To be clear, this has nothing to do with security or any other legitimate concern. The point of it is to give dealers more business and to harm automakers that don't have dealers.
An Oklahoma state rep is currently pushing to ban Tesla from the state (although Tesla could move to tribal lands to avoid such actions if necessary) through House Bill 3994 [1]. They also manage family owned legacy automaker dealerships (per LinkedIn [2]).
Lots of bad faith efforts going to occur on the path to clean mobility due to entrenched interests.
Similarly, Biden is trying to pass EV subsidies for non-Tesla auto manufacturers, and then there was the amusing couple of weeks where he would praise GM and Ford (but not Tesla) for paving the way in EV (Tesla sold 350K EVs in the US while GM and Ford collectively sold fewer than 60K). Moreover, it feels like there's been a surge in anti-Tesla automotive reporting recently (e.g., Roadshow omitting any Tesla cars from its 2022 list of best EVs lol).
One definitely gets the impression that the legacy auto industry is scared shitless and spending out its ass on politicians and media.
I dunno about the Biden admin’s attitude toward Tesla, but some of the other stuff is Tesla reaping what it sows by not having a press office. There’s a reason corporations have press offices! If you don’t engage with the media on their terms, you’re at their mercy. In this sense, the continuous attacks by the Tesla stan army are actually not helpful to Tesla’s long term cause.
If I’m an auto journalist and can’t get a press car, and all of the other manufacturers from Rolls Royce and Porsche on down give me press cars, I’m gonna have some natural bias. This is human nature, not racketeering.
Is "engaging with the media on their terms" a euphemism for paying them for favorable coverage? Because that's pretty clearly the discriminant here. Tesla outcompeted in 2021 and consequently legacy auto manufacturers paid car publications to slant their coverage. Publications like Roadshow even have a "we may or may not get paid by manufacturers for our coverage" disclaimers at the top of their articles.
> In this sense, the continuous attacks by the Tesla stan army
I genuinely have no idea what this might be referring to.
I have a hard time believing it was the inability to contact Tesla and not the large sums of money and implied favors from legacy automotive manufacturers which persuaded politicians to pass (or try to pass) overtly anti-Tesla policy.
Similarly, I have a hard time believing that were it for a Tesla PR department Joe Biden's team would've been aware that it was Tesla that was the driving force behind American EV efforts and not Ford or GM (the latter of whom sold a whopping 26 EVs nationally in 2021 Q4).
Similarly, I have a hard time believing that automotive publications that regularly test drive expensive cars were unable to get their hands on a Tesla because Tesla lacks a press department.
Most of all, I find it conspicuous that all of these things are contemporaneous with Tesla eating legacy automotive's lunch rather than the time Musk dissolved Tesla's PR department a year prior.
There's a couple of different problems here - but both really are related to the question of aligning incentives.
In terms of PR departments, the point is most journalists - for their own long term career/reputations - would prefer to outline both sides of view or at least response quotations for their stories to create an appearance of impartiality. However, when working on deadlines, they won't go out of the way or jump over hoops to do so. The PR department simplifies this massively working towards journalist's incentives and minimizes negative stories & negativity of stories that get put out. Musk seems to be a control freak in that he seems to be unable to delegate PR but the trade-off is he can't personally influence every story.
In terms of the Government piece, the Democratic party has always had a core of unions & worker rights underlying the core base. Musk tweets and Tesla stories have portrayed him been openly hostile & contemptuous of this space and seemingly without anyone to gainsay him, there's no one to bridge. So is it any surprise Tesla gets ignored?
Make no mistake - EV is rapidly becoming commoditized. With full-self driving no were near the horizon for anyone in terms of production, margins are going to shrink as there becomes very little to differentiate products and these mis-steps can risk the business.
I’m not buying it. A PR department isn’t going to make the difference between whether or not the press or politicians do their job on a basic level. I specifically don’t believe that Tesla’s PR department would have nullified Biden’s (and other politicians) incentives to please Detroit.
> Make no mistake - EV is rapidly becoming commoditized. With full-self driving no were near the horizon for anyone in terms of production, margins are going to shrink as there becomes very little to differentiate products and these mis-steps can risk the business.
There’s a wide chasm between Tesla and generic EVs with respect to manufacturing processes, battery technology (especially range), software integration, charging network, and so on. Tesla’s mote is formidable and will endure for the foreseeable future irrespective of level 5 FSD.
Why would Elon firing his PR department motivate legacy autos to pour money into politics and media? Seems to me that Tesla’s success at taking over the market is a much more likely motive.
Don’t hate the player (the PR employee). If you want to influence government policy you have to play the game. If you don’t, great, that’s good (I support that), but don’t complain when the President doesn't mention Tesla. Engineers are sometimes too sheltered, we only complain on our forums here after the fact.
The PR department is irrelevant, precisely no one is complaining about any fallout that could credibly be traced back to Musk dismantling the PR department. “The game” isn’t funding a PR department, it’s lining the pockets of politicians and media, and I’m glad Tesla and Musk aren’t playing it (and yet they’re winning!).
and yet, I still want them to do it, because I think people should have the option of not needing to guess if their car has a new janky update before heading off in the morning.
If you don't like OTA updates, why not just buy a car from a manufacturer that doesn't do them? Why make it so that even people who do like them can't have them?
Because, once you're on the road, your car affects everyone else around you.
Have automatic updates for your personal machine. If Windows pushes out a patch that bricks your computer, I'm not impacted. If Tesla pushes out an OTA that is buggy and causes accidents, I'm affected.
If it keeps Teslas with 'aggressive' autopilot from doing illegal and dangerous things like rolling stops, I don't care what concern it stems from.
Anything to stop bad companies from sending out half-baked patches is a net good. I don't care if you have to roll back an update on your SaaS site. Sucks, but we'll move on. But continuous integration should stay far away from what amount to missiles that weigh tons and go in excess of 60 miles.
> If it keeps Teslas with 'aggressive' autopilot from doing illegal and dangerous things like rolling stops
It doesn't do that. Tesla would still be able to push out whatever updates it wants. Drivers would just have to either go to a service center or leave the state to install them.
The more structural obstacles there are to pushing out patches, the fewer patches will be pushed out. The company would have an incentive to rigorously test the patches before sending them out.
Currently, with OTA patches, you can throw out any patch you want because it's self-serve. But if customers were inconvenienced by having to go to a service center for an update, Tesla would have a larger incentive to get things right the first time. It would damage their brand reputation if a customer had to go to a service center for a patch and then turn around and schedule two more appointments to download bugfix patches.
Ah, the old "make it harder for customers to install safety updates" argument.
> The company would have an incentive to rigorously test the patches before sending them out.
This doesn't stop traditional automotive companies from releasing buggy software (they have far more recalls per unit time than Tesla), and anyway the issue you noted above wasn't a bug, it was a perfectly safe _optional_ feature that NHTSA didn't like.
> Drivers would just have to either go to a service center or leave the state to install them.
Tesla service centers are not a "new motor vehicle dealer", you would need to leave the state to install updates
> Except for experimental low-volume not-for-retail sale vehicles, cause warranty and recall repair work to be performed by any entity other than a new motor vehicle dealer
The part about OTA update being a benefit to all - yes I agree.
But aggressive autopilot being evil? That's a stretch.
Sure human operators need to stop every time, look both ways, proceed cautiously. Makes sense.
But a car with sensors operating 24x7, never getting tired or taking an intersection for granted? The rules can be different. Should be different. Because there's no added risk; because we can adjust them should an issue arrive, and they will always, always do it the new way.
Computers are different from humans. They drive differently. Insisting they behave the same way, is like insisting drivers have a person on foot proceeding ahead honking a horn and waving a flag (as early localities insisted drivers of the new automobile should have to do) because they might scare the horses.
Yes, in states where stopping at a stop sign is the norm, come to a stop even when no one is there. This ensures predictability, which is important for both driver and pedestrian safety.
If you are in a state like Massachusetts where a rolling stop is the norm, then for the same reason it's best if the car does not come to a complete stop, regardless of whether someone is there, because it makes your vehicle's motion more predictable. Ideally laws in those states should also be adjusted to reflect what people actually do.
It is important to know both that no one else is at the intersection and that no one else is approaching the intersection. Consider a two-way stop where an approaching car is exceeding the legal speed limit (as is the norm everywhere I've driven in the US); would you still argue that a rolling stop makes no (zero) impact on safety?
Genuine question: are there any states where full-stopping at a stop sign is the norm whether or not anyone is there? I've done a whole lot of cross-country driving and I've never been to a place where the norm is coming to a complete stop even if there is traffic present.
In my state (SC), the people I personally know who have been ticketed for a rolling stop have typically been the only vehicle at the intersection (sometimes the only vehicle on the entire road). I suspect this is because the infraction is more visible (it's illegal either way afaik).
I'm not talking about "if no one is there will you be ticketed", I'm asking where is it the norm to come to a full stop even if there are multiple people at the intersection? In the midwest and southwest it's common to "California roll" through a stop sign, and given the name of the maneuver, I assume this is also common out west.
My first time visiting Boston I came to a four-way stop leaving the airport. I was surprised that not a single car came to a full stop in the other three directions. They did not yield to the vehicle on the right as is the norm in my state; since I had come to a full stop, they perceived me as voluntarily yielding the right-of-way. When I explained to a local that the other drivers weren't yielding, the reply was "oh, you didn't come to a full stop, did you? Don't do that.".
I live in SC. For us a full stop is still the norm, regardless of whether there are other cars at the intersection, though this is changing as more people move here from other states. I have seen other drivers make a rolling stop, but nowhere outside of Boston have I ever been penalized by other drivers for obeying the sign.
In SC, the law requires a stop (making no distinction between a full stop and a rolling stop) at a stop sign or red stop light, prior to the intersection, either at the stop line or before the near side of the crosswalk. AFAIK the only provision we have for a rolling stop is for emergency vehicles.
'Because there's no added risk; because we can adjust them should an issue arrive, and they will always, always do it the new way.' <-- That's quite a sterile, abstract way to say 'We can change how the computer behaves if the computer causes an accident.' Which is to say if the computer damages property and non-consenting people.
Rolling stops are illegal because they're dangerous. If a computer malfunctions, runs a stop sign and t-bones me, I could very well be paralyzed given the multiple structural issues with my spine.
My bodily health is more than an 'issue' which can be coded away. Because Tesla can fix their code, but that code fix isn't going to fix my back.
If the computer malfunctions, it could t-bone you even though it fully stopped first. It's basically anthropomorphizing the computer to say that a full stop makes it safer.
Ensuring sufficient reaction time isn't the only reason for a full stop. A full stop is a signal to those around you that you are yielding the right of way. If there is another way to signal (e.g. all other cars are a autonomous vehicles communicating wirelessly and there are no pedestrians) then there is no need for a full stop. In other cases the lack of a clear signal can create confusion as to who has the right of way, leading to a collision.
I see nothing in parent/grandparent/ancestors of my comment qualifying whether or not other cars are at the intersection at the time of the rolling stop. The original comment that spawned this discussion was "If it keeps Teslas with 'aggressive' autopilot from doing illegal and dangerous things like rolling stops, I don't care what concern it stems from."
There is a another thread on which is definitely about whether rolling stop are okay when nobody else is at the intersection (note this is subtly different from whether nobody else is around, especially for a two-way stop instead of a four-way).
> Rolling stops are illegal because they're dangerous.
> If a computer malfunctions, runs a stop sign and t-bones me, I could very well be paralyzed given the multiple structural issues with my spine.
How are these two statements in any way related? A computer malfunction can cause an automated car to run a stop sign and t-bone you whether or not that car is programmed to slow down to 0mph or 5mph.
You're stuck thinking about the way that humans make mistakes, computers can still make mistakes, but they're of a completely different nature.
Consider a thought exercise - In your opinion which of these is a better way for an automated car to approach a stop sign?
1) Slow down to 5mph as it approaches the intersection and evaluate if it's safe to continue; if it is the car may continue without coming to a complete stop. The acceleration profile is designed so that the car has 15 seconds total to evaluate the situation before continuing through the intersection.
2) Slow down to a stop at the intersection and immediately continue if it's deemed to be safe. The acceleration profile is designed so that the car has 10 seconds total to evaluate the situation before continuing through the intersection.
If we were talking about a human there is certainly an argument that option 2 is safer since human are generally bad at multitasking and requiring a period of time dedicated to looking for hazards is a good idea. However, a computer does not have the same issue and option 1 could be considered safer due to the increased time spent observing for hazards.
People cause accidents too. If an autodrive car becomes (is already?) safer than a person, should we not revisit the rules?
Rolling stops are illegal for a person, because a person suffers from inattentiveness and has slow reactions. The rules for a computer start from a different place, and likely will end up different from the rules for people.
Traffic is never going to be completely safe. We accept that now. To hold autodriving to a higher standard is reasonable, but there's a limit. And it should be a rational limit, not just hyperbole.
Strawmen like 'I'll be maimed!' are not really helping the conversation.
> If a computer malfunctions, runs a stop sign and t-bones me,
If the computer malfunctions then the software telling it to come to a complete stop at a stop sign isn't going to save your bacon. You're t-boned either way.
The vast majority of stop signs have plenty of visibility, and could and should have been yield signs instead. Still, as a driver that isn't looking for extra hassles, I need to come to a full stop - despite the time wasted, jostling of my cargo, and needless clutch wear. Self driving vehicles shouldn't be allowed to escape the draconian regulations that we all have to bear, rather the regulations should be reformed for everybody. And this uniformity goes doubly for driving, where being able to understand and predict the actions of other vehicles is paramount.
I hadn't seen anyone say they should "escape the draconian regulations". Non-Tesla's are perfectly capable of rolling stops. I'm assuming a cop can and will still write tickets for this. Not stopping at a stop sign is on the driver whether it's a Tesla or not.
The comment I responded to was directly advocating having different laws for self driving cars, or at least I took it that way:
> Sure human operators need to stop every time, look both ways, proceed cautiously. Makes sense. But a car with sensors operating 24x7, never getting tired or taking an intersection for granted? The rules can be different. Should be different.
That seems strange - a robot should have the same rules as a human? Why-ever would that be? It's a reasonable idea, that we will have more nuanced rules for automation vs human operators.
There is significant added risk. Failure to observe traffic rules by a subset of vehicles means that the drivers - human or otherwise - have to guess the intention and actions of vehicles with less information, and greater uncertainty. A driver rolling through an all-way stop intersection is a threat to other vehicles, as it is not signalling it's intention to abide by the round-robin rule of all-way stops, e.g.
And frankly, in the case of Tesla autopilot, the vehicle is not autonomous, and the responsible party for the motion of the vehicle is still the operator, regardless of how much the computers are doing on their behalf. Under these circumstances, correct observation of traffic laws is mandatory.
That's ignoring the premise that there's no one else there. Humans do it all the time (rolling stop when alone at an intersection) and civilization hasn't fallen.
With my phone I can choose to ignore updates. Anyone who has ever owned a computer knows that an update can make something worse. In fact today I was in a discussion about rolling back a BIOS patch.
There should be an easy way to opt out of you're not interested.
The person you're replying to isn't advocating for a way to opt out of updates. He doesn't want you to be allowed to get them even if you do want them. That's just as much not owning your cars as being forced to take updates you don't want is.
You’re more concerned with Autopilot versus your average human driver? That seems like a terribly uneducated position to take, statistically speaking.
Tens of thousands of people are killed annually in auto accidents where 94-96% are caused by human error. Twelve deaths, total (as of this comment), are directly attributable to Autopilot being active.
I would suggest you reconsider your approach to risk management, as it seems to be emotion vs data driven.
Autopilot only works in the easiest of driving situations. Human drivers include those on drugs.
While I think autopilot type things can be much better than humans, so far the data I've seen doesn't make it possible for me to make a fair judgement on how good/bad it is.
In any case the point remains: you cannot just roll out updates to something safety critical like that. You need a lot of careful work to ensure the update is better than the old in all ways, and that nothing can go wrong.
I'm guessing that high drivers account for at most a sliver of incidents, but moreover, I don't see the purpose in comparing autopilot to only the best drivers. If someone who would otherwise drive high uses autopilot, isn't that an improvement? If someone who would otherwise text while driving uses autopilot, isn't that an improvement? If someone who is just a worse driver than autopilot uses autopilot, isn't that an improvement?
Depends on the goals. If you want people to use autopilot then you need to convince them that it is better than good humans. If you only want the high, then compare to them.
I already know humans are bad drivers, but autopilot is a big unknown.
The goal is improving safety. To improve, autopilot needn't be safer than the best driver, it only needs to be safer than the average driver. Even if adoption depends on proving autopilot is better than good humans, if it fails this test, then people just don't use it and safety is no worse than it is today.
> autopilot is a big unknown
We really just need to know whether Teslas get into more accidents or more severe accidents than other vehicles (modulo some adjusting for demographic differences among drivers). If Teslas contribute to fewer/less-severe accidents, then it demonstrates that they're safer overall. Whether they get there via autopilot or other means is an implementation detail. In any case, Tesla isn't just going to ship unsafe software because it would destroy their brand (every Tesla incident gets 1000x more scrutiny than other brands).
> Autopilot only works in the easiest of driving situations.
Oh no no no, this is not accurate at all. I’ve used Autopilot over 60k miles across the entire US, even in the worst of weather through steep mountain grades at night at speeds up to 85mph (both the Rockies and TN/KY/NC). I disagree with your assessment. It has occasional disengagements during challenging scenes (road lines disappear suddenly, heavy construction causing sharp lane changes) but is an extremely robust system overall.
Your average human’s driving ability is not so great, in comparison.
Humans don't have them very often. We might get things wrong, but we don't disengage in the same way. (humans do sometimes die at the wheel of course, and in general are not good drivers)
That immediately is not a safe thing to do. Humans are known to need a minute to change their mindset to concentrate on driving. Tesla tries to keep humans alert and in control, but they have passed the point where humans feel like they don't need to pay attention.
Of course throwing anything to humans only will skew the data against humans.
I think a lot of people are buying the media FUD. The media breathlessly covers Tesla recalls while ignoring those of other manufacturers (despite that the latter are far more numerous and costly).
What is it specifically, though, that saves 100 by killing 12? If you can justify killing 12 people by putting a bigger hypothetical number next to it, why 100? Why not 100 million?
There are certainly more than 100 million people who did not die since Autopilot came out. Neither you or I can prove it didn't save 100 million people. Therefore if it were to kill 12 million people, that's a net 88 million fewer deaths. What is wrong with that logic?
I need to be educated on where the linkage between lives cost and lives saved is.
The education I need is on how people know that something in Autopilot saves X lives at the cost of Y lives. What, in the physical world, does that? Where does the premise of a tradeoff come from?
The difference is scale. If you make an error, it’s just you that makes the error. If an autopilot dev makes an error, it is amplified to every car on the road. Right now, there’s not a lot of those cars on the road, so we have no idea what that amplification even looks like. We’ve had 12 deaths, with what? Half a million cars on the road? A single bug will hurt maybe a single person right now. If all cars were driving themselves, a single bug could kill millions in a single moment.
dealerships exist because they comprise a nontrivial amount of state revenue and campaign contributions. they could offer negative value (which arguably, they do) and states would still refuse to deregulate.
if you want them gone, in the fourth epoch of neoliberal capitalism, youll need to find an alternate source of tax revenue that not only protects corporate tax cuts but champions austerity and doesnt interfere with deficit spending at the state level.
The bill specifically uses the language "new motor vehicle dealer". So no, this isn't even a win for right-to-repair. It's just the rent-seeking dealer lobby looking to stop themselves from being cut out as the middlemen.
It has to do with security. OTA are an atacker's wet dream. And the potential to do harm is big. It's not that cars run obscure OSs. With the ARMification of devices you have a global platform opened to global threats.
But if dealers no longer add enough value to the process of buying a car, doesn't that just mean that they either have to find new value to add or simply cease to exist because their business model is no longer viable?
It's not even that, really. Older cars (for the most part) had nothing to update, so it's not like these updates are really taking anything away from the dealers that they were doing before. They are just mad that they aren't able to add a new revenue stream by being the only place to go to get software updates.
In a free market, that would be the case. But because of a bunch of laws similar to this proposed one, it's basically illegal for most automakers to not have dealers.
Whenever people bring up how the big old automakers are going to eat Teslas lunch any day now, they leave out the dealership factor. It makes the big automakers uncompetitive regardless of how great their product is.
Try to buy any of these new Ford EVs and be ready for a nightmare of sketchy pre-orders and dealer markups all while being treated like dirt while buying a $70k car.
To be fair, Tesla itself has quite a bit of sketchy pre-ordering and not exactly a great reputation for treating its customers that well after the sale.
Automatic over-the-air updates to cars do constitute a national security concern. A nation-state adversary who infiltrates the manufacturer can brick the cars, impeding the ability of essential employees to commute to work. And if the cars can be updated while on the road (either by design or by exploiting a vulnerability), it can be even worse.
How is this any different than everything else that is connected the grid. Sure a nation state could brick some cars or force them to drive without a driver. Shutting down critical infrastructure such as electricity generation, or water purification would do far more damage.
>Shutting down critical infrastructure such as electricity generation, or water purification would do far more damage.
I mean, this is true right now, and so those things (and hospitals and governments) are being targeted right now. Lots of stories about ransomware recently, and allegedly it's part of the hostilities in Ukraine.
When there is enough centralized infrastructure controlling enough cars, then targeting it will do more damage, so it would be logical to do more of it.
OTA updates are important for any tech product and (hot take) electric cars have more in common with phones that gas cars. They need OTA updates. I would like to see it in law that these upgrades are not required and should be the users choice. I'm pretty sure that's what this bill was supposed to be before it got politics.
123 comments
[ 2.1 ms ] story [ 173 ms ] threadLots of bad faith efforts going to occur on the path to clean mobility due to entrenched interests.
[1] https://engage.tesla.com/articles/1134-oklahoma-oppose-hb-39...
[2] https://i.imgur.com/oGCg1Yv.jpg
One definitely gets the impression that the legacy auto industry is scared shitless and spending out its ass on politicians and media.
That’s a racket, buddy.
> In this sense, the continuous attacks by the Tesla stan army
I genuinely have no idea what this might be referring to.
Similarly, I have a hard time believing that were it for a Tesla PR department Joe Biden's team would've been aware that it was Tesla that was the driving force behind American EV efforts and not Ford or GM (the latter of whom sold a whopping 26 EVs nationally in 2021 Q4).
Similarly, I have a hard time believing that automotive publications that regularly test drive expensive cars were unable to get their hands on a Tesla because Tesla lacks a press department.
Most of all, I find it conspicuous that all of these things are contemporaneous with Tesla eating legacy automotive's lunch rather than the time Musk dissolved Tesla's PR department a year prior.
In terms of PR departments, the point is most journalists - for their own long term career/reputations - would prefer to outline both sides of view or at least response quotations for their stories to create an appearance of impartiality. However, when working on deadlines, they won't go out of the way or jump over hoops to do so. The PR department simplifies this massively working towards journalist's incentives and minimizes negative stories & negativity of stories that get put out. Musk seems to be a control freak in that he seems to be unable to delegate PR but the trade-off is he can't personally influence every story.
In terms of the Government piece, the Democratic party has always had a core of unions & worker rights underlying the core base. Musk tweets and Tesla stories have portrayed him been openly hostile & contemptuous of this space and seemingly without anyone to gainsay him, there's no one to bridge. So is it any surprise Tesla gets ignored?
Make no mistake - EV is rapidly becoming commoditized. With full-self driving no were near the horizon for anyone in terms of production, margins are going to shrink as there becomes very little to differentiate products and these mis-steps can risk the business.
> Make no mistake - EV is rapidly becoming commoditized. With full-self driving no were near the horizon for anyone in terms of production, margins are going to shrink as there becomes very little to differentiate products and these mis-steps can risk the business.
There’s a wide chasm between Tesla and generic EVs with respect to manufacturing processes, battery technology (especially range), software integration, charging network, and so on. Tesla’s mote is formidable and will endure for the foreseeable future irrespective of level 5 FSD.
Have automatic updates for your personal machine. If Windows pushes out a patch that bricks your computer, I'm not impacted. If Tesla pushes out an OTA that is buggy and causes accidents, I'm affected.
0: https://teslamotorsclub.com/tmc/threads/opt-out-on-updates.1...
Anything to stop bad companies from sending out half-baked patches is a net good. I don't care if you have to roll back an update on your SaaS site. Sucks, but we'll move on. But continuous integration should stay far away from what amount to missiles that weigh tons and go in excess of 60 miles.
It doesn't do that. Tesla would still be able to push out whatever updates it wants. Drivers would just have to either go to a service center or leave the state to install them.
Currently, with OTA patches, you can throw out any patch you want because it's self-serve. But if customers were inconvenienced by having to go to a service center for an update, Tesla would have a larger incentive to get things right the first time. It would damage their brand reputation if a customer had to go to a service center for a patch and then turn around and schedule two more appointments to download bugfix patches.
> The company would have an incentive to rigorously test the patches before sending them out.
This doesn't stop traditional automotive companies from releasing buggy software (they have far more recalls per unit time than Tesla), and anyway the issue you noted above wasn't a bug, it was a perfectly safe _optional_ feature that NHTSA didn't like.
Tesla service centers are not a "new motor vehicle dealer", you would need to leave the state to install updates
> Except for experimental low-volume not-for-retail sale vehicles, cause warranty and recall repair work to be performed by any entity other than a new motor vehicle dealer
But aggressive autopilot being evil? That's a stretch.
Sure human operators need to stop every time, look both ways, proceed cautiously. Makes sense.
But a car with sensors operating 24x7, never getting tired or taking an intersection for granted? The rules can be different. Should be different. Because there's no added risk; because we can adjust them should an issue arrive, and they will always, always do it the new way.
Computers are different from humans. They drive differently. Insisting they behave the same way, is like insisting drivers have a person on foot proceeding ahead honking a horn and waving a flag (as early localities insisted drivers of the new automobile should have to do) because they might scare the horses.
Until then, when a semi-automated car comes to a stop sign, they should stop.
I'll send you a horn and flag in the post, for the next drive.
If you are in a state like Massachusetts where a rolling stop is the norm, then for the same reason it's best if the car does not come to a complete stop, regardless of whether someone is there, because it makes your vehicle's motion more predictable. Ideally laws in those states should also be adjusted to reflect what people actually do.
Now, if a cop saw you do a rolling stop because you thought there was no one there, he could stop you (in KS anyways).
In SC, the law requires a stop (making no distinction between a full stop and a rolling stop) at a stop sign or red stop light, prior to the intersection, either at the stop line or before the near side of the crosswalk. AFAIK the only provision we have for a rolling stop is for emergency vehicles.
Rolling stops are illegal because they're dangerous. If a computer malfunctions, runs a stop sign and t-bones me, I could very well be paralyzed given the multiple structural issues with my spine.
My bodily health is more than an 'issue' which can be coded away. Because Tesla can fix their code, but that code fix isn't going to fix my back.
But isn't this discussion about whether rolling stops are okay when nobody else is around you?
I see nothing in parent/grandparent/ancestors of my comment qualifying whether or not other cars are at the intersection at the time of the rolling stop. The original comment that spawned this discussion was "If it keeps Teslas with 'aggressive' autopilot from doing illegal and dangerous things like rolling stops, I don't care what concern it stems from."
There is a another thread on which is definitely about whether rolling stop are okay when nobody else is at the intersection (note this is subtly different from whether nobody else is around, especially for a two-way stop instead of a four-way).
> If a computer malfunctions, runs a stop sign and t-bones me, I could very well be paralyzed given the multiple structural issues with my spine.
How are these two statements in any way related? A computer malfunction can cause an automated car to run a stop sign and t-bone you whether or not that car is programmed to slow down to 0mph or 5mph.
You're stuck thinking about the way that humans make mistakes, computers can still make mistakes, but they're of a completely different nature.
Consider a thought exercise - In your opinion which of these is a better way for an automated car to approach a stop sign?
1) Slow down to 5mph as it approaches the intersection and evaluate if it's safe to continue; if it is the car may continue without coming to a complete stop. The acceleration profile is designed so that the car has 15 seconds total to evaluate the situation before continuing through the intersection.
2) Slow down to a stop at the intersection and immediately continue if it's deemed to be safe. The acceleration profile is designed so that the car has 10 seconds total to evaluate the situation before continuing through the intersection.
If we were talking about a human there is certainly an argument that option 2 is safer since human are generally bad at multitasking and requiring a period of time dedicated to looking for hazards is a good idea. However, a computer does not have the same issue and option 1 could be considered safer due to the increased time spent observing for hazards.
Rolling stops are illegal for a person, because a person suffers from inattentiveness and has slow reactions. The rules for a computer start from a different place, and likely will end up different from the rules for people.
Traffic is never going to be completely safe. We accept that now. To hold autodriving to a higher standard is reasonable, but there's a limit. And it should be a rational limit, not just hyperbole.
Strawmen like 'I'll be maimed!' are not really helping the conversation.
If the computer malfunctions then the software telling it to come to a complete stop at a stop sign isn't going to save your bacon. You're t-boned either way.
> Sure human operators need to stop every time, look both ways, proceed cautiously. Makes sense. But a car with sensors operating 24x7, never getting tired or taking an intersection for granted? The rules can be different. Should be different.
And frankly, in the case of Tesla autopilot, the vehicle is not autonomous, and the responsible party for the motion of the vehicle is still the operator, regardless of how much the computers are doing on their behalf. Under these circumstances, correct observation of traffic laws is mandatory.
With my phone I can choose to ignore updates. Anyone who has ever owned a computer knows that an update can make something worse. In fact today I was in a discussion about rolling back a BIOS patch. There should be an easy way to opt out of you're not interested.
Tens of thousands of people are killed annually in auto accidents where 94-96% are caused by human error. Twelve deaths, total (as of this comment), are directly attributable to Autopilot being active.
I would suggest you reconsider your approach to risk management, as it seems to be emotion vs data driven.
https://www.tesladeaths.com/
While I think autopilot type things can be much better than humans, so far the data I've seen doesn't make it possible for me to make a fair judgement on how good/bad it is.
In any case the point remains: you cannot just roll out updates to something safety critical like that. You need a lot of careful work to ensure the update is better than the old in all ways, and that nothing can go wrong.
I'm guessing that high drivers account for at most a sliver of incidents, but moreover, I don't see the purpose in comparing autopilot to only the best drivers. If someone who would otherwise drive high uses autopilot, isn't that an improvement? If someone who would otherwise text while driving uses autopilot, isn't that an improvement? If someone who is just a worse driver than autopilot uses autopilot, isn't that an improvement?
I already know humans are bad drivers, but autopilot is a big unknown.
> autopilot is a big unknown
We really just need to know whether Teslas get into more accidents or more severe accidents than other vehicles (modulo some adjusting for demographic differences among drivers). If Teslas contribute to fewer/less-severe accidents, then it demonstrates that they're safer overall. Whether they get there via autopilot or other means is an implementation detail. In any case, Tesla isn't just going to ship unsafe software because it would destroy their brand (every Tesla incident gets 1000x more scrutiny than other brands).
Oh no no no, this is not accurate at all. I’ve used Autopilot over 60k miles across the entire US, even in the worst of weather through steep mountain grades at night at speeds up to 85mph (both the Rockies and TN/KY/NC). I disagree with your assessment. It has occasional disengagements during challenging scenes (road lines disappear suddenly, heavy construction causing sharp lane changes) but is an extremely robust system overall.
Your average human’s driving ability is not so great, in comparison.
Humans don't have them very often. We might get things wrong, but we don't disengage in the same way. (humans do sometimes die at the wheel of course, and in general are not good drivers)
Of course throwing anything to humans only will skew the data against humans.
If 20,000 people (made up number) die in a year from human error, and that is worthy of concern, isn't it worse if 20,012 die?
There are certainly more than 100 million people who did not die since Autopilot came out. Neither you or I can prove it didn't save 100 million people. Therefore if it were to kill 12 million people, that's a net 88 million fewer deaths. What is wrong with that logic?
I need to be educated on where the linkage between lives cost and lives saved is.
https://www.epa.gov/environmental-economics/mortality-risk-v...
https://en.wikipedia.org/wiki/Value_of_life
https://www.americanprogress.org/article/putting-a-price-on-...
https://www.npr.org/transcripts/835571843
https://www.nytimes.com/2020/05/11/upshot/virus-price-human-...
The education I need is on how people know that something in Autopilot saves X lives at the cost of Y lives. What, in the physical world, does that? Where does the premise of a tradeoff come from?
I know, right? What gives Tesla the right to sell cars that don't pump poison into the atmosphere, or which improve road safety? *shakes fist*
[1] https://www.thedrive.com/news/43329/toyota-made-its-key-fob-...
I simply want a toggle, at least software and preferably also hardware (e.g., being able to pull the SIM).
The less Phone Home stuff in my car (and its black box / data recorder), the better.
I agree this smells of being anti-competitive against any manufacturer lacking in brick-and-mortar dealerships (e.g. Tesla).
Try to buy any of these new Ford EVs and be ready for a nightmare of sketchy pre-orders and dealer markups all while being treated like dirt while buying a $70k car.
By all means take the touchscreen and LTE radio out, they mean nothing to me.
I mean, this is true right now, and so those things (and hospitals and governments) are being targeted right now. Lots of stories about ransomware recently, and allegedly it's part of the hostilities in Ukraine.
When there is enough centralized infrastructure controlling enough cars, then targeting it will do more damage, so it would be logical to do more of it.