, self mastery, and humanism are all crucial values, and I'm sick and tired of ceding more and more of my basic actions to machines I can't control
(3) policy, especially around technology, is too often shaped around the perceived needs of major metros while ignoring everyone else. California for instance is soon to ban fuel powered generators, a disaster for the mountain regions where long power outages are the norm and in cold snowy winter conditions
(4) the surveillance and top-down control implications of all this are terrifying.
I mean, you seem very interested in litigating this question and seem to think it does matter a lot.
I will say that:
1. I don't like spontaneity or road trips. I've done it; don't like it; way too much time on the road. Being in a car passing through the outdoors is not the same as actually being outdoors. Definitely a fly + preplan the itinerary + optimize for minimal travel time guy. I particularly enjoy travel to outdoorsy destinations that don't require a car (e.g. the Swiss Alps, Japanese Alps, western Cornwall, the Cote d'Azur). I'd love to have this same option in Yosemite, Big Sur, Southern Utah, etc.
2. I'm not sure the evolution of cars from human-driven to machine-driven is meaningfully more extreme than any other technical advancement of the past 200 years. And I definitely don't think >5% of the population wants less technological advancement given, you know, revealed preferences. Is this any different than ceding "autonomy" over washing your clothes?
3. I'm totally fine with making policy for metros and rural areas differently. Ideally, fuel-powered generators would be allowed in some counties/census tracts and prohibited in others.
4. More terrifying than the status-quo where people just get hit by cars all the time?
I mean, this obviously seems like a difference in values. I absolutely am biased towards the technocratic approach. I suppose the relevant question is...how many of you are there, given your "Work with us or fight us: those are the options" approach? Are you one person or a constituency?
Also, do you want a constitutional amendment on this? I'm not sure any law can be guaranteed in perpetuity given the nature of elections + legislation.
I'm at work now and will respond properly later. A few quick points:
1. By "I'm not interested in litigating this question", I specifically meant the *why* of "why don't I want to forcibly delegate all human driving to machines". It's a statement of philosophical principle that it ought not matter why, basic norms of liberty-as-baseline alone ought be sufficient.
2. Right now I just have a WordPress site at a cheap domain with one half-written post. I spun it up and started thinking about it only last week after seeing yet another tech founder crowing about how no humans will be driving in a decade. I plan to start spending 5-10 hours a week working on this as a side project, probing the lobbying/legal space, building a mailing list, and investigating best courses of action.
3. Based on surveys and various other sources of information, I do believe more Americans are in my camp than your camp when it comes to driving. I'd roughly put it at 45:25 (30% fuzzy middle) on the question of personal preference, and 70:15 on the issue of bans, if I had to guesstimate.
4. Preliminarily, I think a well written federal law could get 80-90% of this, but an amendment may be another approach.
5. As to your point #4, yes, I will absolutely take that tradeoff. Freedom over safety, in all things.
You are awfully eager to consign a bunch of other people's children to readily preventable deaths just so you can enjoy the romance of the open road. The parents of victims of traffic violence might have a different perspective, and you would come across as a more reasonable and decent person if you at least acknowledged that perspective.
I already did so in acknowledging that there is a tradeoff that exists - more broadly I'm a big advocate of people in policy discussions taking this road rather than pretending their approach has no tradeoffs and is all good and no bad.
Beyond that my view on this is exactly homologous to Charlie Kirk's now-(in)famous view on guns. I have no interest in playing these silly games of weaponized empathy where I have to do a bunch of performative crying before I can straightforwardly make my point. It's the COVID discourse all over again.
Car accidents kill 2,000 children a year. Drowning kills 1,000, basically the same order of magnitude so a useful comp. If I insisted that your right to a swimming pool callously disregards hydraulic violence against children and that parents of said children would find my freedom-first approach loathsome, you'd rightly dismiss me as insane. Bad things happen - it's the law of large numbers.
(To clarify, I'm not implying that's your view: I'm making an analogy to try and find your limiting point on the "safety first" approach.)
You are probably correct about 45:25 (30% fuzzy middle) on the question of personal preference and 70:15 on the issue of bans, but that strikes me as a softly-held belief. Most people haven't actually seen/used Waymo yet; to them, it's abstract.
We will see how those numbers evolves once 1) Waymo goes national and 2) once Waymo has been around for a decade and is a normal fact of life, rather than a novelty
If I am to guess at how things will actually evolve, I'd say that come 2035, most 16-18 year olds won't bother to learn how to drive and there will just be generational turnover on this as the 21st century progresses. Same with the gradual phase-out of manual cars, even if nobody is actually going to ban them.
That’s really awful. I wouldn’t want snake oil peddlers to have threatened to ban more effective medications without a law locking in the legality of their more dangerous treatments in perpetuity just because people like them.
Human driving is inherently faulty. 40,000 auto accident deaths per year represents death on an unforgivable scale. It's persistent death — 40,000 *each year* — and yet, this number actually seems to have *decreased* since the time when I went through driver's training, a while back, when I was advised that the number of annual highway deaths was more like 50,000…
So, human drivers suck. No argument. We all see examples of bad, and even criminally bad driving, just about every day. And then, there are the drunk (and/or high) drivers, whose bad judgment and selfishness make this unforgivable circumstance even more unforgivable.
The one best and most excellent feature of our system of human driving however, is its intrinsically compartmental architecture. A bad driver is only one bad driver, and not a systematic flaw of bad driving which affects more than a single car. Although there are tens of thousands of these individual bad driving flaws each day in the U.S., at least they are all *isolated* flaws.
The problem with computer-operated cars is the same problem with computer-operated anything. Because software is so incredibly complex, and because computers themselves operate software with only with the most unforgiving inflexibility, there is an astronomical number of potential failure points in the process of performing any task — let alone one as complex as operating a motor vehicle. Any human mistake in design or coding at any one of these failure points could create a catastrophic disaster affecting all users of the system, and not just a single jerky driver who thinks he or she owns the road while driving 85 mph along the breakdown lane of a highway, past queues of crawling traffic during rush hour…
Plus, there are security risks. Any computer can be hacked, therefore any car which is operated by computer is vulnerable to a potentially fatal software intrusion, perpetrated by a bad actor — a psychotic actor, or a sociopathic actor with a bad motive. Imagine the consequences of hacking an entire self-driving network, extending along all roads within a whole city, or every city. In a whole country…
It's an enormous risk that can't be ignored. Any decision to abandon epidemic, although isolated, cases of bad human driving would need to consider the risk acquired by replacing it with an inhuman system with so many potentially fatal points of failure which, if encountered, could create death and chaos on a mass scale.
I can see how some sort of undetected flaw in a self driving system update could somehow cause chaos - all the cars might stop one day. But I don’t see how they could cause death and destruction - all of these systems have a much deeper control that makes them stop when things look dangerous, so if something made them start doing things that causes danger, they would all just stop.
The design feature of self-driving software, to stop in case of danger or failure of some kind, is reasonable — but what if this fail-safe is triggered during a driving situation where it's actually unsafe to stop?
Also, who knows what sort of danger human passengers in computer-driven cars might face, if their cars all become inoperable at the same time, in a catastrophic scenario — or even in a single car, during an isolated "client side" failure?
Think of how often software versions are updated to correct "bugs". No matter how carefully programmed any code may be, it's oppressively difficult or maybe even impossible to get 100% right, even after many patches, so it's reasonable to assume that there will always be software flaws — and every flaw is a potential failure point.
…
Another possibility to keep in mind is a scenario where a hostile force might hack the software, reprogramming it to begin driving each car it operates very dangerously. We've seen military forces hack into enemy computer systems before, or otherwise booby-trap enemy hardware, to create disruptions and cause casualties. The massive failure point of using the same software to determine the movement of millions of cars enabled this possibility.
…
Think of nuclear power as an analogy. It's tempting to generate gigawatts of carbon neutral power, and while the designers of nuclear plants definitely go through a lot of trouble to perfect their architecture and safety systems — because these systems are human-engineered and constructed, and because they're subject to the unpredictable forces of nature while they're in use, there's simply no way to make them truly fail-safe. The consequences of such failures can be so great that even a slim possibility of one must be, and is, taken into account while deciding whether to exploit such a tempting route to massive amounts of carbon-free energy, or not.
…
I can't say whether these risks are more or less dangerous than the risks we face now from flawed human driving, but I am saying that they should be taken into account while deciding whether computer-driven cars are a good idea.
There’s many reasons why it’s a good thing that we started with a few dozen cars in Phoenix and moved to a few hundred in Phoenix and SF and are now at a few thousand in five cities, rather than overnight putting millions of them on the roads in all cities at once. Understanding the vulnerabilities to system updates and failures is part of that.
Failures may happen at any time. Security may be breeched at any time. It's the same thing with any large-scale, inherently risky ambition — no matter how carefully planned a project is, there will always be risks which should be taken into account while judging whether they're worth assuming, just to replace other risks which are already in place…
Or whether there might be other, less risky options with more upside to consider.
The old people bashing - while very stereotypical - is simply wrong on its face. 18-29 is pretty significantly more accident-prone as a demographic both in terms of per 100k miles and per licensed driver. Deaths go up after 75, but that appears to be in part the result of 80 year-olds being quite frail so crashes involving them are more likely to kill them.
As a cyclist commuter in a major city, I’m a hardcore autonomous vehicle fan. But I think at the end of the day this boils down to whether we want to be a country of automation accelerationism, or we want to keep a status quo in place where we have 3 million people employed as truck drivers and another few million working for rideshare companies. To me it seems clear that pretending away the issue of driver automation won’t work out in the medium to long term. But I also understand that these careers are highly engrained in the American ethos going back a century.
I appreciate that this article takes into account that Americans actually do really enjoy driving their own cars. In the future, I would love to switch on my AI-assisted self-driving technology on long road trips when I want to read, work, or eat a road trip snack. What leads to the most traffic accidents is distracted driving, I would venture to say, people trying to scarf down their sandwich while trying to help kids eat in the back seat has led to way more accidents than a self-driving taxi ever will. Ignoring technology that can make the roadways safer is foolish.
What I am scared of isn't the lack of jobs available for Uber drivers, but the safety standards used to determine that AI can not only drive in San Francisco but also on all of the highways and byways in the rest of America, in thick traffic, slippery roads, and in road construction.
I really wish some of these rah-rah self-driving cars articles would acknowledge that while self-driving cars are great WHEN THEY WORK, there are still huge obstacles to making them work outside of areas like San Francisco that have easy driving weather and have been mapped and ring-fenced to within an inch of their lives. We're still a long way from when you can get in a self-driving car in Grand Rapids, Michigan in January and ask it to take you up to the Ferris State University campus in Big Rapids. These things are currently best thought of as a new form of public transit in certain large cities, not as a genuine replacement for the uses for which the vast majority of American drivers need a car.
I think this probably understates how quickly things can improve. It seems entirely likely that by 2035-2040, a majority of vehicle miles driven will be fully autonomous. A decade isn't a long time.
I don't understand why you think an AV wouldn't be able to drive you down the freeway, even if it is a longer trip? It can't possibly be more complicated than downtown San Francisco, with bike lanes, buses, pedestrians everywhere, huge amounts of traffic, cars blocking intersections, emergency vehicles, etc. And (for what it's worth) mapping is a relatively minor consideration. We already have all of the information these AVs typically need, including route information and speed limits.
They can't see the road, or other traffic, in snow. They struggle to deal with streets that have worn lane markers, or are undergoing construction. The sensor technology and on-board processing needed to deal with suboptimal driving conditions can't yet be deployed in anything like an economical matter.
I do think you're likely underestimating the current capabilities of these AVs. LA and San Francisco have plenty of extremely worn lane markers and plenty of constructions sites (in fact, I'd argue those cities have a significantly greater density of construction sites - including sites that are blocking lanes - than suburban Michigan). And yet, they're vastly safer than drivers in those cities.
It's true they haven't yet been tested in snow, but I would gently disagree that it would pose as much of a problem as you believe.
I've lived in both California and Michigan. Road quality in California is much, MUCH better, simply because of the environment. Michigan roads are subjected to frequent freeze-thaw cycles, California roads generally are not (and certainly not in LA or San Francisco). and as a result Michigan roads break down much more frequently and severely than they do n in California. They require much more frequent maintenance and repair than California roads. The amount of road closures and diversions in California is nothing like what it is in Michigan or states with similar weather. AVs really struggle with it. And as for snow: it's not that AVs haven't been tested in it. It's that no sensor/compute solution that could be deployed economically has solved it yet. Humans can drive on a road that's been completely obscured by snow because they can tell where the road and lanes should be; AVs can't, at least not consistently or safely.
I was actually working in the auto industry back in 2017, when everyone was pouring huge money into AV development and the industry was very confident that Level 5 AV technology would be a solved problem by 2022. It's not, which is why the industry has been retreating back to Level 2 and 3 systems in their commercial vehicles. I've been watching companies like Waymo and May Mobility drive their test cars around the Detroit area for almost a decade now; they're no closer to going live with an actual commercial service than they've ever been. The technology is at a place where AVs can serve the same function as a bus route or subway; but in most places, the economics don't support that.
The point is that Boston is absolutely a place where it will make sense to launch soon (if they can handle Austin and Atlanta weather, they’ll have proof they can work in some snow days on top of gigantic thunderstorms), so Boston shouldn’t ban them!
Yep, not to mention that they do still require remote human operators to intervene at times, though (probably?) not too often.
This doesn’t contradict the article’s thesis though. In a way it should reassure people worried that human drivers will be banned.
I’m surprised old people are so opposed. Many of the people I know in their early 70s are scared to drive but feel like they have to in order to live their lives. I told some of them that in a few years they will be able to push a button and a self-driving car will affordably take them wherever they want and they were all very in favor.
It’s not so surprising that older people are more against a new technology than younger people.
Absolutely no argument whatsoever on the safety aspects here. Just want to gently push back on the notion that ending the ability of people driving their own cars doesn't require a ban. It will emerge naturally via insurance markets. As the share of self-driving vehicles increases the risk differential between human and self-driving vehicles will become more and more stark and the cost to insure a non-self driving vehicle as a daily will, at some point, become prohibitive.
The risk of driving is what it is and premiums will be priced accordingly - if self driving cars are safer then premiums will go down for self driving cars, not up for everybody else because insurance is priced on risk x damages, not differential. It's possible that legislatures could change this (say demanding $1m limits if you deign to drive) but why would they?
And this is kind of rhetoric, already fairly widespread in tech circles, is exactly why those of us who deeply value human driving are unwilling to trust that AV rollout isn't the thin end of a wedge for a ban, despite the protestations of Kelsey and others.
I've been hearing for my entire life that the government is coming to take [x] because [x] is unsafe. Guns, alcohol, gas stoves - all of which are still available, and probably more available than most times in the past. (Drugs are an edge case, but here the tendency has been toward liberalization too. Same with gambling.) And then I've been told that the government is going to mandate more conduct in the name of safety. Eating broccoli, famously. Which also has not happened.
I find it extremely unlikely that in a democratic society, we are going to end up with a ban on cars in the name of safety, when a lot of people really like driving cars and will probably not vote for the people who are like "hey let's ban driving cars."
FWIW, my above comment was meant to reply to Casey, not you, but I'll leave it up.
Anyway, those things remain legal in large part due to very aggressive political efforts against those who have sought to ban them (and at times, partially, did). Many west coast cities, for instance, *are* banning new installs of gas stoves. And more broadly, Rich Trumka, head of the US Consumer Safety Commission at the time, *did* explicitly say a blanket ban on gas stoves was on the table. This produced an enormous backlash which ensured it would not happen. The backlash (and preemptive action) matter. The gun rights movement is another example - gun laws are where they are today because many people fought very hard for it, not because it just magically fell into place.
Furthermore, if truly nobody is coming to take our cars, then a law enshrining this should be zero-cost and obvious common sense. The reality though, as we see in these very comments as well as the proclamations of various technologists, is that many *do* seek to ban human driving, which is exactly why such a law is needed.
Fair, I think. The thing we need, maybe, is more democracy and less government-by-the-regulators, because the gas-stoves-ban-discussion is one of those things that was floated by a regulator and, to your point, was quashed through the opposition of the people. These sort of paternalistic banning beef because of methane suggestions (which we're currently seeing on the right in the form of vaccine availability; I'm not meaning to suggest it only goes one way) seem much more often to be done through the administrative state than the elected branches.
Ultimately, I would have as much issue with a law that says "you get to drive a car" as I do with a law that says "you get to breathe air." I don't think either is singularly necessary, but since that's the case, there's no harm in it. If people change their mind, then the law can change with it.
But all this is well downstream of the underlying issue: driverless cars appear to be safer and better than human drivers, and their acceptance is a good thing. I noted that Kelsey Piper said in a separate column today that if X deaths are a consequence of the first amendment, then she'll take that trade, because the first amendment is that important. I personally don't get that worked up about driving, but I think it's a reasonable position to think that driving is so important that we have to be willing to make sacrifices to accept it. We do with drugs and alcohol and guns, after all. Kelsey's answer to the first question ("We should allow driverless cars, which are good") seems (a) true, and (b) a real stretch from the second question ("Should we stop humans from driving?")
I do think that on some time scale a ban is reasonable. The people who thought cars shouldn’t be allowed because some day we would ban horses in city streets were factually right, but morally wrong.
Good point.
“they are scared that as soon as Waymos are safer than human drivers, we will take their cars away.”
This is a common fear that leads people to resist technological and social change, and arguably liberalism more broadly. Some other versions of this fear I’ve seen:
- Zoning reform means banning single family homes.
- Incentivizing electric cars means taking away your ICE car.
- Accepting gay and trans people means everyone will be pressured to identify as queer or nonbinary in some way, and to reject heterosexuality and traditional families.
It’s basically “allowing a new thing will mean the old thing will be banned or heavily socially sanctioned”. I won’t say this fear is never rational, but most cases I’ve seen have been.
It is probably worth acknowledging that--once AVs are widespread enough to command large market-share--the rational thing to do *is* to ban human driving, at least in high-traffic areas (if anyone even wants to drive?) Whether or not that becomes a culture war talking point remains to be seen...
Just like the point of banning fake meat (in Florida) and banning the word “oat milk” (in Europe) is that people are worried that some day we will ban real meat and real milk, and they don’t like the idea that they themselves will be happy with that when the time comes. They want to lock in their values now.
For years, I've wanted a self-driving car. On longish trips, I'd love to do something else instead of staring out the windshield at six lanes of traffic bounded by endless billboards. I live in a small town in Florida and -having- a car is useful but -driving- my car is often a time-waster.
Open up your vision, having self-driving cars isn't just a bus-replacement.
It’s also a bus enabler! The city has enough buses to run the rush hour schedule 24/7 - it’s just the driver costs that make it not economical to do so.
That's true but focused on urban dwellers.
The US is a -very- large country where long distance travel is a norm. Even in extended cities like LA, I've known people who made 150 mile daily commutes. Mostly freeway travel, self driving, privately owned cars would be a god send.
Americans make long trips by car as a norm. I live in Florida and airline service goes from hub to hub, Atlanta to Orlando, Atlanta to Miami, there's very little schedule air service between e.g., Pensacola and Miami, a distance of 670 miles. Most states outside the Northeast have long, skinny routes. Tennessee for example is the longest state east to west and lets not think about the Great State, Texas.
All of these are prime markets for self driving privately owned vehicles.
It doesn’t really matter how big a state is. No one chooses their destination based on whether they are crossing state lines. They choose based on whether there is a closer destination where they can do the thing they want to do.
The vast majority of driving is in town, and the vast majority of town driving is on trips that are under 15 miles each way. In any case, self driving helps buses far more than it helps cars (because it’s a predictable route and it is cost-efficient to run on a constant schedule all day), even if buses and cars serve different types of travel markets.
At bare minimum, then, an explicit law protecting the right of drivers to drive their own vehicles (including motorcycles, too) on all public roads, *in perpetuity*, with robust enforcement mechanisms, ought to be a precondition to the mass rollout of self-driving cars. Too many times we have seen the safetyists turn the novel and optional into the mandatory. Hundreds of millions of us simply will not stop driving the old fashioned way, nor give up our vehicles, and we could easily throw a big fat wrench into the works should we choose. Work with us or fight us: those are the options. For the rollout to go smoothly, people like me need to trust that this isn't the thin end of the wedge to banning human driving (which a fair number of technologists are already excitedly blabbing about), so it ought to be very much in the interest of AV proponents too.
The law will of course need to also cover punitive differential tolling and registration fees designed to keep human drivers off the road via the backdoor.
Earnest question--why do you feel so emotionally strongly about this? Why is it important to be able to drive, especially in big, congested metros? It just seems kind of low stakes to me.
Frankly I'm not interested in litigating this question. It doesn't and shouldn't matter.
I'll just say that:
(1) if you've never taken a spontaneous self-directed road trip for days or weeks on end (or experienced the thrill of the open roads from the seat of a motorcycle), we're not arguing from remotely the same plane and cannot and will not see eye to eye. This is not purely a technocratic calculation.
(2) even beyond that and into the mundane, autonomy, self mastery, and humanism are all crucial values, and I'm sick and tired of ceding more and more of my basic actions to machines I can't control
(3) policy, especially around technology, is too often shaped around the perceived needs of major metros while ignoring everyone else. California for instance is soon to ban fuel powered generators, a disaster for the mountain regions where long power outages are the norm and in cold snowy winter conditions
(4) the surveillance and top-down control implications of all this are terrifying.
I mean, you seem very interested in litigating this question and seem to think it does matter a lot.
I will say that:
1. I don't like spontaneity or road trips. I've done it; don't like it; way too much time on the road. Being in a car passing through the outdoors is not the same as actually being outdoors. Definitely a fly + preplan the itinerary + optimize for minimal travel time guy. I particularly enjoy travel to outdoorsy destinations that don't require a car (e.g. the Swiss Alps, Japanese Alps, western Cornwall, the Cote d'Azur). I'd love to have this same option in Yosemite, Big Sur, Southern Utah, etc.
2. I'm not sure the evolution of cars from human-driven to machine-driven is meaningfully more extreme than any other technical advancement of the past 200 years. And I definitely don't think >5% of the population wants less technological advancement given, you know, revealed preferences. Is this any different than ceding "autonomy" over washing your clothes?
3. I'm totally fine with making policy for metros and rural areas differently. Ideally, fuel-powered generators would be allowed in some counties/census tracts and prohibited in others.
4. More terrifying than the status-quo where people just get hit by cars all the time?
I mean, this obviously seems like a difference in values. I absolutely am biased towards the technocratic approach. I suppose the relevant question is...how many of you are there, given your "Work with us or fight us: those are the options" approach? Are you one person or a constituency?
Also, do you want a constitutional amendment on this? I'm not sure any law can be guaranteed in perpetuity given the nature of elections + legislation.
I'm at work now and will respond properly later. A few quick points:
1. By "I'm not interested in litigating this question", I specifically meant the *why* of "why don't I want to forcibly delegate all human driving to machines". It's a statement of philosophical principle that it ought not matter why, basic norms of liberty-as-baseline alone ought be sufficient.
2. Right now I just have a WordPress site at a cheap domain with one half-written post. I spun it up and started thinking about it only last week after seeing yet another tech founder crowing about how no humans will be driving in a decade. I plan to start spending 5-10 hours a week working on this as a side project, probing the lobbying/legal space, building a mailing list, and investigating best courses of action.
3. Based on surveys and various other sources of information, I do believe more Americans are in my camp than your camp when it comes to driving. I'd roughly put it at 45:25 (30% fuzzy middle) on the question of personal preference, and 70:15 on the issue of bans, if I had to guesstimate.
4. Preliminarily, I think a well written federal law could get 80-90% of this, but an amendment may be another approach.
5. As to your point #4, yes, I will absolutely take that tradeoff. Freedom over safety, in all things.
You are awfully eager to consign a bunch of other people's children to readily preventable deaths just so you can enjoy the romance of the open road. The parents of victims of traffic violence might have a different perspective, and you would come across as a more reasonable and decent person if you at least acknowledged that perspective.
I already did so in acknowledging that there is a tradeoff that exists - more broadly I'm a big advocate of people in policy discussions taking this road rather than pretending their approach has no tradeoffs and is all good and no bad.
Beyond that my view on this is exactly homologous to Charlie Kirk's now-(in)famous view on guns. I have no interest in playing these silly games of weaponized empathy where I have to do a bunch of performative crying before I can straightforwardly make my point. It's the COVID discourse all over again.
Car accidents kill 2,000 children a year. Drowning kills 1,000, basically the same order of magnitude so a useful comp. If I insisted that your right to a swimming pool callously disregards hydraulic violence against children and that parents of said children would find my freedom-first approach loathsome, you'd rightly dismiss me as insane. Bad things happen - it's the law of large numbers.
(To clarify, I'm not implying that's your view: I'm making an analogy to try and find your limiting point on the "safety first" approach.)
You are probably correct about 45:25 (30% fuzzy middle) on the question of personal preference and 70:15 on the issue of bans, but that strikes me as a softly-held belief. Most people haven't actually seen/used Waymo yet; to them, it's abstract.
We will see how those numbers evolves once 1) Waymo goes national and 2) once Waymo has been around for a decade and is a normal fact of life, rather than a novelty
If I am to guess at how things will actually evolve, I'd say that come 2035, most 16-18 year olds won't bother to learn how to drive and there will just be generational turnover on this as the 21st century progresses. Same with the gradual phase-out of manual cars, even if nobody is actually going to ban them.
That’s really awful. I wouldn’t want snake oil peddlers to have threatened to ban more effective medications without a law locking in the legality of their more dangerous treatments in perpetuity just because people like them.
I take Waymo about once a week (in Palo Alto) and it is just such a better experience than driving. It’s tech
innovation at its very best.
Human driving is inherently faulty. 40,000 auto accident deaths per year represents death on an unforgivable scale. It's persistent death — 40,000 *each year* — and yet, this number actually seems to have *decreased* since the time when I went through driver's training, a while back, when I was advised that the number of annual highway deaths was more like 50,000…
So, human drivers suck. No argument. We all see examples of bad, and even criminally bad driving, just about every day. And then, there are the drunk (and/or high) drivers, whose bad judgment and selfishness make this unforgivable circumstance even more unforgivable.
The one best and most excellent feature of our system of human driving however, is its intrinsically compartmental architecture. A bad driver is only one bad driver, and not a systematic flaw of bad driving which affects more than a single car. Although there are tens of thousands of these individual bad driving flaws each day in the U.S., at least they are all *isolated* flaws.
The problem with computer-operated cars is the same problem with computer-operated anything. Because software is so incredibly complex, and because computers themselves operate software with only with the most unforgiving inflexibility, there is an astronomical number of potential failure points in the process of performing any task — let alone one as complex as operating a motor vehicle. Any human mistake in design or coding at any one of these failure points could create a catastrophic disaster affecting all users of the system, and not just a single jerky driver who thinks he or she owns the road while driving 85 mph along the breakdown lane of a highway, past queues of crawling traffic during rush hour…
Plus, there are security risks. Any computer can be hacked, therefore any car which is operated by computer is vulnerable to a potentially fatal software intrusion, perpetrated by a bad actor — a psychotic actor, or a sociopathic actor with a bad motive. Imagine the consequences of hacking an entire self-driving network, extending along all roads within a whole city, or every city. In a whole country…
It's an enormous risk that can't be ignored. Any decision to abandon epidemic, although isolated, cases of bad human driving would need to consider the risk acquired by replacing it with an inhuman system with so many potentially fatal points of failure which, if encountered, could create death and chaos on a mass scale.
Not so fast!
I can see how some sort of undetected flaw in a self driving system update could somehow cause chaos - all the cars might stop one day. But I don’t see how they could cause death and destruction - all of these systems have a much deeper control that makes them stop when things look dangerous, so if something made them start doing things that causes danger, they would all just stop.
The design feature of self-driving software, to stop in case of danger or failure of some kind, is reasonable — but what if this fail-safe is triggered during a driving situation where it's actually unsafe to stop?
Also, who knows what sort of danger human passengers in computer-driven cars might face, if their cars all become inoperable at the same time, in a catastrophic scenario — or even in a single car, during an isolated "client side" failure?
Think of how often software versions are updated to correct "bugs". No matter how carefully programmed any code may be, it's oppressively difficult or maybe even impossible to get 100% right, even after many patches, so it's reasonable to assume that there will always be software flaws — and every flaw is a potential failure point.
…
Another possibility to keep in mind is a scenario where a hostile force might hack the software, reprogramming it to begin driving each car it operates very dangerously. We've seen military forces hack into enemy computer systems before, or otherwise booby-trap enemy hardware, to create disruptions and cause casualties. The massive failure point of using the same software to determine the movement of millions of cars enabled this possibility.
…
Think of nuclear power as an analogy. It's tempting to generate gigawatts of carbon neutral power, and while the designers of nuclear plants definitely go through a lot of trouble to perfect their architecture and safety systems — because these systems are human-engineered and constructed, and because they're subject to the unpredictable forces of nature while they're in use, there's simply no way to make them truly fail-safe. The consequences of such failures can be so great that even a slim possibility of one must be, and is, taken into account while deciding whether to exploit such a tempting route to massive amounts of carbon-free energy, or not.
…
I can't say whether these risks are more or less dangerous than the risks we face now from flawed human driving, but I am saying that they should be taken into account while deciding whether computer-driven cars are a good idea.
There’s many reasons why it’s a good thing that we started with a few dozen cars in Phoenix and moved to a few hundred in Phoenix and SF and are now at a few thousand in five cities, rather than overnight putting millions of them on the roads in all cities at once. Understanding the vulnerabilities to system updates and failures is part of that.
Failures may happen at any time. Security may be breeched at any time. It's the same thing with any large-scale, inherently risky ambition — no matter how carefully planned a project is, there will always be risks which should be taken into account while judging whether they're worth assuming, just to replace other risks which are already in place…
Or whether there might be other, less risky options with more upside to consider.
The old people bashing - while very stereotypical - is simply wrong on its face. 18-29 is pretty significantly more accident-prone as a demographic both in terms of per 100k miles and per licensed driver. Deaths go up after 75, but that appears to be in part the result of 80 year-olds being quite frail so crashes involving them are more likely to kill them.
As a cyclist commuter in a major city, I’m a hardcore autonomous vehicle fan. But I think at the end of the day this boils down to whether we want to be a country of automation accelerationism, or we want to keep a status quo in place where we have 3 million people employed as truck drivers and another few million working for rideshare companies. To me it seems clear that pretending away the issue of driver automation won’t work out in the medium to long term. But I also understand that these careers are highly engrained in the American ethos going back a century.
I appreciate that this article takes into account that Americans actually do really enjoy driving their own cars. In the future, I would love to switch on my AI-assisted self-driving technology on long road trips when I want to read, work, or eat a road trip snack. What leads to the most traffic accidents is distracted driving, I would venture to say, people trying to scarf down their sandwich while trying to help kids eat in the back seat has led to way more accidents than a self-driving taxi ever will. Ignoring technology that can make the roadways safer is foolish.
What I am scared of isn't the lack of jobs available for Uber drivers, but the safety standards used to determine that AI can not only drive in San Francisco but also on all of the highways and byways in the rest of America, in thick traffic, slippery roads, and in road construction.