Question: One aspect that you did not discuss on the cancer issue is how early(ier) diagnoses AND TREATMENT may be affecting morbidity and mortality. I don't know if the "JAMA" article discusses this but it would certainly be a consideration to any conclusion one might make about whether there is a real increase in cancer vs. a perceived increase in cancer based on outcome.
And so it came to pass that both sides became afraid of change, not simply moderation, but truly afraid of the future unless one could know, with zero uncertainty of what change would bring. This certainty being, of course, impossible, the people calcified and became as stone while wishing for things to improve.
The thing about those 40K deaths is we could already avoid a huge percentage of them with relatively small interventions like speed cameras and daylighting. The fact that we choose not to do this is a real indictment of our society. Nobody, and I really mean nobody, was willing to inconvenience drivers to save those people before self driving cars. So I am not sure why they'd be willing to inconvenience drivers after. The revealed preference of America is that inconveniencing drivers is never worth it no matter what.
Hmm kinda but buffing up traffic enforcement does and it's like the easiest thing in the world. Hoboken was literally like we just want to stop people dying and they didn't have a death in 7 years. It is legitimately not hard to stop people from dying on the road, the US just has no political will for it.
For the reasons stated by Buzen it definitely is not easy to stop people from dying on the road as there are incredibly diverse roadway systems and municipalities that don’t have single solution options. Unless checkpoints go up on every road, drunk driving will continue to be a third of all fatalities. Speeding reduction would require a camera every few miles on every road in the country. Distracted driving will always be impossible to enforce unless cameras in everyones car are mandated by law. None of this is easy.
Yes you do realize the entire point of this two week old comment thread is saying the people don't want to save those 40k lives. That's literally the whole thing I was saying.
The premise of the column is that Derek is reframing self driving cars in a way that saves lives to convince people they are good. I am saying nobody cares about those lives anyway, so I don't feel anyone will be convinced by such a framing. I wish people cared but they do not.
Well, towns in CA have been given the ability through a new state law to lower speed limits significantly, particularly around school areas. Some have reduced speeds to a slow as 15mpg (when "THE CHILDREN" are present).
Maybe pushing your car around is the ultimate solution?
One of the biggest reasons to be against self-driving cars is how cheap they make driving, which can massively increase pollution as well as decrease the ability for better public transit options to take hold (much better on safety AND pollution). Not to mention, fully autonomous drivers can easily go wrong and cause massive traffic problems, with nobody to really hold responsible. Everyone has seen videos of confused Waymos driving around completely empty, causing traffic and pollution to serve nobody.
When Self-driving Cars are really working in all circumstances, owning a car becomes meaningless. For most people being driven to their desired destination simply becomes a service like public transport as an additional option. Think of all the space in our cities blocked by parked cars that sit there more than 95 percent of the time.
And if you do this and attach a marginal cost to each car trip ($0.35 per minute or so), then people more intentionally consider if each trip is worth taking in a car. People will reevaluate if they'd rather pay $1 or walk 18 minutes.
But if they replace car ownership then we must assume people will still hail them to and from work, meaning we still need as many cars as we did before and we still have to store them somewhere in between rush hours. I am not against this tech. It will make driving safer, but it will not solve other car-related problems like congestion.
Nope. Autonomous cars will put multiple people going to the same place in single cars, resulting in LESS cars on the roadways.
Also, since these would be rental cars, they would be part of a fleet and redirected to serving other needs during work hours, such as package/food delivers.
With human drivers removed from the highways (and hopefully, all roads), speeds would be much higher because autonomous cars would be networked and constantly communicating.
And don't forget that each year into the future is going to result in machines (AI/robots/automation) replacing more and more human workers, so there will actually be LESS people commuting to jobs.
…but, obviously, the car industry - operating under its current working assumptions - is against this. They would have to change from producer of vehicles to mobility provider.
I honestly don't think manufacturers will care, nor is it certain that hordes of people will want to do ride shares. People already spend enormous sums of money just to drive to work and run errands. Cars and trucks are a status symbol and are not going to easily be replaced for many.
I would alsopoint out that Uber and Lyft drivers trolling for rides in major cities are also major problems for traffic. It is not just a Waymo problem.
A trip to NY city will provide ample evidence of this as more than half the uber and Lyft cars travel around the city empty looking to capture a rider.
Most of the uber and Lyft vehicles actually burn gasoline as opposed to an EV contributing to ground level pollution in the city. The same can be said for Taxis in NY. Accountability is an issue but at least there is significant data associated with Waymo and other driverless cars including reams of video. Though some Driver operated cars have video and sensor data it is not as widely available and depends on the driver and the age of the vehicle.
Reducing traffic seems like one of the most effective uses of self-driving technology. Driverless buses and trains allowing increased frequency would reduce traffic, and avoiding toxic human-driver behavior such as tailgating, aggressive merging and so forth. All things being equal driverless technology could plausibly be much better for traffic. Air-pollution can can already be solved by electric cars, so that doesn’t seem like a good objection. You’re also not addressing the safety issues pointed out as being most impressive.
I can assure you, there are way more humans causing massive traffic problems than a few meme vids on YT, e.g. every fender bender on a crowded road can tie up traffic for hours. Countries without self driving cars are not without traffic jams after all.
As mentioned below, in a world of self driving cars, there's no reason to even own a car, so the total number of cars will be way, way less... Most cars - which take enormous resources and cause enormous pollution to build when you take into account all of the aspects of mining the resources, producing the parts, assembling the car, etc, and in the end disposing of the car - spend most of their time sitting idle. An efficient system of on demand self driving cars could handle all of the trips we take now in a fraction of the number of cars, eliminating the waste associated with little-used cars. It's trivial for an AI to figure out how to combine multiple passengers going to and from the same places at the same time into the same car, thus handling high volume times like rush hour and preventing the need for the fleet volume to be scaled just to the high volume times. Fewer cars then has downstream impact on other resource usage like parking, parts production and maintenance, etc.
Ultimately the argument is more based on safety than economics but the economics are definitely positive as well.
The cars are electric. And the more they are adopted, the less cars will need to exist. They will drive 24/7 except for repair and charging (except some resting during low demand). So less lithium mined, and less city space taken up by parking.
Self driving cars are better than regular cars in every way. They’re safer as you pointed out. But they’re also better at enabling sprawl and causing congestion, than regular cars have.
It’s controversial but indisputable that building for cars made American cities worse places to inhabit (and, since it’s a focus here: less healthy). Just compare them with most any European city.
In my opinion, self driving cars are part of the Perfect City, but given our car-centric US development, if we don’t implement them carefully, our cities will become objectively worse places to live.
I am not a Waymo apologist and am not sure I am ready to ride a Waymo myself, but in SF recently, I realized that many or most of these cars are electric and don’t pollute (directly).
Also, for cities that developed after WWII and don’t have a central core or commuting grid conducive to widespread use of mass transit, super cheap self-driving rental vehicles can be a short cut to mass transit.
I think the traffic jams are a solvable issue where better software and coordination. It’s certainly no different from the fact that transit systems sometimes have service delays and outages.
I think we are decades away from Waymo being ready for a city of chaotic traffic behavior and mass jaywalking like New York and many European cities. All the cities in the US where Waymo have been adopted have much more civilized traffic patterns.
This is a general problem with making something more tolerable. People think, "I can't wait to have a self-driving car, I'll get to take a nap during my hour commute." Except your commute will no longer be an hour, everything will adjust to find the maximum that people will put up with.
Making driving cheap is a good thing! Many people need to drive because of location, safety, disability or what they are transporting. Mass transit is not safe, comfortable, reliable or convenient in most cities in the US and non existent in rural areas. Building new subways or dedicated bus lines is much more expensive than cars that use existing streets. Waymos and Teslas are electric and don’t pollute much.
I agree that self driving cars are much safer than Human drivers..but you need to be careful in comparing Waymo to drivers …Waymo is restricted to certain urban areas, these areas my have an inherently lower fatality rate than the country as a whole..a better comparison would be to limit the analysis to the areas served by Waymo.
I live in San Francisco. It is rather obvious that Waymo’s are much safer than human-driven cars. They drive the speed limit, observe traffic signs/signals and don’t drive while intoxicated. We had a pedestrian fatality in my neighborhood this weekend where a car was driving 50 mph through an intersection with 4-way stop signs and killed a 30 year old man (and drove off.)
I have worried that self-driving cars further undermine public transportation (which is what I prefer,) but we need both better public transportation and more self-driving cars to improve safety and quality of life.
The study cited in the linked article used adjusted equivalent traffic, vehicle and road types for the comparison, so the researchers were not careless or deceptive.
I guess urban areas (maybe?) have lower fatality rates per mile traveled, but the majority of fatalities are in urban areas. And the number includes pedestrians.
I've heard this argument before but I just don't get it. I feel like pretty much anyone would agree that urban driving is way more complex than highway or suburban driving, right?
But complexity should be the issue we're worried about for self-driving cars. On the highway the main danger is failing to maintain attention and drifting out of your lane, which shouldn't be a problem for a robot. The things that are hard for automated systems are edge cases outside of their training data, which I imagine should be way more common in cities. If Waymo has figured that out it seems like the main hurdle is cleared.
They are strictly controlling where their cars are allowed to drive, for a reason…I still think they are safer, maybe not by as much as the article quoted…
It was really only a comment about statistics, if you want to compare things they need to be fundamentally the same…city driving is a subset of all driving so making a comparison without checking to see if they are the same can lead to errors..
This would be a market approach to addressing these deaths, so no specific spending would really be needed. If there are costs wouldn’t these be borne by Waymo?
I don’t think the analogy between vaccines and self-driving cars is apt, for a couple of reasons.
What people don’t understand about autonomy is how strong the Pareto Principle is here. It seems like driverless cars are “almost there”, but the cases they can’t handle are really, REALLY hard. I’m a believer that artificial general intelligence is possible in principle, but I don’t think we’re anywhere near autonomous driving technology that can make a car go anywhere a human can drive it.
Additionally, some companies (Tesla is one, Waymo isn’t) are submitting innocent bystanders to their experimentation with unconscionably little concern for safety protocols. If this were a vaccine, would you be okay with the drug company inoculating random people without their knowledge or consent? I’m fine with cities banning driverless cars from companies that don’t meet a very high standard of caution in how they verify and roll out new updates.
Last, autonomous cars are vastly more expensive than vaccines. Even if the perfect driverless car existed, it would take probably 1,000 to 10,000 times more money to replace the human-driven fleet than it would to vaccinate everybody. The number of lives saved might be the same, but the cost matters. At that price, you have to seriously consider whether you could save even more lives with different spending priorities.
Where do your numbers come from? Also how is waymo operating if the cost are so high? Or are you referring to only the cases that self driving can’t handle? In which case why couldn’t we have self driving cars and human operators for the extra difficult cases?
Yeah, some would be for hire, but not all. You don’t have kids, do you? You don’t sound like you’re acquainted with the squalor of a family vehicle. Who wants to share that mess with random strangers? Many, probably even most, people want their own vehicle, autonomous or not.
The CDC’s vaccine price list has single doses of vaccines going for around $50 in most cases. Experian has the average price of a new car at $50,000 in the US, and there are almost as many cars as people in this country. That average is dominated by non-autonomous cars. Vaccinating everyone in the US would probably cost quite a bit less than $50 a head. Even with economies of scale, replacing every car in the US would probably not cost leas than $50,000 per vehicle, and maybe a good bit more. Therefore, the best case for autonomous cars is that they cost 1000 times more than vaccinations, and I could easily see the ratio going as much as a factor of 10 beyond that.
Waymo is operating at a loss. It is funded by Google’s limitless advertising revenue, because they’re willing to take big risks on possible future revolution.
FWIW, I think Waymo is doing it right. They’re appropriately cautious and safe, and they stick to small areas they can map really well. They don’t advertise any unrealistic timelines for replacing all human drivers. I would vote to allow Waymo in my city. I would vote against Tesla.
Existing cars spend over 90% of the time parked while autonomous vehicles for hire would spend less than 10% of their lifetime charging or being cleaned, so your cost calculation is over 9 times too high.
There is a technical nuance here that I feel you really should highlight. Waymo's approach to self-driving cars is safe and effective. Tesla self driving cars are the poison pill that could wreck the public opinion. Unfortunately it's not obvious to the average person as drinking bleach vs Moderna vaccine since the complexity of the underworking of how self-driving is accomplished is hidden.
I do agree overall that mass adoption of self-driving cars would be an overall boon to Americans.
How is Tesla more dangerous than Waymo? They have only just started testing in Austin but there isn’t enough data to compare their safety versus Waymos.
I have heard that some significant part of Waymo's gains in safety disappear if you simply control for speed -- that is, a huge portion of human-at-the-wheel accidents happen because the humans are driving faster than they can safely control, and even given that a crash occurs, the difference in the rate of hospitalization injury or fatality, between hitting somebody going 25 mph and 35 mph is _enormous_. And people drive 35 on 25 mph streets all the time. In many suburbs they go even faster than that, because the streets are designed for it. ( https://archive.strongtowns.org/journal/2021/8/6/the-key-to-slowing-traffic-is-street-design-not-speed-limits )
I would be very curious to see a version of this research that lays out these stats in detail.
California considered a law a few years back that would've required automatic speed governors that would use GPS and map data to hold cars to the speed limit under most circumstances. It's possible that this would also radically reduce mortality rates. I have the impression though that a lot of human drivers find this idea even more intuitively-repulsive or offensive than simply taking a self-driving ride.
One huge note here is Tesla full self drive is exponentially cheaper to deploy fleet wide, is either safer or as safe, and is quicker to deploy to unmapped roads. Yet progressives hate Elon so will never allow it to go forward. But it's a more elegant and simple tech solution to the problem and would absolutely save lives.
Since the data presented doesn’t include Tesla only Waymo, I don’t think you can realistically expect the same stats to apply to Tesla’s cars. Which has a totally different technology model than Waymo.
Very different tech stack but the data for crashes per vehicle miles driven from tesla FSD is far better than human driver and essentially on par with Waymo. We need a larger sample size out of Waymo to actually fully compare them to Tesla given Tesla has so many more vehicle miles driven on FSD. But contrary to the media narrative driven by reaction to Elon FSD is far safer than humans by the numbers. The lack of lidar is actually probably a plus overall for safety given it allows better AI by decreasing complexity.
Yeah. The fact that Mark Rober was able to Wile E. Coyote his Tesla into a trompe l'oeil road (3/2025 - road that was actually a wall of paper and cardboard), that would have been easily detected with lidar, is a great example of failing to test for obvious requirements.
The technology is very similar, both use detailed maps and machine learning algorithms for driving, both have multiple cameras and object recognition and proximity sensors. The main difference is Waymo also has LiDAR, but it’s not clear how much this improves performance except during low visibility conditions, which are not that common even in the foggy city.
Why do you keep saying "accident"? The vast majority of crashes occur due to choices in physical and regulatory design. You don't need new technology to dramatically reduce fatalities. However, the new technology is welcome, but does need a thoughtful regulatory framework. E.g. existing land use regulations kill people (in addition to other negative externalities) on roads because in most places mobility is impossible without driving. Self driving regulations should take heed to the mistakes of generations past to encourage more pro-social outcomes.
Well you could lower speed limits to a crawl, have incapacitated driving lock outs in all vehicles, ban bicycles and pedestrians and physically jam any cellphone signals in vehicles and you might approach Waymos injury records, but it wouldn’t be preferable or politically possible.
AVs are not a panacea. Many of the greatest benefits of automated vehicles won’t be achieved until fleet penetration is over 80%. Then vehicles can travel more closely together reducing traffic and, potentially infrastructure costs. In the meantime automated vehicles could reduce costs and increase demand for driving leading to increased traffic, pollution, and sprawl. Other commentators are right to note that there are some bad actors in this space who are willing to sacrifice safety and transparency for marketing and sales. Finally, cities are right to be wary of being used as laboratories and should use what leverage they have to achieve sustainable outcomes that preserve livability and accessibility.
Not Just Bikes went really into detail to the downsides of Self Driving Cars. I disagree with a straight up ban but some food for thought for other ways that maybe we should regulate them https://youtu.be/040ejWnFkj0
That video just rehashes technophobic anti capitalist fears while also claiming self driving cars don’t actually work, but are actually driven remotely.
This is by-far the best argument that I’ve come across that driverless cars are unnecessary and even detrimental to improving cities. Most arguments are simply about how some driverless cars had something go wrong. This actually addresses what the problems of driver centric cities are and how self driving vehicles don’t solve those problem while adding problems of their own.
Also clearly the answer to most of the problems of cars is to prevent cars from causing those problems in the first place but making them unnecessary for urban transportation. Trains, public transit, bikes, personal mobility solutions all do a better job at fixing the problems of cars than self driving cars. All without the risk of being completely controlled by unaccountable greedy corporate entities.
Every Tesla owner I know admits to using the self-driving feature much or all of the time. It might be interesting to look into the overall accident rate of Tesla owners vs a matched cohort of owners of cars with no autonomous features. Also, is there any goal of having self-driving cars "talk" to each other on the road? In my mind, that solves all the "traffic jam" and "pollution" concerns in this thread, and might make driving I-5 bearable between SF and LA.
Hi, I'm a Tesla owner who does not use the self-drive feature at all. I use the "driver assist" stuff like adaptive cruise control sometimes, but the I don't trust the "autopilot" at all.
I am confident that self-drive will be safer than me someday. I do not feel that day has arrived yet.
(To be fair, I am usually pretty conservative about my speed on city streets or in traffic, and it's quite likely that many other humans are lot less safe than I am. Although I will cop to the fact that if I have a really open stretch of highway with clear visibility, I enjoy driving fast.)
You could easily see increasing real diagnoses and lower death rates if improvements in treatment are saving those lives. Breast cancer, especially HER+ versions, have seen some pretty great targeted therapies in the last couple decades. Saved my wife’s life!
Couple of thoughts on your provocations: the risk of going first, the vulnerability needed for common ground and the role or behavior of reputation. It takes a long time to earn and shape but can be wiped out or destroyed instantly. With all the accidents, deaths and scenes of self-inflicted traffic jams or honk fest reputation of Cruise, and in general autonomous vehicles was swiftly torn down. Doesnt help that SF and other cities can feel like living laboratories. If you dont know what is the product you’re the product could be, if you dont know who are the mice, in the experiment, you are? Similar pattern for vaccines. Which makes pulling dull edged policy levers that alter the flow of the above a scary proposition. Many orgs and elected officials struggle to successfully co-create w communities, in general, which undermines how to establish, communicate any ‘living in beta’ change, renovation, alteration etc. 2cents. Thx for the work you do to put these together.
Question: One aspect that you did not discuss on the cancer issue is how early(ier) diagnoses AND TREATMENT may be affecting morbidity and mortality. I don't know if the "JAMA" article discusses this but it would certainly be a consideration to any conclusion one might make about whether there is a real increase in cancer vs. a perceived increase in cancer based on outcome.
And so it came to pass that both sides became afraid of change, not simply moderation, but truly afraid of the future unless one could know, with zero uncertainty of what change would bring. This certainty being, of course, impossible, the people calcified and became as stone while wishing for things to improve.
No one likes change, except for wet babies.
Witty comment of the week!! 🤣🤣🤣
This made me laugh out loud
The thing about those 40K deaths is we could already avoid a huge percentage of them with relatively small interventions like speed cameras and daylighting. The fact that we choose not to do this is a real indictment of our society. Nobody, and I really mean nobody, was willing to inconvenience drivers to save those people before self driving cars. So I am not sure why they'd be willing to inconvenience drivers after. The revealed preference of America is that inconveniencing drivers is never worth it no matter what.
Speed cameras do nothing to stop drivers impaired by drinking, fatigue, youthful abandon or using their phones, while autonomous driving vehicles do.
Hmm kinda but buffing up traffic enforcement does and it's like the easiest thing in the world. Hoboken was literally like we just want to stop people dying and they didn't have a death in 7 years. It is legitimately not hard to stop people from dying on the road, the US just has no political will for it.
For the reasons stated by Buzen it definitely is not easy to stop people from dying on the road as there are incredibly diverse roadway systems and municipalities that don’t have single solution options. Unless checkpoints go up on every road, drunk driving will continue to be a third of all fatalities. Speeding reduction would require a camera every few miles on every road in the country. Distracted driving will always be impossible to enforce unless cameras in everyones car are mandated by law. None of this is easy.
Every locality that invests in safer streets has success. If it's so hard how come it's so easy when you try? It's merely a matter of political will.
You don't understand how politics work. It isn't about political will, it is the desire to get elected (or reelected).
Simply because a relatively few people such as yourself want government to act as a mother, does not mean that the majority wants this.
Yes you do realize the entire point of this two week old comment thread is saying the people don't want to save those 40k lives. That's literally the whole thing I was saying.
But, wouldn’t self driving cars still save those 40,000 lives? While also not having to do those simple things? If so where is the downside?
The premise of the column is that Derek is reframing self driving cars in a way that saves lives to convince people they are good. I am saying nobody cares about those lives anyway, so I don't feel anyone will be convinced by such a framing. I wish people cared but they do not.
It's a fair question I wasn't that clear.
Ah, thanks for the clarification. I sadly agree with your assessment.
Well, towns in CA have been given the ability through a new state law to lower speed limits significantly, particularly around school areas. Some have reduced speeds to a slow as 15mpg (when "THE CHILDREN" are present).
Maybe pushing your car around is the ultimate solution?
One of the biggest reasons to be against self-driving cars is how cheap they make driving, which can massively increase pollution as well as decrease the ability for better public transit options to take hold (much better on safety AND pollution). Not to mention, fully autonomous drivers can easily go wrong and cause massive traffic problems, with nobody to really hold responsible. Everyone has seen videos of confused Waymos driving around completely empty, causing traffic and pollution to serve nobody.
When Self-driving Cars are really working in all circumstances, owning a car becomes meaningless. For most people being driven to their desired destination simply becomes a service like public transport as an additional option. Think of all the space in our cities blocked by parked cars that sit there more than 95 percent of the time.
And if you do this and attach a marginal cost to each car trip ($0.35 per minute or so), then people more intentionally consider if each trip is worth taking in a car. People will reevaluate if they'd rather pay $1 or walk 18 minutes.
There's also always congestion pricing...
But if they replace car ownership then we must assume people will still hail them to and from work, meaning we still need as many cars as we did before and we still have to store them somewhere in between rush hours. I am not against this tech. It will make driving safer, but it will not solve other car-related problems like congestion.
Nope. Autonomous cars will put multiple people going to the same place in single cars, resulting in LESS cars on the roadways.
Also, since these would be rental cars, they would be part of a fleet and redirected to serving other needs during work hours, such as package/food delivers.
With human drivers removed from the highways (and hopefully, all roads), speeds would be much higher because autonomous cars would be networked and constantly communicating.
And don't forget that each year into the future is going to result in machines (AI/robots/automation) replacing more and more human workers, so there will actually be LESS people commuting to jobs.
…but, obviously, the car industry - operating under its current working assumptions - is against this. They would have to change from producer of vehicles to mobility provider.
I honestly don't think manufacturers will care, nor is it certain that hordes of people will want to do ride shares. People already spend enormous sums of money just to drive to work and run errands. Cars and trucks are a status symbol and are not going to easily be replaced for many.
I would alsopoint out that Uber and Lyft drivers trolling for rides in major cities are also major problems for traffic. It is not just a Waymo problem.
A trip to NY city will provide ample evidence of this as more than half the uber and Lyft cars travel around the city empty looking to capture a rider.
Most of the uber and Lyft vehicles actually burn gasoline as opposed to an EV contributing to ground level pollution in the city. The same can be said for Taxis in NY. Accountability is an issue but at least there is significant data associated with Waymo and other driverless cars including reams of video. Though some Driver operated cars have video and sensor data it is not as widely available and depends on the driver and the age of the vehicle.
Reducing traffic seems like one of the most effective uses of self-driving technology. Driverless buses and trains allowing increased frequency would reduce traffic, and avoiding toxic human-driver behavior such as tailgating, aggressive merging and so forth. All things being equal driverless technology could plausibly be much better for traffic. Air-pollution can can already be solved by electric cars, so that doesn’t seem like a good objection. You’re also not addressing the safety issues pointed out as being most impressive.
I can assure you, there are way more humans causing massive traffic problems than a few meme vids on YT, e.g. every fender bender on a crowded road can tie up traffic for hours. Countries without self driving cars are not without traffic jams after all.
As mentioned below, in a world of self driving cars, there's no reason to even own a car, so the total number of cars will be way, way less... Most cars - which take enormous resources and cause enormous pollution to build when you take into account all of the aspects of mining the resources, producing the parts, assembling the car, etc, and in the end disposing of the car - spend most of their time sitting idle. An efficient system of on demand self driving cars could handle all of the trips we take now in a fraction of the number of cars, eliminating the waste associated with little-used cars. It's trivial for an AI to figure out how to combine multiple passengers going to and from the same places at the same time into the same car, thus handling high volume times like rush hour and preventing the need for the fleet volume to be scaled just to the high volume times. Fewer cars then has downstream impact on other resource usage like parking, parts production and maintenance, etc.
Ultimately the argument is more based on safety than economics but the economics are definitely positive as well.
The cars are electric. And the more they are adopted, the less cars will need to exist. They will drive 24/7 except for repair and charging (except some resting during low demand). So less lithium mined, and less city space taken up by parking.
Self driving cars are better than regular cars in every way. They’re safer as you pointed out. But they’re also better at enabling sprawl and causing congestion, than regular cars have.
It’s controversial but indisputable that building for cars made American cities worse places to inhabit (and, since it’s a focus here: less healthy). Just compare them with most any European city.
In my opinion, self driving cars are part of the Perfect City, but given our car-centric US development, if we don’t implement them carefully, our cities will become objectively worse places to live.
I am not a Waymo apologist and am not sure I am ready to ride a Waymo myself, but in SF recently, I realized that many or most of these cars are electric and don’t pollute (directly).
Also, for cities that developed after WWII and don’t have a central core or commuting grid conducive to widespread use of mass transit, super cheap self-driving rental vehicles can be a short cut to mass transit.
I think the traffic jams are a solvable issue where better software and coordination. It’s certainly no different from the fact that transit systems sometimes have service delays and outages.
I think we are decades away from Waymo being ready for a city of chaotic traffic behavior and mass jaywalking like New York and many European cities. All the cities in the US where Waymo have been adopted have much more civilized traffic patterns.
This is a general problem with making something more tolerable. People think, "I can't wait to have a self-driving car, I'll get to take a nap during my hour commute." Except your commute will no longer be an hour, everything will adjust to find the maximum that people will put up with.
All wlWaymo cars are electric. On balance, the benefits outweigh the costs/risks, in my opinion. This is what progress looks like.
Making driving cheap is a good thing! Many people need to drive because of location, safety, disability or what they are transporting. Mass transit is not safe, comfortable, reliable or convenient in most cities in the US and non existent in rural areas. Building new subways or dedicated bus lines is much more expensive than cars that use existing streets. Waymos and Teslas are electric and don’t pollute much.
I agree that self driving cars are much safer than Human drivers..but you need to be careful in comparing Waymo to drivers …Waymo is restricted to certain urban areas, these areas my have an inherently lower fatality rate than the country as a whole..a better comparison would be to limit the analysis to the areas served by Waymo.
I live in San Francisco. It is rather obvious that Waymo’s are much safer than human-driven cars. They drive the speed limit, observe traffic signs/signals and don’t drive while intoxicated. We had a pedestrian fatality in my neighborhood this weekend where a car was driving 50 mph through an intersection with 4-way stop signs and killed a 30 year old man (and drove off.)
I have worried that self-driving cars further undermine public transportation (which is what I prefer,) but we need both better public transportation and more self-driving cars to improve safety and quality of life.
The study cited in the linked article used adjusted equivalent traffic, vehicle and road types for the comparison, so the researchers were not careless or deceptive.
https://pubmed.ncbi.nlm.nih.gov/40378124
I guess urban areas (maybe?) have lower fatality rates per mile traveled, but the majority of fatalities are in urban areas. And the number includes pedestrians.
I've heard this argument before but I just don't get it. I feel like pretty much anyone would agree that urban driving is way more complex than highway or suburban driving, right?
Complex but slower speed..so less fatalities for the passengers
But complexity should be the issue we're worried about for self-driving cars. On the highway the main danger is failing to maintain attention and drifting out of your lane, which shouldn't be a problem for a robot. The things that are hard for automated systems are edge cases outside of their training data, which I imagine should be way more common in cities. If Waymo has figured that out it seems like the main hurdle is cleared.
They are strictly controlling where their cars are allowed to drive, for a reason…I still think they are safer, maybe not by as much as the article quoted…
It was really only a comment about statistics, if you want to compare things they need to be fundamentally the same…city driving is a subset of all driving so making a comparison without checking to see if they are the same can lead to errors..
Wamo's are on the highways in certain parts of CA. Possibly some other areas of the country also.
This would be a market approach to addressing these deaths, so no specific spending would really be needed. If there are costs wouldn’t these be borne by Waymo?
I don’t think the analogy between vaccines and self-driving cars is apt, for a couple of reasons.
What people don’t understand about autonomy is how strong the Pareto Principle is here. It seems like driverless cars are “almost there”, but the cases they can’t handle are really, REALLY hard. I’m a believer that artificial general intelligence is possible in principle, but I don’t think we’re anywhere near autonomous driving technology that can make a car go anywhere a human can drive it.
Additionally, some companies (Tesla is one, Waymo isn’t) are submitting innocent bystanders to their experimentation with unconscionably little concern for safety protocols. If this were a vaccine, would you be okay with the drug company inoculating random people without their knowledge or consent? I’m fine with cities banning driverless cars from companies that don’t meet a very high standard of caution in how they verify and roll out new updates.
Last, autonomous cars are vastly more expensive than vaccines. Even if the perfect driverless car existed, it would take probably 1,000 to 10,000 times more money to replace the human-driven fleet than it would to vaccinate everybody. The number of lives saved might be the same, but the cost matters. At that price, you have to seriously consider whether you could save even more lives with different spending priorities.
Where do your numbers come from? Also how is waymo operating if the cost are so high? Or are you referring to only the cases that self driving can’t handle? In which case why couldn’t we have self driving cars and human operators for the extra difficult cases?
Yeah, some would be for hire, but not all. You don’t have kids, do you? You don’t sound like you’re acquainted with the squalor of a family vehicle. Who wants to share that mess with random strangers? Many, probably even most, people want their own vehicle, autonomous or not.
The CDC’s vaccine price list has single doses of vaccines going for around $50 in most cases. Experian has the average price of a new car at $50,000 in the US, and there are almost as many cars as people in this country. That average is dominated by non-autonomous cars. Vaccinating everyone in the US would probably cost quite a bit less than $50 a head. Even with economies of scale, replacing every car in the US would probably not cost leas than $50,000 per vehicle, and maybe a good bit more. Therefore, the best case for autonomous cars is that they cost 1000 times more than vaccinations, and I could easily see the ratio going as much as a factor of 10 beyond that.
Waymo is operating at a loss. It is funded by Google’s limitless advertising revenue, because they’re willing to take big risks on possible future revolution.
FWIW, I think Waymo is doing it right. They’re appropriately cautious and safe, and they stick to small areas they can map really well. They don’t advertise any unrealistic timelines for replacing all human drivers. I would vote to allow Waymo in my city. I would vote against Tesla.
Existing cars spend over 90% of the time parked while autonomous vehicles for hire would spend less than 10% of their lifetime charging or being cleaned, so your cost calculation is over 9 times too high.
There is a technical nuance here that I feel you really should highlight. Waymo's approach to self-driving cars is safe and effective. Tesla self driving cars are the poison pill that could wreck the public opinion. Unfortunately it's not obvious to the average person as drinking bleach vs Moderna vaccine since the complexity of the underworking of how self-driving is accomplished is hidden.
I do agree overall that mass adoption of self-driving cars would be an overall boon to Americans.
How is Tesla more dangerous than Waymo? They have only just started testing in Austin but there isn’t enough data to compare their safety versus Waymos.
Tesla's auto drive software relies on cameras and there are many examples where it has failed.
Check out this story:
https://www.ai-supremacy.com/p/tesla-robotaxi-launch-was-a-scam-elon-musk
I have heard that some significant part of Waymo's gains in safety disappear if you simply control for speed -- that is, a huge portion of human-at-the-wheel accidents happen because the humans are driving faster than they can safely control, and even given that a crash occurs, the difference in the rate of hospitalization injury or fatality, between hitting somebody going 25 mph and 35 mph is _enormous_. And people drive 35 on 25 mph streets all the time. In many suburbs they go even faster than that, because the streets are designed for it. ( https://archive.strongtowns.org/journal/2021/8/6/the-key-to-slowing-traffic-is-street-design-not-speed-limits )
I would be very curious to see a version of this research that lays out these stats in detail.
California considered a law a few years back that would've required automatic speed governors that would use GPS and map data to hold cars to the speed limit under most circumstances. It's possible that this would also radically reduce mortality rates. I have the impression though that a lot of human drivers find this idea even more intuitively-repulsive or offensive than simply taking a self-driving ride.
One huge note here is Tesla full self drive is exponentially cheaper to deploy fleet wide, is either safer or as safe, and is quicker to deploy to unmapped roads. Yet progressives hate Elon so will never allow it to go forward. But it's a more elegant and simple tech solution to the problem and would absolutely save lives.
Since the data presented doesn’t include Tesla only Waymo, I don’t think you can realistically expect the same stats to apply to Tesla’s cars. Which has a totally different technology model than Waymo.
Very different tech stack but the data for crashes per vehicle miles driven from tesla FSD is far better than human driver and essentially on par with Waymo. We need a larger sample size out of Waymo to actually fully compare them to Tesla given Tesla has so many more vehicle miles driven on FSD. But contrary to the media narrative driven by reaction to Elon FSD is far safer than humans by the numbers. The lack of lidar is actually probably a plus overall for safety given it allows better AI by decreasing complexity.
Yeah. The fact that Mark Rober was able to Wile E. Coyote his Tesla into a trompe l'oeil road (3/2025 - road that was actually a wall of paper and cardboard), that would have been easily detected with lidar, is a great example of failing to test for obvious requirements.
The technology is very similar, both use detailed maps and machine learning algorithms for driving, both have multiple cameras and object recognition and proximity sensors. The main difference is Waymo also has LiDAR, but it’s not clear how much this improves performance except during low visibility conditions, which are not that common even in the foggy city.
Tesla self driving cars in action https://youtu.be/mPUGh0qAqWA
Progressives are already blocking Waymos as Derek said in the article, so additional hate for Elon isn’t much of a factor.
Why do you keep saying "accident"? The vast majority of crashes occur due to choices in physical and regulatory design. You don't need new technology to dramatically reduce fatalities. However, the new technology is welcome, but does need a thoughtful regulatory framework. E.g. existing land use regulations kill people (in addition to other negative externalities) on roads because in most places mobility is impossible without driving. Self driving regulations should take heed to the mistakes of generations past to encourage more pro-social outcomes.
Well you could lower speed limits to a crawl, have incapacitated driving lock outs in all vehicles, ban bicycles and pedestrians and physically jam any cellphone signals in vehicles and you might approach Waymos injury records, but it wouldn’t be preferable or politically possible.
AVs are not a panacea. Many of the greatest benefits of automated vehicles won’t be achieved until fleet penetration is over 80%. Then vehicles can travel more closely together reducing traffic and, potentially infrastructure costs. In the meantime automated vehicles could reduce costs and increase demand for driving leading to increased traffic, pollution, and sprawl. Other commentators are right to note that there are some bad actors in this space who are willing to sacrifice safety and transparency for marketing and sales. Finally, cities are right to be wary of being used as laboratories and should use what leverage they have to achieve sustainable outcomes that preserve livability and accessibility.
Not Just Bikes went really into detail to the downsides of Self Driving Cars. I disagree with a straight up ban but some food for thought for other ways that maybe we should regulate them https://youtu.be/040ejWnFkj0
That video just rehashes technophobic anti capitalist fears while also claiming self driving cars don’t actually work, but are actually driven remotely.
This reply rehashes technophilic, pro-capitalist hopes while implying Waymo AVs aren't driven remotely, when in fact they are
https://rodneybrooks.com/predictions-scorecard-2025-january-01/
This is by-far the best argument that I’ve come across that driverless cars are unnecessary and even detrimental to improving cities. Most arguments are simply about how some driverless cars had something go wrong. This actually addresses what the problems of driver centric cities are and how self driving vehicles don’t solve those problem while adding problems of their own.
Also clearly the answer to most of the problems of cars is to prevent cars from causing those problems in the first place but making them unnecessary for urban transportation. Trains, public transit, bikes, personal mobility solutions all do a better job at fixing the problems of cars than self driving cars. All without the risk of being completely controlled by unaccountable greedy corporate entities.
Very enlightening. Thanks for posting the link.
Every Tesla owner I know admits to using the self-driving feature much or all of the time. It might be interesting to look into the overall accident rate of Tesla owners vs a matched cohort of owners of cars with no autonomous features. Also, is there any goal of having self-driving cars "talk" to each other on the road? In my mind, that solves all the "traffic jam" and "pollution" concerns in this thread, and might make driving I-5 bearable between SF and LA.
Hi, I'm a Tesla owner who does not use the self-drive feature at all. I use the "driver assist" stuff like adaptive cruise control sometimes, but the I don't trust the "autopilot" at all.
I am confident that self-drive will be safer than me someday. I do not feel that day has arrived yet.
(To be fair, I am usually pretty conservative about my speed on city streets or in traffic, and it's quite likely that many other humans are lot less safe than I am. Although I will cop to the fact that if I have a really open stretch of highway with clear visibility, I enjoy driving fast.)
You could easily see increasing real diagnoses and lower death rates if improvements in treatment are saving those lives. Breast cancer, especially HER+ versions, have seen some pretty great targeted therapies in the last couple decades. Saved my wife’s life!
Why wouldn't AI driven cars be safer than Ape driven cars?
Crudely put, but basically right.
Couple of thoughts on your provocations: the risk of going first, the vulnerability needed for common ground and the role or behavior of reputation. It takes a long time to earn and shape but can be wiped out or destroyed instantly. With all the accidents, deaths and scenes of self-inflicted traffic jams or honk fest reputation of Cruise, and in general autonomous vehicles was swiftly torn down. Doesnt help that SF and other cities can feel like living laboratories. If you dont know what is the product you’re the product could be, if you dont know who are the mice, in the experiment, you are? Similar pattern for vaccines. Which makes pulling dull edged policy levers that alter the flow of the above a scary proposition. Many orgs and elected officials struggle to successfully co-create w communities, in general, which undermines how to establish, communicate any ‘living in beta’ change, renovation, alteration etc. 2cents. Thx for the work you do to put these together.