"If enough cars on the road were self-driving, then they wouldn’t even need the traffic lights, they could coordinate in other ways, and this would all be moot."
Enough here being precisely 100%, so long as we ban cycling and walking.
There are pedestrian crosswalks already that have a traffic signal that only activates when a pedestrian presses the button.
Cyclists and pedestrians would use these and they would make all the sdcs stop for them. Or just dart out in traffic and most of the time they stop or dodge.
Now, the dystopian future that any rational AGI+ would advise would stick a credit card reader on the request crossing button. The pedestrian would have to pay the cost of delaying everyone else, internalizing an externality :)
"What they have done is made 'Robotaxi’ service go live in Austin for select rides, but these rides remain supervised with a Tesla employee in the driver’s seat."
They're in the passenger seat, which I think is a pretty significant difference.
I routinely see Waymos on the freeway which are not obeying the speed limit. I'm sure about this, because I was surprised the first time it happened and made sure to pace them and check their speed: they were going about 70-71 in a 65 zone.
Just how good Tesla is IMO undersold in this post. It is very much a solved FSD product, minus the regulatory approval. There's no scenario where an HW4 Tesla fails at this point where a human wouldn't and I highly recommend buying one if you can afford to do so.
If Tesla did a better job marketing their newest FSD, I predict they'd quickly capture ~75% of the $50k+ car market in the US.
I agree so much with this, but there is slightly more than just policy. There’s the infrastructure of cleaning stations, charging stations with staff, programming in spots for them to idle while waiting for a ride, and checking parking lots and other areas for nuance little spots. There’s the rider call centers and everything.
It will never be limited by number of cars but that’s only one of many factors for the cab driver aspect.
But, yes, I use FSD every day and it’s a safer and better driver than anyone I know.
Private (non-Tesla) ownership of robotaxis takes care of most of these issues--the person owning the car (or fleet of cars) is responsible for charging, maintenance, etc.
Tesla doesn't publish meaningful safety statistics. In good driving scenarios, Autopilot/FSD is (probably) safer than human drivers, but since Autopilot/FSD can disengage in certain driving conditions, it really doesn't tell you much. I wish they would publish better stats.
Idk I’m in the same camp as Zvi in that “safety” is just a red herring and to some degree publishing statistics means making a concession to the Luddites.
Personally I think that “better than a P99 human driver” is good enough and I’m ~95% confident they’ve already reached this milestone.
Based on owning one for 9 months and reading tons of Tesla forum posts where previously skeptical owners did a complete 180 once FSD 13+ was released for HW4.
I trust my own experience more than any random stats.
I've been holding out on buying a new car for 10+ years waiting for full self driving. Tesla seems close now, but are we completely confident if I buy a new or recently used model Y it will be capable of unsupervised driving soon? Why not just wait a little longer?
You have to buy a car with HW4. Ideally buy the 2026 model because it adds a camera on the front bumper which should improve parking performance.
Theres never going to be a 100% perfect FSD, best you can have is a larger number of 9’s. The current Tesla is 99.9% reliable. The one in 2028 might be 99.99% reliable. Whether the inconvenience of not having a car is worth the wait is up to you.
I’d wait till the next one. Specifically, robotaxis should be feature complete when released. Or just wait until it comes out and see if it’s unsupervised. I wouldn’t buy until that’s going around.
Or you could easily buy a used Tesla HW4 vehicle, probably at half the price of new. I’d target 2024 models. The front bumper camera doesn’t seem mandatory. My 2024 is unsupervised entirely for all sakes and purposes.
Hard agree. With v14.2.2.2, I only intervene to go through a drive through, or if I want to park somewhere specific. But otherwise I just let it drive and park itself and have zero safety concerns. Get in car push button turn brain off. Every drive.
Go try it. Walk in and test drive, or get a friend with HW4 to take you somewhere.
I think the reason behind Tesla not scaling robotaxi faster is that they are holding themselves to a much higher standard than Waymo. Waymo gets enough bad press for holding up intersections or passing stopped school busses. Public backlash might be extreme if it was Tesla, given how polarizing Elon is. And realistically there isn't much value gained by removing the safety passenger on their tiny fleet of operating robotaxis.
But I do believe Tesla's FSD model is ahead of Waymo's, despite having significantly fewer sensors and GPU load. Teslas simply don't get stuck in intersections or freeze up at traffic light outages.
And I cannot stress how perfect v14.2.2.2 is, such that there is basically no room for improvement OF ANY KIND in 98% of scenarios. I cannot remember the last time I had a safety-related intervention. Comments from passengers are like: "that is the most perfect ride I've ever had". It's smooth, efficient, and seamless defensive driving.
Robotaxi scaling is slow. We're not sure why. I think it's a combination of extreme caution and a need to build the infrastructure to efficiently support 10k+ robotaxis in each city. 10x safety improvement over humans may not be enough for Tesla to feel comfortable scaling. But hey if Waymo and Zoox want to deploy autonomous (with an unknown level of remote operator intervention) so they can technically say they are first and start drawing trend line through number of cars on the road, that's great optics. But I think Tesla is playing to win.
It really doesn't matter for the financials if Tesla has a hundred vehicles on the road operating fully autonomously. That just would just prove a point to outsiders, but inside Tesla they know all the data and have nothing to prove.
If you're pessimistic on Tesla, do you have a short position against them since they're currently 200-300 P/E ratio? 2026 should be extremely revealing. Either Tesla will scale robotaxi at an extremely high rate, or we'll bleed into 2027/8 and multiple players will have really solved FSD and robotaxi will enter a competitive low-margin market.
Maybe you think TSLA will stay irrationally high regardless of a robotaxi flop, but Tesla's main valuation is based on hyperscaling of robotaxi and massively displacing traditional vehicle ownership.
p.s. Elon is working for free unless he increases TSLA by about 50% from today's price (first tranche of the $1T package). Tesla the company has invested 10's of billions into FSD and robotaxi alone. The sheer size of the bet should impact your estimated odds of success. Or else you think it's all just a fraudulent money wasting exercise, in which case please short.
During covid I apparently got permanent verbal antibodies to the phrase "out of an abundance of caution" and can't avoid a reflexive allergic reaction to it of PTSD twitches even when, I confess, it is the right phrase to use in the context and being used correctly. Some scars never heal.
What do people think about sending unaccompanied minors in autonomous cars, such as to school or athletic practice? I see this as similar (and actually far safer) to walking to school, which I did by myself starting in 2nd grade (and the only reason I wasn't doing it by myself earlier is because I had older siblings at the same school). I think I would feel comfortable putting similar aged children in autonomous cars by themselves. It's a moot point where I live for now, but I'm interested in other people's perspective.
My oldest child is 5 and can walk to kindergarten/sports on his own.
For letting him in a autonomous car I think it depends on how much control there is for me. E.g the child shouldnt be able to tell the car to go to disneyland instead of school.
The other risk i see is him bumping another car when he opens the door.
How does speed limit enforcement work in Australia? Fines for going 1 kph above the posted speed? Monitoring everywhere? Or the opposite, minimal enforcement or no limits? A quick google search turns up headlines about Australia's speed enforcement getting stricter this year compared to last year, but I don't know what the baseline looked like.
There is some variability from state to state, but in general, speed is enforced through automated cameras that are either permanent installations or a detection car on the side of the road scanning every car that goes past. If caught going above the speed limit (10% over), a fine and demerit points will be automatically sent to the car owner (who can then nominate someone else as the driver).
If speeding in front of a patrol car, the police may also pull you over under the officer's discretion. (similar to the US?)
This results in the majority of drivers following signposted speed limits on streets and roads, though I will admit that on multilane highways, we do get some drivers going regularly 10+ km/h over the speed limit (maybe ~10%). But they are generally the exception to the flow of traffic rather than the majority.
In Switzerland there is some leeway for calibration, but generally when you are more than 5% above the speed limit you will get fined (starting at around 70 USD per instance).
As a Swiss policeman once said on some kind of international conference: if we wanted people to drive 53 km/h in cities (as opposed to 50), we would have written that on the signs.
Obviously not fully true (unlikely that the optimal speed limits are all nice multiples of 10), but I really prefer having the "real" rules written out. As opposed to eg Italy where a speed limit of "40 km/h" can mean anything between 50 and 90, depending on the context.
As an Australian living in France and spending a lot of time in Switzerland, Australia's speed limits seem to me to be insanely low. The highway limit should be 130 (80mph), at least!
>The real final boss for self-driving cars is the speed limit.
Is this true? Opus says the original 1964 study that describes this had serious flaws - such as, when they took out all the "slow" drivers that were actually just doing turns, suddenly it was only the drivers going faster than the speed limit that were in great danger. And of course kinetic energy = velocity squared. Follow up studies show a lot more nuance or negative evidence.
Would be feeling the Gell-Mann amnesia if this claim is false.
This whole post was Gell Mann amnesia moment for me. That claim was pretty egregious (it is pretty well established that increasing speed limits increases road fatalities). But also the claims about self driving progress don't address any of the bear arguments. There was a good substack post here: https://strangecosmos.substack.com/p/self-driving-cars-arent-nearly-a
There is a persistent confusion between highway speed limits and residential speed limits. Kinetic energy determines the *severity* of a collision, but the probability of a collision depends on *delta* v (variance).
In residential settings, delta v is approximately equal to car speed because pedestrians (and bikes, to a lesser extent) might as well be standing still.
On highways, everything is moving. Therefore, you want everything to move at the same speed, so delta v is 0. However, there are two populations of human drivers. One group will drive at the speed they feel comfortable maneuvering, which is around 80 mph for most people. The other group will drive the speed limit. Therefore, lower highway speed limits increase collision frequency. This has been observed repeatedly.
Because the bulk of the evidence I'm getting from Opus here https://claude.ai/share/243144b1-156e-42b8-aded-e7c67e307caa is that higher speed limits on highways do increase fatalities, and that there is no U-curve for injuries vs speed - only higher speeds?
Increased fatality rate doesn't necessarily mean increased collision rate.
The behavior of slower drivers is more influenced by the speed limit than the behavior of faster drivers, so raising the speed limit is a more effective way to decrease variance.
Split the difference: Waymo X, where you may or may not have other pickups along the way, in exchange for reduced fare. (I think that's the name anyway, haven't taken a shared rideshare in like 8 years.) The stronger coordination of AVs might make carpooling attractive again, especially when none of the participants actually have to drive anymore. And I'd imagine there's always a market for unusual riders like Roon who view the human interaction as Benefit Not Cost. It wouldn't work for truckers, but I can see a world where "human drivers" are maybe not mandatory, but a sort of premium service one could specify in an AV just to have someone to talk to. The job becomes significantly less fake if you're paid to play the social role of a good cabbie: information broker, confidant, tour guide, salt-of-the-earth philosopher. Honestly sounds a lot more fun than many forms of customer service work! Look at me, a man in motion - all I need is a pair of wheels...
Another positive implication of AV diffusion and scaling driving costs down + availability up: makes location-based markets such as housing more efficient. Like recently I was looking at a condo that would have been a nearly perfect fit...modern, within budget, public transit literally right out front, tons of amenities. Only problem is, the last bus home from work stops running before my shift ends...and ordering an AV at current prices, five days a week, would meaningfully eat away at the positive cashflow of no longer renting. (Cheaper with rideshares, but getting into a car alone with a random stranger night after night is its own form of cost...) Cutting that a lot closer to my current "lol, my subsidized fare is paid using pre-tax dollars via commuter benefits" makes many more of these but-for-transport transactions viable.
"The path forward for Waymo doesn’t look promising. Tesla has shown that scale doesn’t confer the benefits that many (including me) had hoped. Waymo may be doomed to burn capital forever, or else go the way of Cruise and so many other self-driving car startups. (Unless, of course, there is a surprise breakthrough. But we can’t count on that.)"
Having read in that piece about Waymo's "fleet response" department, I can no longer hear the words "self-driving" without thinking, "No, they're not!"
> The correct answer is to increase our speed limits across the board to the actual limit, beyond which we can and will ticket you.
I'm confused as to why you think this will work.
We are willing to blame a robot for going 5 mph over the limit. We aren't willing to blame a human for doing that. (Or at least, we're not willing to blame them *enough* to make the ticket worthwhile.) No matter what the limit is. The problem seems fundamental.
What might actually help here is to automate speed limit enforcement and ticketing more, down to 5mph (combined with relaxing speed limits by ~5mph to make up the difference).
"If enough cars on the road were self-driving, then they wouldn’t even need the traffic lights, they could coordinate in other ways, and this would all be moot."
Enough here being precisely 100%, so long as we ban cycling and walking.
To be clear: this is a dystopian future I detest.
With 100% perfect self-driving cars, you could just let people walk and cycle in the road without any restrictions!
There are pedestrian crosswalks already that have a traffic signal that only activates when a pedestrian presses the button.
Cyclists and pedestrians would use these and they would make all the sdcs stop for them. Or just dart out in traffic and most of the time they stop or dodge.
Now, the dystopian future that any rational AGI+ would advise would stick a credit card reader on the request crossing button. The pedestrian would have to pay the cost of delaying everyone else, internalizing an externality :)
"What they have done is made 'Robotaxi’ service go live in Austin for select rides, but these rides remain supervised with a Tesla employee in the driver’s seat."
They're in the passenger seat, which I think is a pretty significant difference.
They sit with their finger over an e-stop button (a hacked firmware build makes it the passenger door open button)
Mutes gripping armrests >> fashion that into progress :)
I routinely see Waymos on the freeway which are not obeying the speed limit. I'm sure about this, because I was surprised the first time it happened and made sure to pace them and check their speed: they were going about 70-71 in a 65 zone.
Just how good Tesla is IMO undersold in this post. It is very much a solved FSD product, minus the regulatory approval. There's no scenario where an HW4 Tesla fails at this point where a human wouldn't and I highly recommend buying one if you can afford to do so.
If Tesla did a better job marketing their newest FSD, I predict they'd quickly capture ~75% of the $50k+ car market in the US.
I agree so much with this, but there is slightly more than just policy. There’s the infrastructure of cleaning stations, charging stations with staff, programming in spots for them to idle while waiting for a ride, and checking parking lots and other areas for nuance little spots. There’s the rider call centers and everything.
It will never be limited by number of cars but that’s only one of many factors for the cab driver aspect.
But, yes, I use FSD every day and it’s a safer and better driver than anyone I know.
Private (non-Tesla) ownership of robotaxis takes care of most of these issues--the person owning the car (or fleet of cars) is responsible for charging, maintenance, etc.
Tesla doesn't publish meaningful safety statistics. In good driving scenarios, Autopilot/FSD is (probably) safer than human drivers, but since Autopilot/FSD can disengage in certain driving conditions, it really doesn't tell you much. I wish they would publish better stats.
Idk I’m in the same camp as Zvi in that “safety” is just a red herring and to some degree publishing statistics means making a concession to the Luddites.
Personally I think that “better than a P99 human driver” is good enough and I’m ~95% confident they’ve already reached this milestone.
Based on what? With no stats how can you be so confident?
Based on owning one for 9 months and reading tons of Tesla forum posts where previously skeptical owners did a complete 180 once FSD 13+ was released for HW4.
I trust my own experience more than any random stats.
I've been holding out on buying a new car for 10+ years waiting for full self driving. Tesla seems close now, but are we completely confident if I buy a new or recently used model Y it will be capable of unsupervised driving soon? Why not just wait a little longer?
You have to buy a car with HW4. Ideally buy the 2026 model because it adds a camera on the front bumper which should improve parking performance.
Theres never going to be a 100% perfect FSD, best you can have is a larger number of 9’s. The current Tesla is 99.9% reliable. The one in 2028 might be 99.99% reliable. Whether the inconvenience of not having a car is worth the wait is up to you.
I’d wait till the next one. Specifically, robotaxis should be feature complete when released. Or just wait until it comes out and see if it’s unsupervised. I wouldn’t buy until that’s going around.
Or you could easily buy a used Tesla HW4 vehicle, probably at half the price of new. I’d target 2024 models. The front bumper camera doesn’t seem mandatory. My 2024 is unsupervised entirely for all sakes and purposes.
Hard agree. With v14.2.2.2, I only intervene to go through a drive through, or if I want to park somewhere specific. But otherwise I just let it drive and park itself and have zero safety concerns. Get in car push button turn brain off. Every drive.
Go try it. Walk in and test drive, or get a friend with HW4 to take you somewhere.
I think the reason behind Tesla not scaling robotaxi faster is that they are holding themselves to a much higher standard than Waymo. Waymo gets enough bad press for holding up intersections or passing stopped school busses. Public backlash might be extreme if it was Tesla, given how polarizing Elon is. And realistically there isn't much value gained by removing the safety passenger on their tiny fleet of operating robotaxis.
But I do believe Tesla's FSD model is ahead of Waymo's, despite having significantly fewer sensors and GPU load. Teslas simply don't get stuck in intersections or freeze up at traffic light outages.
And I cannot stress how perfect v14.2.2.2 is, such that there is basically no room for improvement OF ANY KIND in 98% of scenarios. I cannot remember the last time I had a safety-related intervention. Comments from passengers are like: "that is the most perfect ride I've ever had". It's smooth, efficient, and seamless defensive driving.
Robotaxi scaling is slow. We're not sure why. I think it's a combination of extreme caution and a need to build the infrastructure to efficiently support 10k+ robotaxis in each city. 10x safety improvement over humans may not be enough for Tesla to feel comfortable scaling. But hey if Waymo and Zoox want to deploy autonomous (with an unknown level of remote operator intervention) so they can technically say they are first and start drawing trend line through number of cars on the road, that's great optics. But I think Tesla is playing to win.
It really doesn't matter for the financials if Tesla has a hundred vehicles on the road operating fully autonomously. That just would just prove a point to outsiders, but inside Tesla they know all the data and have nothing to prove.
If you're pessimistic on Tesla, do you have a short position against them since they're currently 200-300 P/E ratio? 2026 should be extremely revealing. Either Tesla will scale robotaxi at an extremely high rate, or we'll bleed into 2027/8 and multiple players will have really solved FSD and robotaxi will enter a competitive low-margin market.
Maybe you think TSLA will stay irrationally high regardless of a robotaxi flop, but Tesla's main valuation is based on hyperscaling of robotaxi and massively displacing traditional vehicle ownership.
p.s. Elon is working for free unless he increases TSLA by about 50% from today's price (first tranche of the $1T package). Tesla the company has invested 10's of billions into FSD and robotaxi alone. The sheer size of the bet should impact your estimated odds of success. Or else you think it's all just a fraudulent money wasting exercise, in which case please short.
> i strongly prefer uber to waymo.
I wonder if anyone has ever been sexually assaulted by a Waymo.
Rule 34.
During covid I apparently got permanent verbal antibodies to the phrase "out of an abundance of caution" and can't avoid a reflexive allergic reaction to it of PTSD twitches even when, I confess, it is the right phrase to use in the context and being used correctly. Some scars never heal.
What do people think about sending unaccompanied minors in autonomous cars, such as to school or athletic practice? I see this as similar (and actually far safer) to walking to school, which I did by myself starting in 2nd grade (and the only reason I wasn't doing it by myself earlier is because I had older siblings at the same school). I think I would feel comfortable putting similar aged children in autonomous cars by themselves. It's a moot point where I live for now, but I'm interested in other people's perspective.
My oldest child is 5 and can walk to kindergarten/sports on his own.
For letting him in a autonomous car I think it depends on how much control there is for me. E.g the child shouldnt be able to tell the car to go to disneyland instead of school.
The other risk i see is him bumping another car when he opens the door.
If those get solved i dont see a problem.
Yes, I would prefer that I have a lot of control/tracking, ideally including camera views from inside the car.
Podcast episode for this post:
https://open.substack.com/pub/dwatvpodcast/p/advancements-in-self-driving-cars
"We are going to need a regulatory solution, ideally that allows you to drive at the average observed speed."
As an Australian, America's approach to speed limit enforcement seems insane.
How does speed limit enforcement work in Australia? Fines for going 1 kph above the posted speed? Monitoring everywhere? Or the opposite, minimal enforcement or no limits? A quick google search turns up headlines about Australia's speed enforcement getting stricter this year compared to last year, but I don't know what the baseline looked like.
There is some variability from state to state, but in general, speed is enforced through automated cameras that are either permanent installations or a detection car on the side of the road scanning every car that goes past. If caught going above the speed limit (10% over), a fine and demerit points will be automatically sent to the car owner (who can then nominate someone else as the driver).
If speeding in front of a patrol car, the police may also pull you over under the officer's discretion. (similar to the US?)
This results in the majority of drivers following signposted speed limits on streets and roads, though I will admit that on multilane highways, we do get some drivers going regularly 10+ km/h over the speed limit (maybe ~10%). But they are generally the exception to the flow of traffic rather than the majority.
Not Australian but Swiss here.
In Switzerland there is some leeway for calibration, but generally when you are more than 5% above the speed limit you will get fined (starting at around 70 USD per instance).
As a Swiss policeman once said on some kind of international conference: if we wanted people to drive 53 km/h in cities (as opposed to 50), we would have written that on the signs.
Obviously not fully true (unlikely that the optimal speed limits are all nice multiples of 10), but I really prefer having the "real" rules written out. As opposed to eg Italy where a speed limit of "40 km/h" can mean anything between 50 and 90, depending on the context.
As an Australian living in France and spending a lot of time in Switzerland, Australia's speed limits seem to me to be insanely low. The highway limit should be 130 (80mph), at least!
>The real final boss for self-driving cars is the speed limit.
Is this true? Opus says the original 1964 study that describes this had serious flaws - such as, when they took out all the "slow" drivers that were actually just doing turns, suddenly it was only the drivers going faster than the speed limit that were in great danger. And of course kinetic energy = velocity squared. Follow up studies show a lot more nuance or negative evidence.
Would be feeling the Gell-Mann amnesia if this claim is false.
This whole post was Gell Mann amnesia moment for me. That claim was pretty egregious (it is pretty well established that increasing speed limits increases road fatalities). But also the claims about self driving progress don't address any of the bear arguments. There was a good substack post here: https://strangecosmos.substack.com/p/self-driving-cars-arent-nearly-a
There is a persistent confusion between highway speed limits and residential speed limits. Kinetic energy determines the *severity* of a collision, but the probability of a collision depends on *delta* v (variance).
In residential settings, delta v is approximately equal to car speed because pedestrians (and bikes, to a lesser extent) might as well be standing still.
On highways, everything is moving. Therefore, you want everything to move at the same speed, so delta v is 0. However, there are two populations of human drivers. One group will drive at the speed they feel comfortable maneuvering, which is around 80 mph for most people. The other group will drive the speed limit. Therefore, lower highway speed limits increase collision frequency. This has been observed repeatedly.
"Observed repeatedly" - are you talking about this study? https://www.psu.edu/news/research/story/crashes-increase-when-speed-limits-dip-far-below-engineering-recommendation
Because the bulk of the evidence I'm getting from Opus here https://claude.ai/share/243144b1-156e-42b8-aded-e7c67e307caa is that higher speed limits on highways do increase fatalities, and that there is no U-curve for injuries vs speed - only higher speeds?
Increased fatality rate doesn't necessarily mean increased collision rate.
The behavior of slower drivers is more influenced by the speed limit than the behavior of faster drivers, so raising the speed limit is a more effective way to decrease variance.
Specific studies:
https://www.ibiblio.org/rdu/sl-irrel.html
https://web.archive.org/web/20240709233222/https://minds.wisconsin.edu/bitstream/handle/1793/75382/MS_Thesis_Brose_Kirsten.pdf?sequence=1&isAllowed=y
Pages summarizing results:
https://www.michigan.gov/en/msp/divisions/spec-ops/inside-sod/traffic-crash-reconstruction-unit/traffic-control-orders
https://driving.ca/column/motor-mouth/lower-speed-limits-safer-roads-aaa-study?utm_source=aroundthenetwork&_ga=2.15710707.493274395.1690031067-1624059339.1689853297#038;utm_medium=onnetwork&%23038;utm_campaign=on_network_boosting&%23038;utm_content=driving
Glad to know I can still compete with the AIs...
Thank you!
Is Waymo profitable? I have heard from a Tesla fan boy that they are losing hundreds on each ride.
Split the difference: Waymo X, where you may or may not have other pickups along the way, in exchange for reduced fare. (I think that's the name anyway, haven't taken a shared rideshare in like 8 years.) The stronger coordination of AVs might make carpooling attractive again, especially when none of the participants actually have to drive anymore. And I'd imagine there's always a market for unusual riders like Roon who view the human interaction as Benefit Not Cost. It wouldn't work for truckers, but I can see a world where "human drivers" are maybe not mandatory, but a sort of premium service one could specify in an AV just to have someone to talk to. The job becomes significantly less fake if you're paid to play the social role of a good cabbie: information broker, confidant, tour guide, salt-of-the-earth philosopher. Honestly sounds a lot more fun than many forms of customer service work! Look at me, a man in motion - all I need is a pair of wheels...
Another positive implication of AV diffusion and scaling driving costs down + availability up: makes location-based markets such as housing more efficient. Like recently I was looking at a condo that would have been a nearly perfect fit...modern, within budget, public transit literally right out front, tons of amenities. Only problem is, the last bus home from work stops running before my shift ends...and ordering an AV at current prices, five days a week, would meaningfully eat away at the positive cashflow of no longer renting. (Cheaper with rideshares, but getting into a car alone with a random stranger night after night is its own form of cost...) Cutting that a lot closer to my current "lol, my subsidized fare is paid using pre-tax dollars via commuter benefits" makes many more of these but-for-transport transactions viable.
Was hoping you would cover this post that was linked on MR recently: https://open.substack.com/pub/strangecosmos/p/self-driving-cars-arent-nearly-a?utm_source=share&utm_medium=android&shareImageVariant=overlay&r=ghym0
"The path forward for Waymo doesn’t look promising. Tesla has shown that scale doesn’t confer the benefits that many (including me) had hoped. Waymo may be doomed to burn capital forever, or else go the way of Cruise and so many other self-driving car startups. (Unless, of course, there is a surprise breakthrough. But we can’t count on that.)"
Having read in that piece about Waymo's "fleet response" department, I can no longer hear the words "self-driving" without thinking, "No, they're not!"
> The correct answer is to increase our speed limits across the board to the actual limit, beyond which we can and will ticket you.
I'm confused as to why you think this will work.
We are willing to blame a robot for going 5 mph over the limit. We aren't willing to blame a human for doing that. (Or at least, we're not willing to blame them *enough* to make the ticket worthwhile.) No matter what the limit is. The problem seems fundamental.
What might actually help here is to automate speed limit enforcement and ticketing more, down to 5mph (combined with relaxing speed limits by ~5mph to make up the difference).
I agree, but this seems even more unpopular than self driving cars.
Would cost trillions for Waymo to replace even a third of cars. Need to license their software
Thanks for linking my newsletter!