Cruise vehicle gets stuck in wet concrete while driving in San Francisco
sfgate.com
sfgate.com
Discussion here: https://news.ycombinator.com/item?id=37142065
2)Absolute counts worldwide are meaningless when comparing against crashes in SF.
3)You picked an arbitrary period
4)Whether it makes the news is not reliable for statistical comparison. Millions of crashes go unmentioned. Many of them don't even result in a report to any officials.
Was this a freak accident, or indicative of expected performance.
I guess we'll only find out if more Cruise cars turn themselves into modern art.
It is non-fatal for starters, which makes it in a separate category completely from the worst idk ~200 or so problems on roadways.
Ask how many of those taxi drivers how many have interacted with their smartphone while driving, and you will find that 100% of them have been part of a common roadway problem that leads to fatalities.
So, one Cruise vehicle driving into concrete is about equivalent to 3 million human drivers driving into concrete on an equal rate basis.
Does this hurt or help Waymo? They're being painted with the same brush, but perhaps they can leverage that into regulatory capture. When the inevitable regulations come along they'll likely be set to "difficult, but achievable". Which at this point feels like "Waymo pass, Tesla & Cruise fail".
> In May, a driverless Waymo car blocked a fire vehicle while it was backing into a station.
Granted, people are this dumb every single day. But who is liable for the fines here?
No idea how that would apply specifically here though.
The problem here is dispersed responsibility so no one is responsible
The result is that no one takes responsibility for not actually creating blatantly dangerous events, such as blowing through cones, signs and other obvious indicators of a construction / no-driving zone and endangering construction workers & passengers (cones also initially mark a washout/bridge-out problem). To the software developers and managers, it is just some lower-priority edge case they'll handle later. To the corporate promoters, it's just a cost to be externalized on the public. To the regulators, obviously, they don't want to be seen as "Luddites", so they just approved MORE, not less, of these obviously-not-self-driving cars.
If a person actually had both personal responsibility (e.g., they will get fined or potentially imprisoned for approx. negligent homicide if death results) and authority (e.g., they can stop deployment until it is truly fixed after a car mistakes a truck for the sky and decapitaties the driver or blows through a construction zone) to prevent these situations, they would not happen as much. Who should have such responsibility and authority? Start with the CEO and board.
EDIT: stray words, tenses, added examples
Same with emergency vehicles, which self-driving cars have already been shown to ignore.
I'm not debating whether an AI car should "see" possible construction, even when not marked appropriately (most humans can too). Only pointing out we don't know if this zone was marked/flagged - and if it wasn't, there's a higher likelihood of human driver error as well.
Light control at an intersection out? They'll just blink red, indicating a stop sign. Maybe they're off, and they put up temporary stop signs on the corners, or as you mentioned, a police officer redirects traffic. And then a fire engine will run through it, as they need to.
https://www.google.com/search?sca_esv=557782243&sxsrf=AB5stB...
Looking at those pictures, I think loads of people could have made that mistake for quite a few of them, as it often doesn't seem indicated very clearly, and in some situations going through the wet concrete was the only way the car could go from a parked position.
If it was then clearly the car's fault, if not then the worker's fault. Cones become more critical with the proliferation of non-human drivers.
If a construction zone is unmarked, human drivers are more likely to drive into it as well.
Anyway, AI cars should be smart enough to "see" a potential unmarked construction zone, just as most humans can. I'm only saying construction zones should be marked explicitly for the safety of all involved.
The good news is they can fix the self-driving software then replicate the solution forever in millions of cars.
Now, how to prevent the small number of humans who do the same thing?
These sorts of safety issues have been a always been a problem, with lax markings, etc, but now are more acute with non-human drivers.
I think ultimately it raises the question of whether it's reasonable to offer some accommodations to non-human drivers where they don't perform so well. Do they have to be at the standard of human drivers for everything? That's probably impossible.
Ironic. Cones are the self-driving car’s only weakness, and yet they can’t survive without them
Also the picture in the article shows both cones and workers, the presence of either would generally cause a human to alter how they drive.
What we see consistently is self driving cars that randomly stop in perfectly reasonable driving conditions, and similarly we see repeated cases of them driving through clearly unsafe areas. If a single human driver repeatedly made these same mistakes they would lose their license, but for some reason we allow it from these unsafe self driving systems. If a car manufacturer ships a car that is found to operate unsafely, the manufacturer is required to recall them, and isn't allowed to return the car to market until the problem is demonstrably fixed.
False dilemma: value from self driving cars is a sliding scale, not all or nothing.
I've been having that problem with archive.is links for weeks.
I'm not framing anything...
People had "primal fear" of seatbelts when they were introduced, even though the statistics, like those of AV, prove their absolute benefit in the preservation of life.
They just stop if they lose service?
I agree that "stopping" should happen. But, they need to stop safely (pulled onto the shoulder/parking lane/low volume side road), not just abruptly stop in the middle of the street, blocking the rest of the traffic.[1]
1 - https://www.wsmv.com/2023/08/15/driverless-cars-stall-causin...
The question is whether they learn from it.
https://www.theonion.com/high-unemployment-rate-linked-to-on...
EDIT: or, more appropriately, the case of Prawo Jazdy: an alleged Polish national with hundreds of speeding tickets (Prawo Jazdy is Polish for driving licence)
There is a huge potential benefit. If the cars are going to improve they need to be tested in the real world. So maybe some forbearance is justified?
Just like a learner driver who stalls the engine, forgets to put on a turn signal and can't keep up with traffic.
We've all got to learn sometime.
Also humans can learn from this too. It’s not hard to imagine a new type of safety cone that has an explicit sign for autonomous vehicles to stop far far away.
Isn't that kind of defeating the point of autonomous vehicles if we need to build a world that will accommodate them?
It'd just be done such that new signage is made to be easily readable, so that over time, as old stuff is replaced, the signage becomes more readable for machines.
This sort of thing can also benefit regular vehicles.
I'm not sure what "This sort of thing can also benefit regular vehicles." means in this instance. If the cone is machine readable how does that help me?
The cones being machine readable would allow regular vehicles to also be given the ability to read them more reliably (and potentially with less equipment than needed by a current autonomous vehicle), so, they could do things like warn the driver, which could be valuable in reducing the cases of distracted drivers slamming into closed off lanes.
This comparison doesn't make much sense. We're highlighting a dangerous failure in capability here, not a desire to change out the engine. Either engine works perfectly fine, it's secondary properties that we care about with transition. It's a primary responsibility of an autonomous vehicle to be able to identify and react to obstacles.
We either need the enhanced cones or we don't, but only a fraction of cones being enhanced means they aren't actually that useful or that the situations without them are inherently more dangerous. Are we OK with a situation where autonomous vehicles identify the enhanced cones 90% of the time but the non-enhanced 50%? No, we want both to be very high.
I agree that we should strive to improve the vehicles' ability to detect even unenhanced cones. Doesn't mean that we can't also aim to gradually improve detection accuracy further with enhanced cones. That is, if autonomous vehicles identify regular cones correctly 98% of the time, but enhanced cones can be identified 99.99% of the time, we should be okay with a gradual transition to the latter, even though tautologically it means that the former is less safe.
We already do this with all sorts of things, safety standards improvements often have at least a grace period during which the comparatively less safe things are still allowed to exist and operate alongside the safer ones.
I dont believe thay every car maker will not try to invent their own models
Car companies will never build in failsafes like that, it would be too easy for anyone to put up those signs anywhere. Cars and streets are made for human drivers, there is no way around it and no alternative to making your car AI as good as human drivers. If you can't do that, you can make a much simpler train "AI" and put it on tracks. We did that in the 60s.
Two things:
The promise of self-driving cars is that they can operate themselves in an environment not specifically designed for them. If that promise is broken, and it's the position of the industry that things like traffic signals, construction, emergency services, road markings etc need to be modified for functionality with self-driving cars then I think public will to accommodate them dries up quickly.
That being said, I can imagine having a lot of fun with a physical token that can remotely stop an autonomous car.
The lesson should be that San Francisco isn't the right place to run a highly disruptive experiment.
Do these experiments where the potential for disruption is a lot lower, like a smaller city with less congestion.
Should they be banished too?
But I'll take your bait:
I grew up close to a city, (Worcester MA,) with plenty of weirdo streets like San Francisco, and had plenty of teenage friends who lived in the city or travelled through it.
Needless to say, we didn't get behind the wheel for the first time in Worcester. We stayed in calmer streets until we were ready. When we drove in Worcester with learner's permits, we had adult supervision. Afterwards, we were only allowed to drive without adult supervision once we passed the driver's test. (I should also point out that, at the time, Worcester had a very serious problem with unlicensed drivers.)
Clearly, Cruise's vehicles aren't ready to drive in San Francisco without adult supervision.
But let's get back to the thesis: It is highly inappropriate to equate a machine with a human. Machines have no rights whatsoever. The "We've all got to learn sometime" attitude doesn't apply to this discussion, because an autonomous car is not a human.
What instead applies to this argument is conventional engineering and business development: Where is it appropriate to run Cruise's early marketing experiments, where bugs (and other unanticipated behavior) is expected? What are the appropriate limitations to put on vehicles when they have no driver? For example, what if the first use of the vehicles was a private resort (like Disney World) where the nature of a malfunctioning vehicle is easier to absorb because there are very few passenger vehicles.
Am I to assume you never made a single mistake while doing so?
Of course not. Learner drivers stall, they forget turn signals, they don't keep up well with the flow of traffic. That is fine, that's all part of learning and perfectly normal.
> Clearly, Cruise's vehicles aren't ready to drive in San Francisco without adult supervision.
I don't think that is clear at all. Sure, they're making mistakes, but so did you and I and everyone else that ever learned to drive.
> But let's get back to the thesis: It is highly inappropriate to equate a machine with a human. Machines have no rights whatsoever. The "We've all got to learn sometime" attitude doesn't apply to this discussion, because an autonomous car is not a human.
I don't think that has anything to do with it. Driving a car is not a "right" that all humans have, it's a privilege that can be quickly taken away.
Everyone is afforded the opportunity to learn at some point, and we need self-driving vehicles to have that opportunity too. Fast forward 20 (or whatever it is) years to when self-driving vehicles are much, much safer, they mean long-haul truckers don't have to be away from home in an unsafe and unhealthy profession, they mean less vehicles on the road, etc. etc.
Society NEEDS those benefits, and if we never give the self-driving vehicles the opportunity to learn, we'll never get there.
> What instead applies to this argument is conventional engineering and business development: Where is it appropriate to run Cruise's early marketing experiments, where bugs (and other unanticipated behavior) is expected? What are the appropriate limitations to put on vehicles when they have no driver? For example, what if the first use of the vehicles was a private resort (like Disney World) where the nature of a malfunctioning vehicle is easier to absorb because there are very few passenger vehicles.
Absolutely, those are very important discussions to have. They are being had, by people in a position to make those decisions. Like a million things in our society, if you don't like the decisions they're making, you need to get yourself into that position.
For example: Yes I did stall, but I did not block traffic for 20 minutes. I started the car and moved the shifter from 3rd to 1st. I did not require outside intervention to move my car.
I am also a person with fundamental human rights. A self-driving car has no rights, and deserves no empathy.
Which is why I'm trying to refocus the discussion back to "tech."
For example, when we talk about disruptive tech: The early customers need to be willing to put up with the "faults." In this case, San Francisco's emergency response departments aren't willing to put up with these faults.
Likewise, when we talk about choosing a market for early markets: That means making sure the market is well-chosen to suite the capabilities of the tech. One of the issues in San Francisco is that demand for self-driving cars outstrips the capabilities at this point. IE, it's better to choose a place where Cruise can meet the demand.
Are you saying that driving a vehicle is a fundamental human right?
Becuase it is absolutely 100% not.
> In this case, San Francisco's emergency response departments aren't willing to put up with these faults.
What are you talking about? San Francisco's emergency response departments absolutely ARE putting up with these faults, and the regulators and people who make decisions about if they should or should not be on those streets are deciding they should be.
I understand the self-driving cars can be inconvenient right now, but that always happens when you're aiming for improvement or progress. When lanes get added to a road the traffic suffers during construction. When you renovate your kitchen it's painful to live in during the work, etc. etc. Just because it's inconvenient doesn't mean you shouldn't do it - the eventual improvement will be worth it.!
This argument comes from my interpretation of two books: "Crossing the Chasm" and "The Innovator's Dilemma."
Cruise can choose any market they want: any city, any municipality, any closed road network like a resort. I keep arguing that San Francisco is the wrong choice right now.
Choosing an early market is an important step in developing a technology business. It doesn't matter how wonderful your technology is, your company needs to succeed in its first market in order to move to another market and then to larger markets. Furthermore, your customers must be willing to put up with the bugs and shortcomings of your product compared to existing technology.
If a company overestimates the maturity of their product, they can lose the goodwill of the future customers who don't want to put up with bugs or other shortcomings.
I think the big problem that Cruise is facing is that they will be regulated out of existence before they fully debug their product. (For the sake of argument,) if they had used resorts as their first customer, They wouldn't have to worry about being regulated out of existence because they would only have to deal with a government; and they would have had a much easier situation to debug the product.
There are plenty of regulators and people who's entire job is to figure out if it makes sense in their city or not. Those people are in charge of making that decision for San Francisco, not you.
As a city, San Francisco has a lot of citizen initiated ballot questions. (When I lived in the city they had an initiative to rename the sewage treatment plant after George W. Bush.)
I wouldn't be surprised if an upcoming ballot has an initiative to ban self-driving cars.
DMV tells Cruise to reduce its driverless vehicle fleet in SF by 50%: https://www.sfgate.com/tech/article/firetruck-siren-lights-o...
Proves my point that San Francisco is a horrible place to test market driverless cars.
> In a statement released Friday night, the DMV said it is investigating “recent concerning incidents” involving Cruise's vehicles in San Francisco. In addition to halving the number of vehicles in the city, the DMV called for Cruise, a subsidiary of General Motors, to have no more than 50 vehicles operating during the day and 150 vehicles driving at night until the agency's investigation is complete.
> As SFGATE previously reported at the time of the Aug. 7 meeting, which featured a slide deck on autonomous vehicle performance in the city, “In 2023, the department logged about 50 incidents involving AVs that nearly crashed into personnel, obstructed travel or blocked stations, per the presentation — and five more reports were written up over the weekend.”
> Nicholson provided a statement in response to Chiu’s move, sent to SFGATE on Thursday by the city attorney’s office. “We do not believe the industry has any incentive to remain at the table and solve their problems,” she said. “These incidents with Public Safety are not going away and are in fact increasing.”
There are plenty of options that negate this as a concern (actually it's not even a concern), especially when the alternative is big government autonomous vehicle welfare shoving ads up your ass and nickel-and-diming every move you make to participate in society.
Want to protest something? Sorry the government shut down your ride. Whoops you were mistaken for a terrorist - straight to jail! Did you forget to pay your Waymo bill? Guess you're stuck at home until you get the funds to pay it. Heading to the airport? Sorry sir you can't exit the car while the car is in motion, 40 minutes ETA until you arrive at your destination. But I can walk, it's right there! Sorry sir for your safety and those around you please enjoy this new Netflix special while you wait. You can purchase this bottle of condensed milk for $7.99 and it'll automatically be deducted from your paycheck.
Car-only infrastructure isn't just a really stupid and expensive way of doing things, it also keeps you under control.
That idea doesn't make sense. Where would the space for the dedicated lanes come from, especially on city streets (like the one in the OP)? The existing traffic could use those lanes more effectively, and it would be wrong to waste it because driverless cars suck and their developers are insistently shoving them out the door despite that.
Because driverless cars are safer, faster and more efficient.
Good public transit is WAY safer and more efficient than cars of any sort. It could be faster, too, if we prioritized it over cars like much of Europe has.
Beg your pardon? How is 2.5 tons of steel to carry a handful of people more efficient than a tram?
Public trans? I already said I'd support that, but it's not on the menu is it.
The other half is the stuff you can't do, like carry large items or eat on the train. Just got some large boards for a DIY project and stopped by in-N-out. ~10 minutes there and back. Would have taken an hour w/PT, couldn't carry materials and couldn't stop for food. With a driver-less car, I could still do that.
I'm not suggesting any money go towards paying for driver-less lanes atm though. Still feels way too early for that.
If it took you ~10 minutes to get to the store, it's probably ~2 miles by car. Maybe a bit further. That's easily done on a bicycle (not for large boards, but easily doable for moderate grocery runs or other regular errands). Or, a bus/tram - might take a few minutes more waiting for it to arrive at each end, but still not an hour (if it runs regularly enough, which it obviously doesn't).
How do other countries solve this? Is it just an off-hours thing, or a problem mid-day?
The problem is nobody does anything because nobody else cares enough, tbh.
Nobody cares enough to call the police because they know the police won't do anything. The police won't do anything because they'll arrest people and those people will be let out a few days later with no punishment. Why? Ask the elite representatives in DC I guess. They probably just don't care to solve it or don't want to solve it. There was a dude who vandalized a store and stole some stuff in DC, got arrested, let out 2 weeks later and literally 2 days after being released did the same exact thing.
I've unfortunately seen creeps doing awful things in front of innocent people too and its one of those things that nobody will spend time tracking down I guess.
I hate to be the bearer of bad news, but its closer to 7. Speed limit is 70 mph for most of the drive, haha.
Even if it was 2 miles, that's gonna be close to 30 min each way, an hour spent just on commuting for groceries. I don't think I could justify it tbh, and I like biking enough. I'm usually just carrying stuff though, if I'm leaving the house to go shopping.
If it was going to/from the exact place, perhaps. It's just so convenient to grab the keys and not have to worry about schedules or arranging how you'll carry everything back.
Perhaps if I lived in a city and it was my only choice though.
Also, isn't Musk setting up Hyperloop, or was that another publicity stunt?
Re. Hyperloop: some have said it was a stunt to attack public support for CA high speed rail (which would reduce CA's dependency on cars): https://twitter.com/parismarx/status/1571628269555826688?lan...
EDIT: Re-reading your comment, I realize that you're challenging GP's assertion that we already have bike lanes everywhere, which I agree is obviously false.
For reference, two of the cities metioned are Austin with ~2500 miles of roads and Denver with 1900 miles.
It's a start. But that's all it is.
Bike lanes make sense because bikes can't go as fast as cars and are unprotected in the event of a collision. Bus lanes make sense because buses have dedicated predictable routes, and efficient bus lanes increase the number of people that can get around.
I don't mean doing this today. I'm talking about a future where most vehicles have a driverless capability. The reason why it would be useful is that a lane of automated cars would be safer and more efficient. But the reality we have right now is that driverless cars are intermixed with human-driven cars, which will always be a danger.
It happens all the time
> Accidents per million miles driven for Waymo self-driving cars are 0.59, compared to the general U.S. rate of 2.98
https://blog.gitnux.com/self-driving-cars-safety-statistics/