Dashcam footage shows driverless cars clogging San Francisco
wired.com
wired.com
People would commute up to 4 hours one way, napping in their cars, maybe even longer. Delivery vehicles running all hours of the night, perhaps with no one ever in them at all. People sending the cars to go pick things up or people that they otherwise wouldn't have time in their day to do.
Even if you don't want to live in your car that the pressure will still be exerted on you because you'll be competing with people that do. Cars instructed to circle the block in areas with no parking clogging the streets. You'll need a self driving car yourself to even get a spot since you'll never defeat all the robot vultures.
People will forget how to drive entirely of course so there will be no going back. The whole thing will feed on itself, people will need more self driving cars to get back the time stolen by other self driving cars.
In your hypothetical future, where no humans manually drive anymore, it opens up the possibility of completely redesigning our vehicle infrastructure without having to worry about the training and compliance of a mixed human/AI population on the road.
My optimistic hope is that with fully automated vehicles, mass transit wins out on sheer efficiency in order to negate the wasteful deadheading inherent in personal vehicles.
How does a self driving car deal with a blizzard with white-out conditions or a complete loss of traction due to ice? Yes, humans get into a lot of accidents in these conditions but as a percentage of the time the rate is still remarkably low.
If you want to eliminate large classes of those adverse situations then the best approach we have right now is to switch to rail transport. Public transit is truly the way to go!
Excommunicating humans from all public space and drawing a moat of monopoly owned robo taxis around every block and every destination isn't a utopia, it's a complete nightmare.
In short, what took 25 minutes door to door on any given day before ended up taking ~two hours on the way to work. And for reasons I can no longer remember, getting back home ended up taking twenty minutes longer.
In other words, a total commute time of ~50 minutes (max) was turned into a ~4.5 hour commute due to having to rely on public transportation.
I also remember the specific reason why I went and bought a new car that same weekend (fortunately this happened on a Friday): it was the realization, upon arriving at home (after said ~2.5 hour commute), that I had forgotten my apartment keys at work.
It's pretty neat, I've moved to a European city recently myself and I'm slightly horrified at the prospect of moving back.
The eternal pessimist within me looks at how difficult it is to get anything with infrastructure accomplished, both at a reasonable pace and reasonable cost, in the US and can't help but expect this sort of thing to not play out as wonderfully as that sounds.
We are nowhere near a future where cars exist in hermetically sealed tubes, completely walled off from the rest of existence.
Why do you think this would happen? Self-driving cars are poised to dramatically improve parking overall. One can be dropped off exactly at their destination, and the car can then go elsewhere. Even an elsewhere of "somewhere 5 minutes away" is a large amount of potential parking in a typical urban environment. Additionally, any existing parking lot that chooses to only accept self-driving vehicles instantly gets additional capacity: self-driving cars could cooperatively pack/unpack themselves into space efficient configurations.
And if your counter is "a minority of selfish entitled people", the answer to that is "laws and regulations". Self-driving cars will have to follow the laws of wherever they're driving. If your "circling the block" apocalypse comes to pass, the mitigation would literally be a law/regulation that self-driving cars would have to obey.
Cars are not going anywhere.
Imagine an inner city without cars. Cities in which this is a reality are a bliss to be in. And people get in and out just fine.
But also, while I'd love (love!) to see much better, European-style public transit in US cities, it would take a fantastic level of political will, lobbying, and pitching to even begin to happen. For all sorts of reasons, public transit is wildly unpopular in the US, and construction is wildly expensive.
But a company like Tesla could (if they could figure out the FSD part) just roll out self-driving taxis after a few legal tweaks. It's achievable in a reasonable timeframe (once the tech is in place).
Seems Americans are just too used to staring at a 2 ton steel can in front of them.
Reinvented, but much better. It's the difference between Circuit-Switched vs. Packet-Switched in networking. Trains are extremely inefficient use of resources. They require a dedicated and very expensive path, cannot re-route around problems (e.g. broken tracks or obstructed tracks), and frequently travel at a fraction of their capacity. Properly sized and routed electric vehicles are much more efficient in terms of cost per mile, travel time, and almost any other metric.
> frequently travel at a fraction of their capacity.
The average occupancy of a car with 5 seats is only marginally above 1. So, 20-25%? The average occupancy of a European streetcar is far above that.
> Properly sized and routed electric vehicles are much more efficient in terms of cost per mile, travel time, and almost any other metric.
Just as much as you can optimize those factors in electric cars can you optimize them in trains/trams/buses/...
It was already starting to happen in NYC with Uber's exponential growth before they capped the number of cars permitted. They were a year away from carmageddon.
And interestingly, all of the parking lots in a city, in theory, will be obsolete, opening a whole bunch of opportunities for new urban design. The idea behind self-driving cars is not to own the car, but to rent it...which is why some people are super bullish in the long term on Uber.
The cutover to autonomous cars will be a huge PITA however, and I think this is really what the OP is getting at.
I've had this same thought, but haven't seen it discussed much.
Parking lots essentially switch from random access (which requires expensive unoccupied 'lanes') into a slowly snaking space-filling curve. Cars can crawl slowly from entrance to exit, moving like water through a pipe. The paths can even branch and merge to make efficient use of the entire surface area.
If you also want to charge the cars stored in the parking lot, your solutions fall into two broad categories:
1.) Arrange the parking lot with rows of chargers that automatically plug in. Cars choose the appropriate row based on state-of-charge, twith he algorithm making sure the entire row finishing charging at roughly the same time. This minimizes unnecessary disconnect/move forward/reconnect cycles. Or,
2.) put wireless charging under the parking lot. This would be a lot more feasible than electrifying large swaths of highway.
Personally I still prefer option 1, because the peak charging speeds can be much greater.
Why would they? You still need to get the car to unpark itself when the owner needs it?
Oh, you mean everybody shares cars with everybody? What about letting sections of that "snake" just drive around, stopping at different spots and let people enter and leave? And perhaps let them walk around inside? Congratulations, you've just re-invented public transit.
This is unlikely to ever happen. Even in the human-driver world, there would be huge potential to have cars communicate directly with each other. Never materialized.
If the concept of a remote accident is so appalling to you...just don't get a self-driving vehicle. But other car users are capable of processing statistical risk non-emotionally, and they will get to enjoy stress-free parking once these vehicles become available.
Anyway, in all likelihood, self-driving cars will result in a large pivot away from car ownership. The impact of "your car" being involved in a remote accident becomes moot if it's not even your car.
do you:
- send the car home
no, it would only get halfway home before the end of your appointment
- send it on a rideshare journey
let's say (conservatively, and according to some minimum regulations from the rideshare service) it gets assigned a job after 5 minutes to drive somebody 10 minutes away. it finishes its job at the same time as your appointment ends. you have to wait 10 minutes for it to return to you. again no.
- send it to a parking lot
assuming the parking lot is 5 minutes away and you can tell it to return 5m before your appointment ends this seems reasonable
- drive circles around your appointment
all the benefits of sending it to the parking lot but requires even less effort as you no longer have to remember to tell it your appointment is finishing.
now apply this logic to the general public. what is the most likely outcome?
I believe this scenario was discussed as an argument for congestion pricing, serving as a vital solution to the tragedy of the commons exacerbated by self-driving cars.
I think this would ultimately be for the best. If the streets clogged up so severely with traffic then the value of owning a car would drop precipitously. Even if car owners lobbied successfully to widen all the streets the traffic would just expand to fill the available capacity.
People would finally be forced to seek alternatives!
It’s more likely that roads will be widened and traffic will grow to meet that supply than it is that cars start to go away. Cars only go away when governments announce near immediate bans.
8 hours (4+4) spent in commuting, driving or not, is an entire life. This isn't a realistic concern, but something pulled out of a cyberpunk novel.
Cars can park away from you and come back to pick you up. This should alleviate parking problems, and reduce parking-space-search based congestion on the street.
Learning to drive was never that hard.
You poor innocent child. Spent a couple of years of my life doing pretty much that, in crowded trains with a thousand others doing the same. Some did it for many, many years. It's a thing. (OK, I was lucky about where I worked and only had a 3-hr commute each way.)
Related data point... my commuting choices were
a) drive all or most of the way (saving 45 minutes to an hour each way);
b) drive 15 minutes to a park-and-ride, take a commuter bus;
c) drive 45 minutes to train station, take train in (extra 30 to 45 minutes each way, exta $$/mo).
Often did (b) for practical reasons, much preferred (c) for comfort. Didn't do (a) too often. (d) "black car" (private driver) was an idle astronomically expensive dream. But if it was only a relatively small premium over normal car ownership? I and 100's of thousands of my fellow commuters would have given our left kidneys.
I drove 90 minutes each way in/ around Atlanta and that was considered long; I had friends in NYC who were on the train 90 minutes each way and that was considered long.
I can't think of anyone I knew who had 2 hours each way, every day. (Some days, yes, but not as the norm.)
I do know folks who drive/ ride/ fly and then stay locally for some days, then commute back for the weekend or their WFH days. And most are for 2-3 hour driving commutes, not 4.
I agree with them 8 hours a day of travel "isn't a realistic concern".
I know this sounds nice for just a moment, but if you consider it longer, it's awful. It won't decrease congestion, it will increase it. Why?
1. Cars will still be searching for parking: many people will prefer to have their car nearby. It's not as if every car will altruistically drive to a far-off area. Since self driving cars have more patience than humans, they'll be on the road much longer searching for their parking. It may be more efficient to keep the car circling the block, which is even worse.
2. Parking availability is a major inhibitor of car trips, especially in cities. This is a good thing. If it becomes easier to "park" (ie leave your car driving on the street or send it away), that will induce more car trips, and more car trips means more congestion, until there's a new equilibrium (maybe a 7-minute parking search is eliminated, but it's replaced with 7+ minutes of traffic).
3. Pick-up and drop-off in cities is already difficult with rideshare services. If all personal vehicles are doing it as well, they'll definitely congest the curb lanes more. This is definitely a more solvable problem than the first two, but still an annoyance.
And for what it's worth, the 4-hour commute may be a bit far fetched, but it's hard to deny that many wouldn't mind an extra 20 minutes of commute time in traffic if they can relax, nap, read, watch a show, etc. People will chose that option more, adding more trips and more congestion, until an equilibrium is reached. Maybe it won't be 4-hour commute times but it will be a major increase and added congestion.
All of these extra miles traveled searching for parking, and adding extra congestion, are disastrous to cities and neighborhoods. Sure, the fossil fuel emissions alone would be awful, but suppose (charitably) that all autonomous vehicles are electric, and assume that their electricity generation is emission free (unrealistic for decades). The weight of EV batteries will dramatically increase road wear and tear, and they'll increase the pollution due to rubber tires, which are already the major source of microplastic pollution. And of course, it's a dramatic waste of energy from the power grid. And all of this is ignoring that dedicating that much road space storing to idling and parked vehicles is a no-good, terrible, awful way to utilize public space in a city or neighborhood, when it could be used by some efficient public transit, parks, and safer infrastructure for personal vehicles when necessary.
You'd then need to price the time a vehicle spends on the road without an occupant. Otherwise you're just pushing people to send their cars around the block for an hour to avoid paying market rate for parking.
But it's hard to enforce the "unoccupied vehicle" rule, so it would be much better to just charge for all road use time in high congestion areas. Maybe... congestion pricing?
You're not thinking with an open mind.
If I own a robot car and I know I'm going to be at dinner for 2 hours, I will send the robot car to do a few Uber rides while I'm eating. Or it will go and charge. It doesn't have to just drive around in circles like a human. It can coordinate will all the other robot cars in the area and pick a place where it won't create traffic, or they can distribute traffic among each other. Maybe it can park-share, where if there are 3 cars and 2 parking spots, the 3 cars can coordinate and rotate who goes driving around.
Maybe robot cars will be fungible, so you don't own a specific robot car, but you own a time share, so all the cars are basically like Ubers, and you can call the one closest to you.
The possibilities are endless, don't think that a robot car will just emulate a human.
If all cars are Ubers I’ll just move to some other place that isn’t the ninth circle of rent seeking, sharing economy, dot com nightmare dystopia.
You're right that I did predicate that little rant on the idea that the majority of AVs would be personally owned and not part of a fleet. I'm sure that car share will come into play to some degree, but I do think it's tough to convince people who are used to their car being a personal, (relatively) private space that they can store nearly for free on public roads, to give that up. It's especially hard to convince the automobile industry that the incredibly profitable 1-2 car per household model should be pushed aside in order to manufacture fewer, shared vehicles. Cruise (i.e. GM) will not cannibalize their personal car sale business: they're using this as a way to get test data to build personal AVs. Maybe car share will increase over time, but we're not about to witness some revolution, especially if it reduces consumption or profit.
No one wants their $60k+ self driving car casually risking accidents. And I certainly don't want to end my date night with drunk/messy Uber patron roulette for like $10 profit. Not hating if you do.. just not for me.
I mean assuming I don’t want to leave my sunglasses in the car, or anything else.
Of course, it still costs money and if everyone does that now maybe it's 2+ hours in a car each way and I still won't mostly do it.
I'd also routinely take the car into the city for a work event of some sort rather than dealing with non-trivial hassle of multi-modal driving to the train station and 2 different forms of public transit.
I've really cut down on casually swinging into town to meet someone or do an activity on a weeknight because it's just a hassle to drive with all the traffic and driving back home when I'm probably getting a bit tired.
So it is a combination.
There are already people who live in their vans anyway, if they didn't even need to drive them anymore perhaps areas with overzealous zoning laws that circumvent supply & demand would see an influx of small moving apartments instead.
It would be like a land yacht, and economically sound so long as the price of gas is lower than the price of rent.
But you'll still want to wear a seatbelt because although self driving cars don't tend to cause accidents, they very much are the victim of them.
And consider the electricity demand of keeping your big skooli heated and cooled compared to that of a small and tight modern vehicle, you'll likely be forced into charging last thing every night and first thing every morning while on the road, while for us it is about 25% of battery on the worst nights so we can camp anywhere.
What if that time is spent sleeping? Maybe you don't even need a house, so you'll have more money to spend on a nice car.
It's really uncanny, every time, there it is, man must submit to the technology rather than have technology submit to us. Because a self-driving car is advanced technology, therefore men must move with self-driving cars. Even if that literally includes having to live out of your car.
this kind of commute daily no less seems to be more of a stubborn personal choice than any economics forcing it .
All models are wrong, but some keep your hopes closer
I had a job with an effective ~4-hour round trip commute, from one major city to another in both my car and a train, but only had to do it once a month.
With self driving cars, it will be like a NYC taxi, the goal is 24 hour operations.
So for those cars, the time on the road per user doesn't change much, but the need for parking greatly decreases.
To paraphrase a certain fictional pirate captain, "You'd best start believing in cyberpunk dystopias... you're in one!"
Listen, and understand! That relentless driving machine is out there! It can't be bargained with. It can't be reasoned with. It doesn't feel pity, or remorse, or fear. And it absolutely will not stop... ever, until you are dead!
FTFY
Right now, SDC is operating in the real world, in non-trivial environments (San Francisco), without special road infrastructure to make them work. It’s beautifully backwards compatible, at the cost of not being generalized (the service areas are extensively mapped).
Once SDC take off we’ll likely start getting infrastructure and rules to support them. Think standards for communicating position locally - ie car A broadcasts its position and route to cars B, C, D within 200m, special road infrastructure to make lanes and corners more manageable for SDC, rules against aimless circling. There’s already a carrying cost in the form of gas or electricity plus wear incentivizing aimless driving, also the opportunity cost of not actively moving someone or something, but we can probably introduce some kind of toll or tax on a SDC operating with no humans inside it to further disincentivize this.
A very useful thing about SDC, and something I think people forget about rideshare and taxis, is that they let people move around independently without needing parking for those trips. In dense cities like SF and NYC that’s hugely useful. A single rideshare or SDC can move 10 people on custom routes without any of those people needing to find and pay for parking, and without using any parking infrastructure. That’s great because it disincentivizes wasting more space on parking in aggregate. Over time this should let us build denser.
Of course, public transit could obviate all these concerns, and I’m a big believer in funding way more public transit than we do already, but it will take a lot of time and political will to make that happen in the US. And it still does not offer the flexibility of SDC and rideshare. SDC is fully compatible with existing infrastructure and may give us a way to morph into public transit more smoothly with things like dynamically routed SDBusses and a reduction of parking infrastructure leading to denser urban environments that more easily support public transit. I think we can solve the “spending too much time on the road doing nothing” problem with congestion pricing, which we should really be doing already.
https://s3-ap-southeast-2.amazonaws.com/ehq-production-austr...
That picture has the cars bumper to bumper, which is likely rather better than even the self-driving best case. Cars take up a lot of space.
I can imagine interesting and efficient ways to move large numbers of people that involve self-driving vehicles for part of the journey, but for high capacity at reasonable cost and space utilization, feet, bicycles, busses, trains, subways, etc are dramatically better.
And stop calling for-hire cars "rideshare", because it isn't sharing, it's hiring or renting.
I believe it's quite common. In about 50 y this place will look very different.
Which may be.. but backwards compatibility doesn't absolve you of actually doing maintenance on your legacy systems. Unless the upgrade is free.
Road Pricing
When a scare resource is given away for free, it becomes overused. Somehow we have a hard time seeing that for road access, but it's true anyway.
It's very unlikely that it happens to be the optimal price.
Since they'll autopark or circle anyway, their size no longer becomes a limiting factor for most people.
There won't be any human drivers, so they will all yield to me even if I treat them like slalom poles.
This thought experiment sounds like a performance art dream. Do you remember the artist who created a Google maps traffic jam in London by filling a wagon with mobile phones and walking slowly? Ref: https://www.simonweckert.com/googlemapshacks.html
Another thing: You can sabotage self-driving cars by putting traffic cones everywhere. Or sit in the middle of a major intersect with ten of your friends and read the newspaper (or pick your nose!). The passengers will be furious, and the self-driving cars won't know what to do.
Well assuming no minimum stopping distance per vehicle momentum. Might want to be sure about that first lol
Listen, and understand. That car is out there. It can’t be bargained with. It can’t be reasoned with. It doesn’t feel pity, or remorse, or fear. And it absolutely will not stop, ever, until you ̶a̶r̶e̶ ̶d̶e̶a̶d̶ reach your destination.
Personally, I'd be down for a driverless Uber-like service.
If it involves the car having a data link to someone else's server, then I have no interest whatsoever in self-driving vehicles.
Until they don't, because of prohibitive insurance premiums or some accelerated phase-out of HDC bill. That's the blessing and the curse of the technology market: you don't get to keep your old toys if most of the market has moved on.
Yes, you're probably right. So what? I wasn't even remotely asserting otherwise.
I wonder if we would see as many idle/empty self-driving cars as we do idle/empty human-driven cabs and Ubers/etc. I could see a self-driving system being a lot more efficient because it doesn't have to keep a car on the road for a driver's entire shift.
Actually, yes. However, my take is a bit different from yours.
The future driverless car will not at all look like a Tesla.
It will be a moving office or comfortable room for family trips. This does not mean it has to be massive. Something of the size of a modern minivan could be fantastic if you did not have to worry about driving at all.
In other words, I actually see future driverless cars giving us more time rather than taking it away by being stuck on traffic paying attention to stay in a lane, press the brake or accelerator.
As for people forgetting how to drive. Well, maybe in one or two more generations?
Nobody really likes driving. People either need to go from point A to point B or want to (vacations, eating out, etc.). I enjoy driving, not on the street, on the race track. I do that with some frequency. Driving on the street is boring, dangerous, time consuming and stressful. I hate that kind of driving. I can't wait until I don't have to do it any more.
The scenario you outline was part of Mega-City One's landscape in the 2000AD/Judge Dredd comics: cars on the road all day every day, with people living out of their perpetually moving RVs; something similar was depicted for Termight in the Nemesis the Warlock series.
After giving control over to an AI, suddenly I don't care at all about saving 3 minutes by picking the perfect lane, getting the best spot on every merge, maxing out the speed limit threshold where the local police will pull me over.
That kind of gamification with no tangible reward actually generates the anxiety. When you're no longer the player, there's no more anxiety about those weird time optimizations.
If streets are "clogged" it saves me a few pennies and gives me less reason to worry about a human cutting across to risk my life for no reason. It's actually kinda relaxing.
But you are wrong about one thing: AI have a fear of death. Every aspect of their training is hyper focused on a paranoid level of death avoidance.
When I picked up my father from the airport last month, I did your worst nightmare, my car circled the terminal for about an hour while I jammed out to progressive rock. 90% of the time I was in bumper to bumper on a closed access road with no entrance or exit. From a traffic perspective it was irrelevant whether I was there or not. On the 10% when I reached the terminal on the loop, surprise, I was blocked by loading cars in front of me... The same blockage the car behind me would face.
What will happen is:
Companies will buy fleets of self driving cars.
The best self driving cars will be too expensive for most people to buy.
People will just pay one of those companies for transportation services.
The transportation - serving companies who will use the cars in the most efficient way will survive. They will use AI to optimize the usage of the fleets.
Regulators (governments) will tax the companies per km / time interval / peak hours.
People will pay more if they want to travel alone or want to travel in a straight line. They will pay more if they want more exact timing.
Roads will be safer because they will communicate with the cars and the cars will communicate with other cars.
There will be manned vehicles that will help relieve any obstructions.
In other words, given proper tax rules, transportation will regulate itself in effective and efficient ways.
This is not necessarily a given; after all, most Americans have other little-utilized equipment that nobody bothers to try to "rent out" to maximize usage. Think - washing machines, dish washers, lawn mowers, snow blowers, etc. In fact, there's an entire industry built around storing little-used equipment (storage units).
And we have high-usage vehicles transport already, they're called taxis and they're at use in most major cities; the AI cannot save MORE than the cost of the driver without some magical accounting.
I don't really follow the argument. I don't think of my car as an "investment." In fact quite the opposite, I think of it as an extremely depreciatory purchase that is partly a necessary cost of living (commuting for work, shopping) and partly for leisure (visiting friends and family, going out on the town, road trips). It makes very little sense to me to "want to be on the road as much as possible," especially considering that most of the depreciation for when I eventually sell it or trade it in is likely going to be based on the odometer rather than the calendar.
This means you'll end up not owning a car at all. Like with housing, those who look at it as investment game outcompete and drive up prices for those who just want a place to live in and call their own.
And if you mix together "modern American government" and "billions of dollars at stake for mega-corporations" - just how close to zero is the real-world probability of those "proper tax rules" being made and maintained?
That's a legitimate question I haven't done the research on this. It would seem though that usually it's a municipal employee, probably a union job, so probably paid pretty well (relative to say an uber driver). Also that cost for the driver would be double if you half the size of the bus and run them twice as much which would be a better experience for passengers. It would seem like the cost of drivers could be a real impact but this is me being handwavy I haven't crunched any numbers.
On what are you basing this?
I grew up in NYC which had the most frequent buses I've seen anywhere in the US, and 10+ minutes on a route was common, far far more on off hours. Plus, that route would only get me so far, so I'd have to change buses.
My ~2 mile trip from my apartment to my elementary school required at least two buses, often three.
> a substantial amount of people aren't going to pay $20 to get from A to B when you can get there slightly slower for $2.
NYC MTA trip fare is $2.75. Atlanta is $2.50.
Are you Chuck Wollery? How are you doing 2 and 2?
Comparing to a current system limited in service by the cost of the driver to a future system that won't have this meaningless.
And I've got news for you. Your $2.75 ticket wouldn't be $2.75 without a driver.
Additionally, the vast majority of transit rides are covered by a monthly pass, and the average trip price is less than $2.75.
They won’t park, they’ll just move from one job to the next (like taxis and ubers).
I think the possibility is more around SDCs becoming not simply a business to individual customer service or subscription, but rather SDCs being offered by organizations the individual customer is already a member of as part of that organizational offering. This can begin at large and wealthy retirement communities first (like Sun City Grand in Phoenix), but there's no reason why it couldn't be implemented elsewhere or why an organization that is already setup to handle customer and vehicle relationships can't offer it as well. Think AARP or your auto insurance company.
I used to write them off as over the top worries of uninformed people.
More recently as Elon Musk’s failure to deliver has been made clear, I’ve realized that I was myself taken in by a lot of hype.
Now when I go back and look at some of the cyclical content around self driving cars that has been around for years, I realize that a lot of it makes sense. The most damning criticism is that automobile-focused infrastructure sucks, and self driving cars won’t change that. We need to invest in traditional forms of transit. But in a world with cheap money (the last decade) and winner-take-all mentality, we get huge investment in individualized technology like self driving cars instead of better urban design and transport infrastructure.
This is a disease of the soul caused by reading Freakanomics which makes you think everything that's counterintuitive must actually be true.
Sure!
An entire class of accidents and injuries will be entirely gone, saving countless lives. Insurance will get cheaper for everyone. Cities will be able to adapt and grow more quickly, since traffic management can finally happen in realtime and at scale. Pollution will plummet.
Cars can get smaller and smaller, focusing only on having a safe space for passengers and eliminating all the extra "driver" nonsense. With so many cars on the road, you'd always be able to catch a ride, anywhere, anytime.
Each car could have a delicious sandwich compartment, so that it's always doing double-duty for carrying both passengers and sandwiches. (Every car would have a couple drones it could dispatch to pick up and deliver sandwiches, so that passenger trips can go on uninterrupted.)
People will forget how to drive entirely, of course, so there will be no going back. The whole thing will feed on itself, and we'll enter a golden era of cheap transportation and ever-faster sandwich delivery.
I for one welcome our autonomous sandwich delivery overlords.
Joking aside I'm more curious on the negative aspect of self driving vehicles as I'm not entirely convinced they will explode into the market quite the way it's been pitched.
In the long run however you can't ignore the productivity increases and the cultural changes that will come with it being a net positive.
Governance and ownership is the sticky point for me.
* Cars are all electric, since the one downside major downside (a long time to charge versus gas) is reduced if they can go charge themselves. So less pollution and noise.
* The number of cars may increase, but so will the capacity of roads due to faster speeds and higher density. So a city could possibly only have a couple of car accessible roads.
* Parking can also get denser, as cars can presumably park themselves and coordinate. So you could build a few massive parking garages (for example) to serve thousands of cars, or require them to park outside of the city.
Tire noise dominates at speeds above roughly 30 kph (20 mph), so on reasonably fast roads not really.
https://www.researchgate.net/figure/Propulsion-noise-the-tyr...
Of course, self driving cars might also mean slower cars, because the drivers are less impatient, and if that is true then noise would go down. On the flip side, self driving cars might end up meaning faster cars because they can coordinate better, increasing noise levels.
As someone who enjoys walking and biking, this strikes me a serious negative, not a positive.
(Group parking spots are typically better monitored than streets.)
None of this is coming soon as far as I'm aware mind you. I just don't think the end game of self driving vehicle is bleak - just the long mid game :)
Why? Attrition on a car is mainly miles based. Sharing a car won't make it last more miles. Not accounting for vandalism.
But even today, even in north America, they're barely rational. Most people don't mentally add up the purchase price plus insurance plus gas plus maintenance etc. And the inconvenience of garage and parking and clearing driveway and snow etc etc.
The reason we have them is convenience. Public transit in North America with our huge suburbs is on average crap.
I firmly believe that for modern generation, car is increasingly a liability not pleasure. So if we / they can exchange it for a self riding car that magically shows up when needed but also magically buggers off when not, we / they will.
On the other hand, ride sharing could mean less cars overall and less car ownership. Electric cars mean fewer or no emissions. Totally automated traffic management could enable vastly higher throughput, and otherwise congested roads could have dynamic tolls applied so people won't use them unless they really need to. Delivery vehicles can batch shopping pickups and dropoffs to use less vehicles than currently.
If problems arise, they can be fixed through policy and incentives. Otherwise, traditional economic incentives for efficiency continue to apply.
https://www.theatlantic.com/science/archive/2020/11/car-tire...
But... there's another dimension here. Would you share a car with a random stranger with no "trusted" third-party (i.e. the driver) present?
Hopefully we can use this moment to purge inner city areas of cars.
The national media of course wants to produce content confirming the biases of people who are (in the back of their minds) anxious about the prospect of SDC but don’t have exposure to them. I have a lot of exposure to them - not as an employee though - and think Cruise, Waymo, and Zoox are all doing a great job at being cautious and respectful with their testing programs. Sure they do get stuck sometimes, but human drivers disrupt traffic too, and honestly SDC are already better at being safe drivers around pedestrians and other vehicles than humans IME. It’s just that they are so cautious sometimes they just stop and create situations like this.
Most people who haven’t been living with SDC for years like we have in SF want to hand wring about them based on articles like this, but the reality is quite different. You’ll note you don’t hear much about the cars injuring people or getting in accidents despite the appetite for negative SDC media coverage.
On top of that they have a bunch of electric scooters and electric skateboards and such.
As to how many such incidents there were, it wasn't a lot. Quote from the article:
> Agency logs show 12 “driverless” reports from September 2022 through March 8, 2023
> Sideshows entail street stunts in which parties perform “doughnuts,” or high-speed circles, burn-outs and other risky maneuvers.
In the summer, I can hear side shows in Oakland, which are miles away. Because the night air is still and the particular acoustics of the terrain, I guess, it sounds like it is a few blocks away. I once laid in bed for almost an hour listening to it.
You will see weird speed bumps that form a circle that are designed just to stop people from doing these.
The only side-shows I have seen have been in Oakland and even further into the East Bay.
With that being said, I have witnessed Cruise cars lovingly tapping j-walking pedestrians and traffic abiding cyclists around the lower Haight. I have also seen self-driving cars block buses and the street car.
As for the Cruise cars, I believe they were still being trained so were not carrying passengers yet. They might have ironed out the kinks.
> I have witnessed Cruise cars lovingly tapping j-walking pedestrians
I live in the lower Haight as well and have never personally witnessed that, but don't doubt that it happens. The non-self driving cars here don't interact with the Wiggle super well either though.
The crux of the article seems to be “wow! If you block traffic it really adds up to wasting people’s time!”. Duh. Lots of things do this, many less valuable than driverless cars.
As an example, Seattle’s drawbridges sound like much more of a disaster than whatever time is being wasted by Cruise vehicles. Sailboats can’t even make it under the Fremont bridge, so every rich dude with a boat blocks traffic (including multiple bus lines) for 10 min when they sail by.
I've dreamt of a reverse-toll system where those folks end up paying drivers, pedestrians, and bus passengers some token amount for the delay (not enough to encourage lolligagging but enough to encourage boats to find better travel times or coordinate to spread the tolls.
Special laws sometimes are passed to have "boat times" so the boat has to wait until the top of the hour or similar.
There are some special laws about when the bridge can open during the week, but not on the weekend.
Those bridges don't open 7-9AM and 4-6PM, which shows that we have and should continue to regulate to mitigate disruptions!
Unsafe/stupid driving is accepted as inevitable and rarely enforced against for human drivers (outside accidents). Not a good thing, but definitely the reality these days. Tbh, it’ll likely be much easier to enforce rules on a self driving car, since almost everything is tracked and recorded. Every mistake can be scrutinized for a self driving car, in ways that would never be required for human drivers.
If the purpose is to spot bad behavior and then correct it within these systems, then this is as far from achieving that goal as you can get. I'm not interested in penalizing the corporation, I'm interested in getting the best performance we can.
What natural incentives are there otherwise? If users come to expect that their self driving cab "breaks a few rules" to get them there faster, then how do you balance that?
> Unsafe/stupid driving is accepted as inevitable and rarely enforced against for human drivers
What is your source for this assertion?
> it’ll likely be much easier to enforce rules on a self driving car, since almost everything is tracked and recorded
If the car can't correctly drive itself, then what help is this? How would you sort through this data without the very type of automation which is in the car itself?
> in ways that would never be required for human drivers.
There's much more to roadway enforcement than what the U.S. currently does. China has an entirely different system of police and thus traffic enforcement. European countries love speed cameras. Several U.S. states are big on red light cameras.
Self driving vehicles confidently drive down one way streets and block other vehicles offering no recourse to solve the problem for the other driver. We don't have an enforcement mechanism for this because this hasn't happened before.
Whether this technology is "better than humans" or not is beside the point. It's using roadways designed and still majority occupied by actual humans. This is not being properly accounted for.
You can steal ~899$ worth of shit and the cops don’t show up. Open air drug markets, people shitting in the streets, thousands of homeless people. But yeah let’s pontificate about fines for someone disrupting traffic.
You gonna fine the person that runs out of gas, causes an accident or just mechanically breaks down?
https://sftreasurer.org/business/taxes-fees/traffic-congesti...
> As driverless cars keep racking up the miles, San Francisco transit advocates propose a variety of measures to lessen their impact. Jaime Viloria of Equity on Public Transit, a grassroots group of riders in the Tenderloin neighborhood, says companies operating autonomous vehicles should be fined for causing delays.
Of course, in the case of SF you could argue they have far more highly paid tech jobs than the city can support already, and no need to bootstrap a technology hub....
What they seem to be missing is a “get out of the way” path-planning mode when it doesn’t know how to proceed.
Slack notification comes in: a car is stuck. Click link, get jacked-in and take manual control to move it to a safe location for further debugging.
So about one unplanned stop every 3 days? Seems a bit of a stretch to call it "clogging" San Francisco.
By a regular old delivery truck.
Better statistical data on these incidents is called for.
I think after a few times bailing their cars out of impound, these companies would find a way to "return to profitability".
Not sure what you do in that situation, honestly.
Not pushing a luddite agenda, just curious how it might all play out.
I don’t think they’ll have much trouble in Boston. The current Cruise service area includes a lot of steep hills, narrow roads with parking/pedestrians on each side, and some winding streets. Driverless cars are typically pretty cautious around hazards like this, sometimes a bit too much (like driving up to a stop sign too slowly), but matching what I would want a “safe driver” to do.
I took a Cruise in the rain a few days ago and the rain didn’t seem to cause any problems at all. Even though you can only take Cruise at night, between 9 and 12 pm there have still been plenty of situations where it had to deal with pedestrians walking across middle of streets, opening and closing doors of cars parked on the street, cars stopping in the middle of the road to unload pedestrians, etc. and it’s actually handled all of them well.
When the cars get stuck, I am pretty sure it’s because they are in “mapping mode”. I think it’s weird and likely unscaleable for an entire service area to need to be mapped before opened commercially, but once the cars are actually “working” they don’t get stuck much.
Anyway, I think the only real problem with Boston would be snow and ice (on the ground, I think the rains and fog in SF are close enough that cars could handle snow and ice in the air).As far as I know, no self driving car company has tested much under snowy/icy conditions.
humans have been driving for a century and we still have 30,000 traffic fatalities per year in just the U.S.: human-level automation (AGI) doesn’t solve that. sure, you could tune the driving parameters to significantly reduce that count — but not for free and not without creating new (but hopefully smaller) problems. at some point it just mirrors any other unsolved human system, where you’re left debating values and tradeoffs. think the Ford Pinto memo [1] but where the tradeoffs are “do we tune driving parameters to achieve 10 fatalities/year and 10M man-years of travel time per year, or 1 fatality and 11M man-years of travel time?”
personally i think the tradeoffs in this article are overblown. 99% of us accept that traffic lights are stupid and will delay every 2am no-other-cars-on-the-road trip by a few minutes. but when it’s a new autonomous system suddenly those delays are to be treated as unsolved problems? don’t get me wrong, i like when people acknowledge that there are things around us that ought to be improved, but the distinction here just seems arbitrary. it’s just some weird status quo bias that seems to carry more weight than i can justify.
1: https://en.wikipedia.org/wiki/Ford_Pinto#Cost–benefit_analys...
I just hope that they will continue to extend that tolerance to still allowing human drivers for a long time after automated ones are more safe. It's almost inevitable that there will be a story like this someday about the hazards of letting humans disable their auto-pilot.
The bus- / train-riders have no choice in the matter.
It adds hours onto the driver's day.
It's a shame the mitigation takes as long as the article claims it does.
"They oughtn't to let old ladies carry magnums," he murmured.
Witherspoon-Hobbs nodded agreement across the front-seat corpse. "They oughtn't to let 'em carry anything. God, how I hate Feet," he muttered, looking down at his shrunken legs. "Wheels forever!" he softly cheered.
Literally spent close to half an hour stuck like this once in SOMA behind one of these.
This could be that.
look at the dashcam footage. What's taking up space?
The driverless car, or the long lines of parked cars on each side of the road?
It seems so limited. If I had to bet I’d say we solve general AI first and then driverless cars are somewhat of a byproduct. Like many oldskool specialty AI’s get owned by general purpose LLM’s nowadays.
my problem with busses is that they are huge and, in my experience, mostly empty. they create traffic and their weight damages roads.
as far as public transportation goes, busses are as braindead as it gets.
I wonder what amount of disruption was caused by homeless and mentally ill people during the same period.
Waymo has a lot of miles in quiet suburbs with broad streets and light traffic. How has that helped them when they get to SF?
(Remember this the next time someone says "teachers are the reason our kids are dumb".)
Insufficient. They'll only count it as part of the cost of doing business.
I think every person director or above at these companies should be eligible to be delayed at random for a time equal to the total human-minutes of delay caused to transit riders and people in private vehicles. If a driverless vehicle blocks a train with 60 riders for 15 minutes, those directors and executives should in aggregate spend 15 hours not able to go wherever they want to go, and in a way which they are unable to schedule in advance.