More than 2 million Teslas are being recalled due to unsafe Autopilot

Inadequate driver monitoring means that Autopilot is far too easily misused.

See full article...

 

[ScottTFrazer]

It's weird that you're so willing to dismiss the lived experience of others. Just because something works acceptably for you doesn't mean it's well engineered or safe. In this case, we've got science that says it's not.

His reasons also are sound in my opinion. One of the scroll wheels changes the speed that the car is moving, the other changes the volume of the radio. Both are suboptimal for a system simply designed to show you're paying attention to the road. If I experienced those issues, I'd turn off the feature as well.

It's not going to let you drive into the car in front of you, it will prevent you from accidentally leaving the lane,

This is a thing we have documented proof that it does, in fact, do. https://www.washingtonpost.com/technology/interactive/2023/tesla-autopilot-crash-analysis/ is just one example, it's easy to google many more

 

[chalex]

I hate these arguments because I feel like I have to nitpick and people think I'm being too pedantic but for example when you say "One of the scroll wheels changes the speed that the car is moving", no that's not what it does, it adjusts the maximum speed setting and most commonly (IME) the car is not going the max speed because of traffic. And the whole argument is predicated on a person who is somehow not compatible with using a wheel torque sensor (which is not a Tesla-only thing). So that person does have other options that are easier (for them) to use.

But anyway, none of that is changing in this recall anyway, so it's a side argument.

 

[ScottTFrazer]

I hate these arguments because I feel like I have to nitpick and people think I'm being too pedantic but for example when you say "One of the scroll wheels changes the speed that the car is moving", no that's not what it does, it adjusts the maximum speed setting and most commonly (IME) the car is not going the max speed because of traffic. And the whole argument is predicated on a person who is somehow not compatible with using a wheel torque sensor (which is not a Tesla-only thing). So that person does have other options that are easier (for them) to use.

But anyway, none of that is changing in this recall anyway, so it's a side argument.

Then, I don’t know, stop being nitpicky? You just said “most commonly” which doesn’t mean never, so the point stands. Why is it so important to you that this be acknowledged as “it’s fine, really” instead of being an issue?

Is it just because you own the car and you like it? Cool! I’m glad you like your car, but not everyone else appears to have that same opinion.

But when you say there isn’t a problem with the system, it ignores a large amount of data to the contrary. Why are you nitpicking things in order to ignore that?

 

[cbreak]

I would say that's ultimately a good thing. If this issue was only being treated as a simple software update, the vast majority of people wouldn't think twice about it (or even ignore it). By the NHTSA using their legal power to issue an official recall, even if it causes some confusion with people, it brings the issue to the forefront and makes Tesla owners actively seek out information on what this recall entails.

The fact that it is just a software update does not change the fact that this is an official, legally enforceable, recall.

Also, the software update is probably applied via some form of built-in networking, such as a cellphone network modem. So, tesla is calling the tesla and applying the update via that "call". If it's not the first update, it's a re-"call"

 

[numerobis]

Then, I don’t know, stop being nitpicky? You just said “most commonly” which doesn’t mean never, so the point stands. Why is it so important to you that this be acknowledged as “it’s fine, really” instead of being an issue?

Is it just because you own the car and you like it? Cool! I’m glad you like your car, but not everyone else appears to have that same opinion.

But when you say there isn’t a problem with the system, it ignores a large amount of data to the contrary. Why are you nitpicking things in order to ignore that?

Personally I own the car and I like it and I wish it would be better, e.g. by having actual driver monitoring rather than bullshit nagging.

I don't understand people who get defensive about the car they own being imperfect.

 

[stevenjklein]

how will you separate the possible factors of "autopilot sucks ass" (extremely true) and "reckless dudebro types gravitate towards these cars" to explain any variance?

If “reckless dudebro types” make up a disproportionate number if Tesla owners, that might explain why Teslas have higher accident rate.

(Will autopilot drive faster than the posted speed limit?)

If Teslas have fewer accidents in spite of the “reckless dudebro types”, how else would that be explained?

Note: Since I lack data to form the basis of an opinion, I’m not pro or anti self-driving in general, or Tesla’s implementation in particular.

I’m genuinely puzzled by the massive downvote my post received.

Since you took the time to post a reply, perhaps you can shed some light on this. What am I missing?

 

[ScottTFrazer]

If “reckless dudebro types” make up a disproportionate number if Tesla owners, that might explain why Teslas have higher accident rate.

(Will autopilot drive faster than the posted speed limit?)

If Teslas have fewer accidents in spite of the “reckless dudebro types”, how else would that be explained?

Note: Since I lack data to form the basis of an opinion, I’m not pro or anti self-driving in general, or Tesla’s implementation in particular.

I’m genuinely puzzled by the massive downvote my post received.

Since you took the time to post a reply, perhaps you can shed some light on this. What am I missing?

Check out this article. Author goes into a good bit of detail about how difficult it is to compare accident statistics and especially how Tesla is muddying the waters with the stats they do release:

https://www.forbes.com/sites/bradte...-story-with-their-crash-data/?sh=59f6f560feda
As for the dude-bro hypothesis?

Teslas are expensive cars and are bought primarily by older, wealthier people. Middle-aged wealthy people are in fact the safest class of drivers. Any study of them, compared to the general population, would show the car driven by the safer drivers with a lower crash rate — but it’s not the car, it’s the drivers.

And there's good bits in there about the stuff I've been posting in the last couple pages: how some autonomous systems can lull the driver into a false sense of security.

 

[Snarky Robot]

I’m genuinely puzzled by the massive downvote my post received. […] What am I missing?

Purely guessing, but I’d say at least some are because accidents per mile is an absolutely fucking terrible metric.

I assume you’re aware that San Francisco and Saint Cloud have different temperatures. As do Dallas and Detroit. Miami and Minneapolis. And so on. So, a mile in a place with lots of warm, sunny days is not the same as a place with snow, ice, or rain. Pretty obvious that one will have more accidents, and because Teslas sales are disproportionately in warmer climates, you’d have to control for weather.

Teslas are also still pretty expensive overall. So that means you have to account for wealth (which is, sadly, a factor in accidents). Wealth is also a fairly reasonable proxy for age,(as they tend to have less disposable income). Younger people tend to have more accidents. Also connected to wealth, and the relatively recency of Teslas become a massive seller, you have fewer used cars on the road. Older cars have more issues, meaning more accidents.

There are many, many more factors that would go into an actual comparison. Raw accidents per mile is a preposterous foolish argument to make because it takes so little of reality into account.

So, a decent start (buy still not reasonably accurate) would require you to account for the age of cars (i.e., you can only count a cars under a given age, and still have to weight for proportion of cars that are used), age of drivers, economic status, and location. Once you account for those factors, you’re getting closer to a real result. You’re at least in the ballpark. Raw accidents per mile isn’t even in same hemisphere as a coherent analysis.

Though others may have a different reason, that’s likely one reason people are not enamored with your response.

 

[Delerious]

I agree that the word “recall” is correctly used in this instance, but it’s also fair to say that it’s a word whose meaning has been distorted by circumstance. It was initially used to because it‘s meaning applied exactly (you had to bring the thing back), and now it still gets used but doesn’t mean that at all. Kind of like how people still refer to a band’s “albums” despite the age of digital.

What is a collection of songs issued as a single item supposed to be called if it's not an album?

 

[cbreak]

What is a collection of songs issued as a single item supposed to be called if it's not an album?

A single?

A folder of mp3s? A zip file? An audio-bundle?

 

[Jim Salter]

If Teslas have fewer accidents in spite of the “reckless dudebro types”, how else would that be explained?

There would be any number of possible logical explanations for that outcome. But let's save that particular speculation for after a negative correlation is discovered between Tesla ownership and accident rates, controlled specifically to rule out anomalies caused by age or income distribution.

Because I don't think you're going to find anything of the sort.

 

[Jim Salter]

What is a collection of songs issued as a single item supposed to be called if it's not an album?

That's not the point: the point is that artists are increasingly less likely to release cohesive albums in the first place. Instead, they are increasingly likely to just release singles.

With listening technology that makes it trivial for anyone to assemble a playlist from a ton of sources, you don't need an hour or so of content from any individual source so that you can listen for a significant amount of time without having to break your immersion to mess with the playback device.

The album isn't exactly dead yet, but it's not the overwhelming market expectation that it used to be, either.

 

[Delerious]

That's not the point: the point is that artists are increasingly less likely to release cohesive albums in the first place. Instead, they are increasingly likely to just release singles.

So the collection of singles below isn't really an album even though they call it a digital album (and it's over an hour)?

View: https://doomedandstonedlatino.bandcamp.com/album/latinoamerica-2023

 

[Jim Salter]

So the collection of singles below isn't really an album even though they call it a digital album (and it's over an hour)?

View: https://doomedandstonedlatino.bandcamp.com/album/latinoamerica-2023

Of course that's an album. What point are you trying to make?

 

[dollyllama]

Elon thanks Ars for not once mentioning Fully Self Driving isn't.

Rolling Stone has the cajones to report it.
Elon Musk’s Big Lie About Tesla Is Finally Exposed

 

[Cthel]

What is a collection of songs issued as a single item supposed to be called if it's not an album?

If it's issued on a single piece of media, is it really an album, or just a sparkling compilation?

Spoiler

78rpm albums were multiple discs in a literal album

 

[Snarky Robot]

What is a collection of songs issued as a single item supposed to be called if it's not an album?

It’ll be interesting to see if we keep getting “albums”, or if there will be a shift towards small releases as streaming continues to be the dominant form. As contracts get renegotiated because of streaming, there may be a tipping point where you could do 4 (or fewer) songs as a batch, and release longer stuff you can buy as set lists from shows.

I’m not sure how much I like the notion of “albums” anymore. It feels like an obligation to listen to all of them, when what I’d like is more just “mix a new song into my playlists”.

 

[numerobis]

Rolling Stone has the cajones to report it.
Elon Musk’s Big Lie About Tesla Is Finally Exposed

What’s your issue exactly? Ars has reported on FSD being crap. That just isn’t the topic of this article, which is about a NHTSA recall notice on Autosteer, which isn’t FSD.

 

[wasbee56]

With the limitations of current technology on display, at some point you really might as well just drive the car yourself instead of gripping the wheel in a tense state of catlike readiness, waiting for your car to betray you at any moment.

IMO if I have to oversee the driving, I would just prefer to do the driving

 

[dollyllama]

What’s your issue exactly? Ars has reported on FSD being crap. That just isn’t the topic of this article, which is about a NHTSA recall notice on Autosteer, which isn’t FSD.

It's about reporting important information that's relevant to the recall, which Ars didn't fully do and I merely pointed out. The recall is about drivers not maintaining vigilance in the face of automation that doesn't work well enough [to be called full self-driving]. The matter would've fit well in the article without feeling pedantic. However, it seems I have to be pedantic about it in the face of Ars apologists and perhaps Tesla investor apologists who'd like to bury criticism of FSD, too. While we're on the subject of Tesla autopilot, I'll add that the human factor isn't the only part of the equation that's lacking. I personally experienced phantom braking that nearly caused a rear-ender and did cause a knee injury in supporting myself. There are apparently many reports of P.B. and not just with Teslas.

 

[numerobis]

It's about reporting important information that's relevant to the recall, which Ars didn't fully do and I merely pointed out. The recall is about drivers not maintaining vigilance in the face of automation that doesn't work well enough [to be called full self-driving]. The matter would've fit well in the article without feeling pedantic. However, it seems I have to be pedantic about it in the face of Ars apologists and perhaps Tesla investor apologists who'd like to bury criticism of FSD, too. While we're on the subject of Tesla autopilot, I'll add that the human factor isn't the only part of the equation that's lacking. I personally experienced phantom braking that nearly caused a rear-ender and did cause a knee injury in supporting myself. There are apparently many reports of P.B. and not just with Teslas.

Apparently it also caused some brain damage?

News articles are about a particular piece of news. They aren't about whatever random vaguely related things also are true. Gitlin has written plenty of articles about phantom braking and about FSD.

 

[dollyllama]

Apparently it also caused some brain damage?

News articles are about a particular piece of news. They aren't about whatever random vaguely related things also are true. Gitlin has written plenty of articles about phantom braking and about FSD.

You lack sense and sensibility.

 

[icrf]

Irrelevant. We already have technology that tracks eye movement. The torque sensor is a boondoggle.

Even if it worked perfectly, which would you prefer: a "driver" napping with their hands on the wheel, or a driver with their hands off the wheel, but their eyes firmly on the road?

Agreed, but that could be done with a software update for existing cars that will likely never get additional hardware. I'm trying to not let perfect be the enemy of the good here, unless perfect is just removing ADAS entirely for all those vehicles.

 

[ScottTFrazer]

Agreed, but that could be done with a software update for existing cars that will likely never get additional hardware. I'm trying to not let perfect be the enemy of the good here, unless perfect is just removing ADAS entirely for all those vehicles.

I'm thinking it could devolve into Tesla making the experience objectively worse for the driver, the NHTSA saying it's still not enough. Repeat cycle until it's just a constant alarm until the driver turns Autopilot off

 

[numerobis]

I'm thinking it could devolve into Tesla making the experience objectively worse for the driver, the NHTSA saying it's still not enough. Repeat cycle until it's just a constant alarm until the driver turns Autopilot off

You say that like it's a bad thing.

 

[ScottTFrazer]

You say that like it's a bad thing.

Heck, no If the cars don't possess adequate tech to gauge driver attention, having the driver turn the feature off is absolutely the safest result for the driver and anyone in a 100 yard radius of them

 

[Pooga]

I agree that the word “recall” is correctly used in this instance, but it’s also fair to say that it’s a word whose meaning has been distorted by circumstance. It was initially used to because it‘s meaning applied exactly (you had to bring the thing back), and now it still gets used but doesn’t mean that at all. Kind of like how people still refer to a band’s “albums” despite the age of digital.

If it's issued on a single piece of media, is it really an album, or just a sparkling compilation?

Spoiler

78rpm albums were multiple discs in a literal album

As soon as I saw the digital album comment I was anticipating this particular response. Music albums have been something of a misnomer since we stopped releasing disc collections on 78.

Do recalls come with a fine for the automaker? Recalls requiring parts and labor constitute a huge penalty for the automaker, and hence a strong deterrent against shipping dangerous vehicles. Recalls involving only software have much less deterrent value on their own, and could be shrugged off as cheaper than doing adequate safety reviews or testing.

Gotta love how you never know how your car will behave on a given day, since there might have been an update..

I am, as a professional and very infosec-aware senior sysadmin, VERY focused on applying updates, not sandbagging them. But God damn it's tempting to sandbag the OTA Tesla updates, because in the couple years I've had the car, they've added one feature I actually liked, fixed zero bugs that affected me, and markedly decreased my satisfaction with the interface while introducing less reliability.

When I bought the car, emergency autosteer did not try to pancake me into heavy trucks on Huger St. After a particular update this year, it began trying to hijack control and force me into the left lane whether it was occupied or not, EVERY time I drive that street past a particular intersection. There's nothing there I can even identify as likely to be problematic--no very heavy shadow, jut-out from the curb, etc. I had to disable emergency autosteer entirely after that.

So, yeah, I still apply the damn things. But I have a very uneasy feeling about it every time now.

All of these kind of tie into something I haven't really seen brought up here - allowing system-critical software and especially software that affects the actual vehicle safety to be updated OTA seems like a Bad Idea. My reasoning is two-fold.

First, IMO the lesser threat: We've had examples where even companies whose product is nominally security-related software (e.g., LastPass and Okta) have had their networks compromised. It's no stretch to imagine that something similar could happen to a company that is primarily a vehicle manufacturer. I'm not saying it would be easy for someone to gain the access to send out a malicious update (or spoof the vehicle that an update is coming from the manufacturer when it isn't), but it's not a situation I particularly want my vehicle to be in.

Second, we've got a few decades of examples now of what happens to software development and release cycles when things that were previously difficult to update after release became easy to patch over the Internet. In almost every category of software for devices that were originally standalone but now have Internet connections we see software pushed out early, poorly tested and/or buggy to meet goals that have more to do with company profits than reliability and stability.

I know my 2017 Bolt EV has the ability to connect to my WiFi and a cellular connection for OnStar. I think the only software that can be updated through one of these connections is the infotainment system. That is, I've received one update from the Internet over WiFi but also had to bring my car to the dealer for a software-based recall. While this is clearly the more arduous route for me, I feel it's the better option for the software stack that affect the car's controls and safety features. It puts more pressure on the manufacturer to get it right the first time or face costly recalls that can't just be pushed out overnight. It lessens the chance that some newbie developer's flaky code accidentally bricks the important bits of my car. Plus it makes it unlikely that the control layout will be rearranged at the whim of someone at the company, forcing me to relearn how to do things after an update.

 

[dkazaz]

After much consideration and driving a Tesla for 3 years (which I quite like) I have conluded that Tesla has a lot of great ideas for automobile innovations, that are, sadly, so poorly implemented that the best you can say, is that hopegully other more serious auto makers will copy them, giving us better cars in the process.