California DMV accuses Tesla of making false claims about Autopilot and FSD
theverge.com
theverge.com
Then Musk came along and completely ignored the research (which he must've been really aware about considering Google was in negotiations to buy Tesla) and even took the advertising a step further. He KNEW people wouldn't pay attention but sold the cars with that marketing anyways. IMO the blood of the 15 people who have died due to autopilot is directly on his hands as well as the thousands of other injuries due to autopilot accidents
Tesla's still have problems. They made up 273 accidents involving driver-assistance systems in 2021. 70% of all such accidents.
No, because it hasn't been demonstrated to even work as promised, let alone "isn't perfect".
That data is also super misleading as they compare Autopilot highway crash rates with statistics for an average car (which is older and has fewer safety features than newer cars) that includes city driving (which is much less safe than highway driving). It's designed to advertise Autopilot as safer than it really is. In reality, it's about as safe as any new car with standard safety features (AEB, collision avoidance, lane keep etc).
Teslas drive like drunk drivers, whether using AP or FSD.
Tesla fans might feel like they are safer, but from the perspective of someone else on the road, they are objectively less safe, unless you consider swerving, randomly slamming on the brakes, and tailgating to be examples of safe driving.
And Tesla's AP system has more recorded fatalities than all of their competitors, combined. By the numbers, NHTSA data shows that Teslas are 50x more likely to get into an accident using AP or FSD than a Ford or Toyota; and the nearest legacy automaker is still 1/3rd as likely to get into an accident as a Tesla using AP/FSD.
Is that a AP/FSD thing? I ask only because just this weekend there was a Tesla on my rear bumper in a 30mph zone, and I had the cruise set to 34mph. I did wonder whether it was the computer, or if the driver was just an ass. Regardless, after a number of miles I grew tired of it, let's give 'em a healthy brake tap. Just enough to get the hint, I don't want anyone spilling coffee on themselves. So I give the brakes a good jab, and immediately back off.
Next thing I know, I hear the tire chirp of ABS kicking in, and I look in the rearview to see the Tesla stopped in the middle of the road. Oops, sorry (not sorry?), not exactly what I was going for. OTOH, maybe you or the car will quit riding people's asses now. And I'll admit I got a good chuckle out of it.
IF we would have the number of how many times the driver would have been hurt if he did not take control then we could have a better guess how many kilometers a Tesla would self drive until would kill or hurt you,
Perfection is not a requirement for advancement.
Conclusion FSD is better then humans.
Nobody said that installing a good drive assistant is less safe, the issue is with idiots using this stats to imply the AI itself is better then human drivers wich is a bad conclusion to draw from this.
To prove FSD is better then humans, send a Tesla from A to B and note hoe many kilometers until it screws up. Repeat 1000 times then do the same experiment with 1000 humans. Don't forget to repeat after each update.
Go to Tesla subbreddit are all hard core fans are and read about this topic, you will see many example of users having issues with this feature, the FSD is not safe, it will kill you if you don't pay attention, it might kill some innocent person because some random dude is alpha testing updates and maybe is also live streaming it for fun.
I REPEAT, Tesla is not sharing the data so if someone is telling you that the AI alone is better then a human then he is lying, there is no data to show this.
I would bet Autopilot/FSD is unsafe alone(without the human ) and the issue then raises because of marketing where someone say rents a Tesla, he believes Elon and does not pay attention and kills someone. I seen even Tesla fanbos acknowledging that FSD is not safe and just read a comment where one such fanboy had to explain his relative that FSD is a scam and not to buy it. Also there is a post ATM on Tesla reddit about someone that does not trust his relative driving his car with the autopilot because this person is failing to pay attention when using the AI, the dude was asking how could he make it impossible for this relative that shares the car with to use the AI.
No they are. They won't break out FSD and they 100% could
> their autopilot data suggests it's safer than human pilots on highways
Their autopilot data compared to every single type of car, with every single type of driver, only on highways.
So we're comparing every 1994 Corolla with a teenage driver to every 2021 EV with a 30 yr old driver.
The reality is AP should really be compared to modern cars which tend to come with LKAS and AEBm and stats should be adjusted to reflect the differing driver populations.
But Tesla would rather throw up comparisons that make themselves look good, but only past muster if you don't think about it too much.
Is there new data I missed? Last I was aware Telsa had published data comparing autopilot to human drivers, but not adjusting for road type, weather, road condition, construction, and any number of other factors in addition to the ones you mentioned.
So basically Telsa determined that autopilot on easy-mode driving (highway, clear enough conditions for autopilot to be willing to operate), with a human backup, has less crashes per-mile than humans in the full range of conditions (including those where autopilot will opt out) with no backup.
If that's not an example of "lies, damn lies, and statistics" I don't know what is.
The narrow section excludes older vehicles, skews away from the most at risk demographics of drivers, removes the toughest highway situations from the equation since AP will disengage, the list goes on.
Also: you already have autopilot on the freeway: it's called cruise control + lane assist. That's not as sexy as saying autopilot though.
https://www.forbes.com/sites/bradtempleton/2020/10/28/new-te...
https://medium.com/@MidwesternHedgi/teslas-driver-fatality-r...
If you question was if self driving is safer then humans the answers sr probably no, if it were even close Tesla would have shown the numbers not kepp them hidden.
And that doesn't include the thousands of accidents related to Tesla's AP/FSD systems.
Tesla's advanced driving systems are objectively less safe than teenage or drunk drivers, by miles driven, and the sooner AP/FSD is banned, the better.
Google had hands-free driving working on highways a decade ago, but decided they didn't want to take the incremental approach of L2 -> L3 -> L4 exactly for the reasons you mentioned and jumped straight to making L4 (fully driverless) work.
Mercedes-Benz had hands-free driving working on highways in 1987 (before Google even existed). However, their approach relied on expensive specialized hardware and wasn't commercially practical.
It's up to 21 now, including the 2 fatal motorcycle accidents referred to in the Ars Technica article from this weekend.
After watching thorough video reviews, the differences between the best systems are fairly minor.
Making it all about Tesla will do the industry a real disservice. Thankfully NHTSA seems to be aware of the issues and is taking their time to understand fully before making adjustments to regulations.
Did that happen? I thought that was just a private thing between Musk and Brin.
> Tesla's still have problems. They made up 273 accidents involving driver-assistance systems in 2021. 70% of all such accidents.
Is that broken down anywhere in terms of actual miles, and even more importantly high way miles compared to non high way miles.
Even without any driving assistance, people will go in their their phones and have lapses of attention.
The data shows that people are more irresponsible with the autopilot technology. Does this fault fall on the drivers or manufacturer.
Tesla chose to go with "autopilot" because they sell people what they wanna hear. It's irresponsible and unsafe
> This is another version of the "guns don't kill people, people do" argument.
And no actually this is quite the opposite argument.
> The data shows that people are more irresponsible with the autopilot technology. Does this fault fall on the drivers or manufacturer.
Just like the fact that tools are specifically built in a way that makes them much more likely to be used to kill people, the fact that Tesla's are made and marketed the way they are makes them much more likely to be lead to people being irresponsible
I think you got the arguments backwards buddy. you're exactly the one who's making a version of the "guns don't kill people, people do" argument
The question of if Tesla sells a more dangerous product is different from the false advertising claim. Both could be true, false, or mixed.
I personally think FSD is false advertising but the name doesn't make the product anymore dangerous then if it would have been if called "super cruise", "driver assist", or whatever with identical features.
The features themselves may also be more dangerous, but the question of if that puts blood on their hands is the gun debate all over again.
Fast cars and cheap cars are also more dangerous, ect
It's hard to see ambiguity about "Full Self Driving". What's next, will they argue that "The Car Drives Itself, You Can Take a Nap" isn't meant to be taken literally?
The car can in no way fully ‘drive’ itself unless you define ‘driving’ as not including many things that are universally thought of as necessary skills to operate a car.
[edit: rereading your comment, I see we’re actually in agreement!]
https://www.tesla.com/autopilot
That paired with the name of "Full Self Driving" really does a good job of implying that the car drivers itself...
Whether Tesla will actually achieve FSD is debatable. Personally, I think they will. Yes, even with vision-only. It's going to be several more years, but I think they'll get there eventually.
But what do you call it when you're selling that feature to customers today?
Being able to answer that confidently requires some understanding of the problem domain which not everyone has. This is especially true for these big claims where people might think that a large company wouldn’t rationally make claims like this which would expose them to liability unless it really was almost done, not realizing the degree to which Elon will in fact take that risk just to cast himself as the hero of a 90s sci-fi novel’s back story.
Not really.
All it really takes is a brief look at history to determine that Elon Musk has been saying FSD is less than 6 months away for well over 5 years. Some basic pattern matching skill should instill the idea that FSD is not actually coming anytime soon, and that it's all marketing bullshit.
I want to agree with you on this, but Tesla is very much bordering on false advertising.
The car was fun to drive, but the build quality and service were terrible. And the continuous upsell to FSD, even 8 years ago, was such a hassle.
The parent isn't suggesting looking at the ratio of happy and sad posters, which would definitely be selection bias, they are saying look at the rate and severity of complaints on it's own.
Of course there are defects, at that volume that would be expected.
Parent didn't say anything about rates.
The cars themselves have more range and charge faster, but its not an order of magnitude. Maybe charge times on a road trip have dropped by ~50%.
Service has generally gotten worse as well, as the company has really struggled to scale. They've added a lot of mobile service and service centers, but they still can't keep up with the pace of growth. Good thing is that most of the newer cars aren't needing service much, but when they do it can be problematic.
You're definitely not alone
The FSD name makes more sense for the software they've got in public beta than their current offering. Interestingly the only difference between "enhanced autopilot" and FSD right now is the ability to stop at stop lights.
Also if you fly like a maniac you’ll get your license taken away, which is very much not the case for driving.
FSD shouldn't be allowed. EAP is a decent name for what it actually does.
Comparing a pilot to a regular car driver is just completely off point. You are basically throwing out all the training pilots have to do just to make your point.
Pilot use technical terms that they learn in both training but also in the culture around pilots. Words used by pilots in that context can have a different meaning than the literal interpretation of those word outside that context by non-pilots. The jargon is different, and thus advertisement towards pilots using that jargon will be interpreted in that context. The word "autopilot" directed towards pilots will be interpreted in that context, meaning they will expect the technology to work as training and culture has made them to expect.
Driver training is very different from pilot training, and the jargon and lingo is different from pilots. An car with autopilot that is limited to work with the restrictions of a airplane autopilot, ie only work at airport or while high in the air, would likely not be what the driver expected. The same word but different context.
What I think is more important is understanding what the general population of buyers of these cars think it means. Whether they are wrong or right, what they think it means effects how they drive with autopilot on, and whether they are looking ahead or on their phone.
The only real way to know this would be to measure it, and I doubt that is knowledge that Tesla wants to know.
As a side note, the Boeing pilot will spend a great deal of time looking at charts or their company laptop once they are at cruise with their auto-pilot on.
You just don’t hear Toyota calling their self-driving “Toyota Autosteer.”
This all seems to be driven by arrogance and marketing.
NHTSA is making headway, but its going to take time to reach accurate conclusions.