Weren’t they also found to be disengaging so/fsd seconds prior to impact and reporting it as driver error (because of their BS “driver must always be ready to take control”, with a more rapid response than we expect from trained pilots)
TBF, this is what level 2 and 3 are supposed to do. It's a shade more driver friendly than driving headlong into an obstacle and hoping the driver wrests control away from the computer.
Level 3 is supposed to be smart enough to disengage and ask the driver to drive. Level 2 is not required to be that smart, and most level 2 cars are not.
I thought FSD applied only to uncontrolled access surface roads, and the bay bridge would be simply autopilot.
I know that in my car, the FSD stuff ends and regular autopilot takes over on highways and other controlled access roads. I've never driven the bay bridge in my Tesla (since I live thousands of miles away), but I remember it being a controlled access highway, so I assume it was the AP logic, not the FSD logic that was in control.
Correct. On the highway, every tesla would be on the old 2016-born autopilot stack, not the FSD model. the only cars that have v11 fsd highway stack are internal tesla test cars.
I was curious about this and could not find an authoritative answer- it appears that there was a Lane Change feature in Autopilot that got moved to FSD, and autopilot shouldn't auto-change lanes at all. But I see lots of conflicting information. Some articles mention FSD, others just say "driver assist". I can't find the SCI on the NHTSA site.
I came back to re-read this post and since I'm unable to find any real corroborating information, I'm inclined to agree that it makes much more sense it was on Autopilot.
(I also haven't seen any commentary from musk, but it's hard to search twitter)
Much as Teslas navigate through "un-occluded space" I wonder if they do not sufficiently factor in their own Occlusion (blocking) of space.
What kind of idiot STOPS ON A DIME with speeding cars hot on his tail? Only a Computer...
It's unfortunate that the Driver did not have the sense to Slam on the Gas -The idea of "Slow == Safe" can be disastrous...
As an aside, I wish that Stoplights would factor in Aggregate Kinetic Energy of the Traffic they manage. I feel a bit bad every time my Single Stopped Car gets a Green, causing a group of 10 Cars at 40 mph to Stop, then later Re-Accelerate. If the stupid Junction had waited some time, the 10 Cars would happily carry their 40 mph through, allowing me to go once the Large Kinetic Energy Traffic had passed uninterrupted.
For simplicity lets assume all the cars in this discussion are the same mass and travel at the same speed except when they slow and stop for a traffic light or are accelerating back up to speed after the light changes. Let E be the kinetic energy of each car when it is not dealing with a traffic light. Let's assume no regenerative braking. Let's assume that a car idling at rest uses R energy
So we have 1 car that approaches the light, gets a red, and stops. It has to shed E kinetic energy. Then 10 cars approach the intersection on the perpendicular road. Before they reach it the light changes and they all have to stop, shedding 10 E kinetic energy. The first car resumes its journey. The light changes again and the other 10 resume.
The net result is that encountering the intersection caused this 11 car system to shed 11 E kinetic energy more and to consume 11 R idling energy more than would have been consumed if the intersection had not been there (e.g., if the crossing had been an overpass).
What he's proposing is that some sort of smart traffic light system could notice the 10 approaching cars and hold their green a little longer so that the 10 cars do not have to stop. Suppose it has to hold the light long enough that car 1 has to wait twice as long.
The energy cost of the intersection on the 11 car system under that scenario is E shed kinetic energy and 2 R idling energy. That's going to be much less that 11 E + 11 R.
> What kind of idiot STOPS ON A DIME with speeding cars hot on his tail? Only a Computer...
Humans do it on a pretty regular basis unfortunately. I have seen people slam on the brakes at highway speed for minor road debris, small animals, and a couple of times for no apparent reason. The only thing you can do as a driver is minimize your own risk by leaving sufficient following distance.
No offense, but what makes you the authority on all driving everywhere? Just because you haven't seen it doesn't mean it doesn't happen on a pretty regular basis.
Dolphins die on a pretty regular basis and I've never seen it. That doesn't mean it doesn't happen on a regular basis...
The follow distance thing really pisses me off. I was recently read ended as a result of the person behind me both not paying attention and having been way to close to me to stop in time. And since other drivers don’t understand the concept of a safe following distance they cut in front of me and now we’re practically bumper to bumper.
Ironically this is one of the areas where driver assist can be helpful. Adaptive cruise control that maintains following distance though in practice I don’t think people use it much.
All drivers should drive with the assumption the person in front of them will, at some point, instantly turn into a wall like a Wile E Coyote drawing of a mountainside cave.
Indeed. Transmission failures, bearing failures, tire blowouts, and throwing a rod all can cause a car on a freeway to stop quickly. The onus of avoiding this is entirely on the following car.
If the stopping car had been an old Toyota and we all had the same footage... well, it probably wouldn't even make the news in the first place... but if it did, we'd all be blaming the rear-enders. But because it's a Tesla, and there is a possibility that AI was involved, suddenly we shift our blame from the rear-enders and tailgaters to the computer.
We've always had double standards for humans and machines.
Unless the cohort of AI/FSD cars are sharing vehicle health/telemetry and know when another is “about” to have a blowout, I posit a rear end can still happen.
Yes brakes will apply sooner and follow distance is likely to be safer/wider.
I’ve always thought of it more like “will slam on their breaks” than “turn into a wall”. In most states, rearending is your fault unless the person in front of you stopped faster than is normal (i.e. was hit head on by an oncoming car), and I think that sets a good standard for what you should be prepared for.
If a tree falls on the car in front of you and you hit them because they stopped instantly, it’s hardly your fault.
Have you ever driven the 101 in the Bay Area? I’ve never seen so many purely shit drivers in my life. And I lived in Florida for years. People routinely merge across from their on-ramp to the express lane doing 5-10 mph under the speed limit, brake suddenly when the car a dozen or more car lengths ahead of them flashes their lights, etc. It’s a pure shitshow.
I don't know why Tesla etc are not required to report the number of crashes per 1000km in different modes publicly. Until those numbers are available none of us can really judge, least of all based on single incidents...
They have never published any data. Also, those number are also obviously misleading. Comparing my 2008 extended van I bought for $5k to a luxury vehicle is obviously pointless and useless
In addition, those "safety reports" continue to spout misinformation that the National Transportation Safety Board is currently suing Tesla for
Well it is a general principle of civility that we should believe people by default, unless we have good reason not to. I'm not sure what you're referring to when you say "track record" though.
Maybe you should look into why the National Transportation Safety Board sued Tesla for spreading misinformation
See also:
- The dangerous "we're almost at full self-driving" myths Musk continues to spread (which is probably part of the reason Google didn't acquire them given that they've been ahead on the research of this and knew how BS that is)
- The battery swap lies that they made a huge announcement about in order to take advantage of a California subsidy and then simply... never actually allowed any Tesla owner to swap batteries
- Tesla lying to shareholders about warranty expense by categorizing it as "goodwill" in order to inflate their gross margin and net income (since at least Q2 2018)
- On top of the California battery swap stuff, they've been sued for subsidy fraud in multiple countries including Canada and Germany. Much of it's "profitability" has simply been selling regulatory credits to other OEMs
- Musk using Tesla to bail out his cousin's failing solar company SolarCity and not disclosing the personal connection to shareholders
> it is a general principle of civility that we should believe people by default, unless we have good reason not to
That is appropriate in social situations, not in situations with significant consequences. If lives are at stake (or even lots of poeple's money), the burden of proof is on the person making the claim.
For example, that is true in the legal system, in science, and when you give a presentation to your boss. Saying 'you should believe me by default' doesn't go over well in those situations.
Well that number is made up, but yes I do have highly favourable feelings for that company, and for full disclosure, I am also a shareholder. I believe in their mission to accelerate the world's transition to sustainable energy. (Note: this does not mean I approve of everything Elon does and says)
Oh, give me a break. They were pointing out the absurdity of the position I was responding to with my tongue-in-cheek comment: anyone can say they did some analysis and produce “data”, but it’s hogwash if they don’t release the data that analysis was done on. Especially if the entity doing the analysis is a freaking company with a huge incentive to fudge the data, and a track record of dishonesty and hype bordering on fraud. It’s completely preposterous to claim with a straight face that Tesla has released safety “data” in any meaningful sense here.
I'm just holding both sets of data to the same standard proposed by GP. If there's a reason my question can't be answered, I'd love for someone to point it out
When you say "anecdotes", what are you referring to here? Because we're talking about the summary data published by Tesla. We're not talking about someone's story or personal experiences.
Luxury vs non luxury is always the first thing brought up with this report, but that's not the biggest issue IMO. Mainly it compares the easiest miles where people feel comfortable using Autopilot vs all miles. I wonder how many miles per accident there are with cruise control engaged.
the "new luxury vehicle" vs "all vehicles on the road" is relevant to Tesla's claimed overall safety of Tesla vehicles in general. But I agree wrt the Autopilot "safety" claims that the most relevant detail is the fact that highway crashes are already more than 2x less likely
At this point we still don't have any actual public evidence that driving assistance technology is even safer than driving without it. If Tesla's so confident in it they should release data because all the actual academic metastudies still have the same conclusion: we don't have any data to say anything about its safety.
While it would be definitely interesting, I am not sure it would be really helpful. This is really comparing apples and oranges here: you will have statistics for modes that are used in specific situations against more global statistics. Without more data (which would raise privacy questions), it will be hard to correct the numbers for the usage.
It's been floating around for few days. No major media has picked up the story. This seems very fake, yeah.
FWIW: the original Klippenstein story was sourced only from 2-3 sentences in the police report too (a report that was explicit about its inability to verify the claim, no less). The fact that an eyeball-driving story like this has gone this long with so much as an on the record quote from the driver...
My original intuition looking at this was "Autopilot doesn't have this behavior", and the longer this goes the more convinced I am that the driver just fibbed to the cops. This absolutely looks like a classic distracted driving accident, and if it were any other vehicle, even ones with assistive technology, no one would think otherwise.
Meh. I think criticism/public outcries of self driving is neccessary. Needs tight regulations because as much as I think self driving is entirely moronic, it is becoming reality, and that reality needs a way for the law and justice systems to be able to deal with it acccordingly.
The tech itself is impressive but not really a leap forward for humanity or transportation. More sideways meamdering than anything.
He's questioning whether the story even has addressable merits.
The story says the NHTSA report confirms FSD was on, but the story cites as its source and appears to have just cribbed their story from the CNN story.
The CNN story just says that the NHTSA confirms that Tesla's "controversial driver-assist software" was engaged and later says that the NHTSA has been investigating both Autopilot and Full Self Driving. There's nothing in the CNN story that tells if "controversial driver-assist software" is Autopilot or FSD.
CNN does not link to the actual NHTSA report. I could not find it with a brief search, and found on Reddit people who are much more familiar with NHTSA report procedures also failing to find out.
I wouldn't call the submitted article a "fake story" like he did. It's more a pointless story that has no good reason to exist and should be replaced with the CNN story and the CNN story's title.
>The CNN story just says that the NHTSA confirms that Tesla's "controversial driver-assist software" was engaged and later says that the NHTSA has been investigating both Autopilot and Full Self Driving. There's nothing in the CNN story that tells if "controversial driver-assist software" is Autopilot or FSD.
Umm, this is from the CNN article linked to as a source at the bottom...
>The Tesla’s driver told authorities that the vehicle’s “full self-driving” software braked unexpectedly and triggered the pileup on Thanksgiving day. CNN Business was first to report last month the driver’s claim that “full self-driving” was active.
I guess I wasn't clear. There's nothing in the CNN article to support that the NHTSA found that FSD was engaged, which is what the submitted article said.
The driver said it was engaged but drivers are not always correct. They can confuse Autopilot with FSD (especially given Tesla's misleading naming of "Autopilot"). Drivers even occasionally will lie after an accident if they think it might in any way be their fault.
I had a someone hit me once and it was amazing how blatantly they lied to the police about it.
A friend of mine had recently moved to someplace in the hills near Los Gatos, and I was trying to drive from Cupertino to their place for the third time.
The first two times I had missed a left turn to the road they were on, and there was no place to turn around for a mile or two so missing that left turn was quite annoying. On this third visit I was being very careful to memorize the names of the crossroads and the landmarks that I would see as I neared their road so I would not miss it (this was before consumer GPS was common).
On this third trip, my trip ended with me waking up in my car stopped beside the road a little ways before the left turn, with someone standing beside my car asking me if I was alright. I apparently then lost consciousness again.
My next memory is me still in the car, and I could see another car on the other side of the road that had evidently been in an accident, and was being tended to by police and an ambulance crew. I passed out again.
Next memory is the ambulance crew putting me in the ambulance.
According to what the other driver told police they had been going the opposite way as me approaching while I waited to make my left turn. Then at the last second I suddenly started my turn and they didn't even have time to react, hitting me at 55 mph (the speed limit there), which knocked my car back the way I'd come.
I found that story totally believable. It was pretty much the only way I could imagine our cars ending up where they did, given my planned route and the utter impossibility that I could miss the left turn a third time.
However, all the witnesses told a different story. Apparently, and I still cannot believe this even though dozens of witnesses and a bunch of physical evidence say otherwise, I missed the freaking left turn again! I drove the mile or two to the first place I could turn around, came back, and was in the right lane slowing down to start my right turn when the other car came from behind, going around 75 mph, and slammed into me.
That knocked my car across both lanes in that direction, across both lanes of heavy oncoming traffic miraculously missing everyone, where it came to rest as described above.
The driver of the other car then leapt out, not checking on his passengers (one of whom was seriously injured apparently having come out of his seat and slammed into the windshield cutting up his face quite badly) or me but instead trying to hide the open bottles of beer they had been drinking behind some bushes.
The physical evidence also backed this up. When I got hit I locked the breaks (I don't know if I actually was able to consciously do that or it just happened somehow) and the police found skid marks from my car from the point of collision all the way to where my car ended up. Also, my rear window popped out at the point of collision, without shattering, and fell to the street where it then shattered leaving a nice indicator of where the collision actually happened.
I spent a night in the hospital for observation for a concussion, and spent the next week or so kind of in a fog. A couple days after I got home I got a pamphlet in the mail from the hospital called something like "What To Expect After a Minor Concussion" that among other things warned me I might get some strong feelings of déjà vu.
Someone screwed up at the hospital (or someone there has a twisted sense of humor) and they mailed me another copy of that pamphlet the next day. Reading again that I might get str...
Really? It's published on a laptop review link farm populated mostly with presumptively-auto-generated content, not written by a journalist with a byline in any major publication. It makes a blockbuster claim for which it fails to cite sources (which should be trivially easy, since it's claiming the DOT "released a report"!). It's by an author who seemingly (see his history page) generates one hit piece on Tesla or Musk properties every few days.
If there's ever a fake source, this is it. If there was an NHTSA report corroborating this, then the direct link would be everywhere.
It’s relatively easy to take over the controls which instantly disengages FSD or Autopilot. Why this Bay Bridge driver didn’t do that just boggles me. It’s almost like these drivers have lost or no sense of a mechanical aptitude and so when something like this happens they freak out and freeze. No mechanical sense at all and just stays there to get rear ended. No athletic sense either. Driving school ought to incorporate playing tag or pushing around a grocery cart in a supermarket for a day and equate all that you do as an analog to driving mechanics and athleticism. (Just don’t use IKEA carts or you’ll go mad.)
It’s really easy to step on the accelerator or brake or turn the wheel which instantly makes the car human operated again.
To start doing that, a driver needs situational awareness. Getting that takes time. If you truly believe it's fully self driving, you think you can afford to look away and loose that awareness.
This gap between 'sometimes needs attention' and 'never needs attention' is hard to bridge and very dangerous, obviously
I’ve seen the video. This driver had Autopilot on and experienced “phantom braking”. If you drive a Tesla you’ll have experienced phantom braking at some point, a false positive condition when the Autopilot cameras thinks there’s a wall up ahead, for example if going under an overpass. It will literally brake. You step on the accelerator and cuss once or twice at Elon and you’re good to move on. This Bay Bridge driver could not make a decision and let the car brake to a stop. It’s not uncommon. Anecdotally I would say Teslas get rear-ended quite often.
>“If you drive a Tesla you’ll have experienced phantom braking at some point, a false positive condition when the Autopilot cameras thinks there’s a wall up ahead, for example if going under an overpass. It will literally brake. You step on the accelerator and cuss once or twice at Elon and you’re good to move on. This Bay Bridge driver could not make a decision and let the car brake to a stop. It’s not uncommon. Anecdotally I would say Teslas get rear-ended quite often.”
This level of apologism is shocking. “Everyone knows Teslas emergency brake to a stop in the middle of highway traffic randomly, bro. Just gotta be ready for it! No biggie!”
Yikes. Ever considered rental cars? Borrowing a friend’s car that you’re not familiar? Or just 100+ years of automotive history in which random no-input braking was considered completely unacceptable and not even a thing?
Hey you’re misquoting me. I’m no Tesla apologist actually looking to get rid of it but this is the way you live with Autopilot. I’m done with Musk and his hypocrisy. All I’m saying is that everyone and their neighbor now drive Teslas but don’t have a mechanical sense for what lane keeping assist is.
The byproduct of the migration of the modern automobile into a rolling living room will, no doubt, result in people driving who should not. I don’t mean to sound like a “standard shift get off my lawn and into a jeep” old man but…
Yes, it’s all anecdotal evidence from both my time as a vehicle owning adult and in my nearly 20 years picking up people from car accidents as a firefighter.
The issue is the FSD, not the driver. Also, drivers might not react quickly enough; what reaction time is needed to OODA. Finaly, if drivers are 100% engaged, they have no use for FSD.
If this exact same thing had happened 20 years ago, we would all agree that the people to blame are the ones behind. The person who rear-ends another is almost always at fault.
So even if FSD made the very poor decision to slow down or stop right there (and I will point out we have good reason to think this was not FSD, but rather autopilot or human driving), the blame still goes to the rear-ender.
Coming to a sudden stop in a busy lane may well be necessary if there's some unexpected obstacle in front of you. Which is why road rules require maintaining sufficient distance to the car in front of you that if it has to brake hard, you have enough time to react and do the same thing without ramming into it.
IIRC police has already said that at least some cars involved in the accident did not maintain such a safe distance, and their drivers thus bear their share of responsibility for the crash.
> No, not if someone comes to a sudden stop in a busy lane.
Why not? You’re meant to maintain enough distance to the car in front of you at all times that you should be able to come to a complete stop without rear-ending them.
The Tesla switched fucking lanes. You're can't fault someone in the second lane for not predicting some asshole will cut them off and slam on the brakes. The robot is a laughably bad driver. If a human had the collective record of either Autopilot or FSD they'd be uninsurable and have long ago lost their license.
Yeah... But in the vast majority of cases, rear-ending is caused by some combination of inattention and tailgating rather than the malice of the leading car. Let's not forget that humans are really bad at driving in general, and that this is extremely frequent.
And even if the leading car was non-malicious and merely incompetent, it's your responsibility to not hit anything in front of you. For liability, you worry about what's in front, not what's behind.
The only way you get out of being liable for rear-ending someone is if the leading car maliciously caused you to rear-end them, or if something unpredictable happened that would have made it impossible for you to avoid hitting the other car.
> The only way you get out of being liable for rear-ending someone is if the leading car maliciously caused you to rear-end them, or if something unpredictable happened that would have made it impossible for you to avoid hitting the other car.
This is literally what happened in this situation. If you have a dashcam and something like this happens, you won't be found at fault.
The driver behind the Tesla ought to have noticed the Tesla signal its lane change and would have had several seconds to ensure they were leaving sufficient space for the Tesla. If the driver had been attentive and slowed down prior to the Tesla changing lanes, as he should have, then this would have been totally avoidable.
If you went to court for having rear-ended someone in a similar set of circumstances, that's basically what the judge would tell you. Would that be frustrating to hear? Yes, but it's how it works.
It's called autopilot. The person engaging it may come to the understanding that it is the car who is driving, not they. Yet the reality is they must be more attentive than a regular driver because the Tesla can behave unpredictability. Much like overseeing an unprepared student driver.
It's also really easy to criticize from an armchair.
Nothing to do with mechanical sense or athleticism, all to do with attention and surprise. Self-driving encourages people to reduce their vigilance - after all, that's what they bought it for, so they could do less driving - while at the same time it introduces a whole new category of surprises.
People are trained for normal driving. They are not trained to deal with this kind of emergency.
A self-driving system that can't be relied on and needs constant vigilance has turned the "driver" into a driving instructor who needs to do nothing except for occasionally correct the mistakes of the computer. This is a harder job than regular driving.
It drives me nuts that we could expect people who are not driving to maintain the same level of engagement as someone actively driving. It is absurd.
Not to mention as peoples skills at controlling a car atrophy, via being a passenger, we expect them to be able to immediately switch on in a high pressure high context situation and make good decisions.
I will happily use a car that drives itself. But only once I am sat facing backwards, purely a passenger.
>It’s relatively easy to take over the controls which instantly disengages FSD or Autopilot.
What a naive assertion. Have you read takeover studies? Driver assistance systems put drivers in a state of inattention (universally, not just ones named “Autopilot” or “Full Self-Driving”, which undoubtedly makes it worse). Regaining attention and building a model of actors and hazards around you takes time — too much time to make the necessary split second maneuver to avoid an accident at highway speeds.
So, if you constantly have to maintain full attention even with FSD engaged, what are you really buying for $15k? You don’t get to sleep. You don’t get to watch a movie. You don’t even get to relax your awareness at all. If you do, the company that you paid $15k to for “Full Self-Driving” will say YOU were the one being irresponsible.
Basically, I see Tesla’s FSD as the following proposition: “Would you like to pay $15k to allow an overly-cautious and sometimes-erratic beginner driver to drive you around everywhere? Most of the time, this driver will drive so cautiously that it will frustrate you and confuse other drivers on the road. This driver may also get confused by shadows and abnormal visual phenomena, and this may lead to inexplicable emergency braking or incorrect path selection at any time. To help this driver, we’ll install a second steering wheel and pedals at your passenger seat so that you can take over driving at any time (which you WILL be required to do regularly and immediately). If you guys get in an accident, YOU will be at fault, no matter what.” Sooner or later, people will stop buying this garbage and just opt to drive themselves.
> The accident in the Bay Bridge tunnel has occurred with Tesla's Full Self-Driving Beta mode active indeed, just as the driver claimed.
This is almost certainly false. This kind of highway uses Autopilot, not FSD beta. In fact in the tweet that they show that simulates that same drive you can see on the screen that it's using Autopilot, not FSD beta.
this is the exact kind of bullshit that corporations have gotten away with for too long. As a consumer, I do not give one flying fuck what the new hip-slang for computer vision is, call it "put your life in the hands of a company run by the most burned out divorced dad person of the year".
having "self driving" in the name is probably the cause of the confusion. As far as non technical people are concerned, tesla soft ware for driving the car engaged. Whether it was one product or another is immaterial to the fact that tesla software caused the crash
But this is hacker news so for us, here, the distinction is important.
Which is what this whole absurd subthread is about. Redox pointed out that the article attributes the issue to the wrong product and you all are saying that it doesn't matter.
Well perhaps you are not interested in the topic and it doesn't matter for you. For me it matters.
---
Imagine an article about a Google Chat issue was posted on hacker news. Redox99 points out that the issue is actually in Google Meet. And then you, bob and boucher appear and you reply that it doesn't matter which one it is, both are google chat apps. Well I believe it matters a lot.
The sub thread you replied to started with the poster literally saying “as a consumer…”. It was not a discussion of the technical minutiae, but of the overall situation
Finding out which system it was using is pretty much the first step of any kind of investigation, which this is.
As a consumer you probably would want to know if this is caused by the autopilot that you are using and have been using for years, or if it's the beta FSD that you have to opt in.
Thats an interesting claim, that the highway would trigger use of autopilot and not FSD. How do you know that? I think Tesla, who would know for a fact the status of FSD during this accident, might want to tweet that FSD wasn't on if it wasn't. But they haven't. Also, as a human, why would you hit the breaks in the fast lane with cars approaching? This does not happen very often, especially on the incredibly crowded Bay Bridge.
I actually was there at this accident and swerved just in time to miss the cars piling up (you can see my car in both videos of the accident just getting out). It was a truly scary thing.
FSD does not support highways yet, so it automatically switches to autopilot when you enter a highway. You can tell because they use completely different UIs, visualizations, settings, etc (FSD and autopilot are fully separate software).
Tesla has been promising FSD v11 which will support highways for a very long time already, but it isn't out yet.
Literally impossible is kind of a strong statement, but yeah people have driven there and shown that it uses Autopilot there (because it's classified as highway on the map data), which is the expected behavior as FSD does not support highways yet.
Interesting, I wonder what the behavior is when someone turns FSD on when they are already on a highway as claimed in this case, as opposed to someone who has it on before entering. Maybe that's the issue, if you turn FSD on when already on a highway, the car starts FSD, senses its on a highway then given its programming that FSD should not operate on a highway, consequently hits the brakes? Tesla knows whatever happened in fact for sure.
You can't "choose" which one[1]. It's the same button/stalk. The car chooses by itself.
This is almost 100% the driver just confusing the name or a slip of the tongue, which is understandable. He bought FSD, he probably has it enabled and uses it, and knows that on highways the UI changes. Just doesn't know that when the UI changes it's the old autopilot.
[1] Technically you can disable FSD beta, thus it will always use autopilot
Regardless of how it actually works, wouldn’t a corporation have a very strict policy of never ever allowing any talk about ongoing situations like this on sm/marketing/pr channels?
Doesn't that make it even more unsafe? Two different systems that behave in different ways, one may require more intervention than the other. Car switches between the two when it wants and the driver is supposed to know the intricate differences between the two?
FSD beta, as the name implies, is for people that want to play with the experimental version. You have to request access to the beta, and then you may or may not be granted (although since ~1 month ago basically anybody that requests it will get accepted in a few minutes).
So it's kind of expected that its users are a bit more aware of it's intricacies.
Also, even though it switches and the UI is different, I don't think it makes much of difference as a driver. Both systems are Level 2 and require constant driver supervision.
Autopilot although much simpler, is more tried and true. The delays of FSD v11 are most likely because they are having trouble achieving equal performance to Autopilot on highways.
Your whole spiel about how people have to request access to FSD doesn't seem very relevant. The feature is being sold to the general public to be used on public roads. That means it has to be safe enough to be used by the general public on public roads, "beta" sticker or not, access request process or not.
Even if you buy FSD, you have to request it be in the FSD beta feature. If you aren’t in the FSD you get some features above non FSD like parking assist and summon. Everyone gets Autopilot which is really just the combination of adaptive cruise control and lane keeping.
Twist the steering wheel against the direction the car is turning it
Tap the brakes
These all immediately drop out of AP.
The driver also has the option of pressing down on the accelerator to make the car go faster and cancel AP’s anomalous braking.
One situation I can think of is using Navigate on Autopilot and the car thinks the exit is much further along than it is, tries to take the “exit” which is actually a concrete wall, AEB kicks in due to imminent collision.
But then why haven’t other Teslas had the same issue there?
Your counter claim is apparently false based on the attempt at recreation in this linked video. The driver engage FSD while driving into the same section of the Bay Bridge and tunnel
132 comments
[ 5.3 ms ] story [ 168 ms ] threadI know that in my car, the FSD stuff ends and regular autopilot takes over on highways and other controlled access roads. I've never driven the bay bridge in my Tesla (since I live thousands of miles away), but I remember it being a controlled access highway, so I assume it was the AP logic, not the FSD logic that was in control.
(I also haven't seen any commentary from musk, but it's hard to search twitter)
What kind of idiot STOPS ON A DIME with speeding cars hot on his tail? Only a Computer... It's unfortunate that the Driver did not have the sense to Slam on the Gas -The idea of "Slow == Safe" can be disastrous...
So we have 1 car that approaches the light, gets a red, and stops. It has to shed E kinetic energy. Then 10 cars approach the intersection on the perpendicular road. Before they reach it the light changes and they all have to stop, shedding 10 E kinetic energy. The first car resumes its journey. The light changes again and the other 10 resume.
The net result is that encountering the intersection caused this 11 car system to shed 11 E kinetic energy more and to consume 11 R idling energy more than would have been consumed if the intersection had not been there (e.g., if the crossing had been an overpass).
What he's proposing is that some sort of smart traffic light system could notice the 10 approaching cars and hold their green a little longer so that the 10 cars do not have to stop. Suppose it has to hold the light long enough that car 1 has to wait twice as long.
The energy cost of the intersection on the 11 car system under that scenario is E shed kinetic energy and 2 R idling energy. That's going to be much less that 11 E + 11 R.
Humans do it on a pretty regular basis unfortunately. I have seen people slam on the brakes at highway speed for minor road debris, small animals, and a couple of times for no apparent reason. The only thing you can do as a driver is minimize your own risk by leaving sufficient following distance.
It's very rare. I've never seen it, IIRC.
Dolphins die on a pretty regular basis and I've never seen it. That doesn't mean it doesn't happen on a regular basis...
Ironically this is one of the areas where driver assist can be helpful. Adaptive cruise control that maintains following distance though in practice I don’t think people use it much.
If the stopping car had been an old Toyota and we all had the same footage... well, it probably wouldn't even make the news in the first place... but if it did, we'd all be blaming the rear-enders. But because it's a Tesla, and there is a possibility that AI was involved, suddenly we shift our blame from the rear-enders and tailgaters to the computer.
We've always had double standards for humans and machines.
Yes brakes will apply sooner and follow distance is likely to be safer/wider.
I just don’t yet trust it’s good enough.
If a tree falls on the car in front of you and you hit them because they stopped instantly, it’s hardly your fault.
2. It's impossible to be safe if someone is (illegally!) tailgating.
tesla.com/VehicleSafetyReport
In addition, those "safety reports" continue to spout misinformation that the National Transportation Safety Board is currently suing Tesla for
(Edit: if you’re going to downvote me, please explain how a literal interpretation of exactly what I wrote is technically incorrect.)
Plenty of analysis is based on data that isn't released. And do you have data that says otherwise? Why should we believe you?
See also:
- The dangerous "we're almost at full self-driving" myths Musk continues to spread (which is probably part of the reason Google didn't acquire them given that they've been ahead on the research of this and knew how BS that is)
- The battery swap lies that they made a huge announcement about in order to take advantage of a California subsidy and then simply... never actually allowed any Tesla owner to swap batteries
- Tesla lying to shareholders about warranty expense by categorizing it as "goodwill" in order to inflate their gross margin and net income (since at least Q2 2018)
- On top of the California battery swap stuff, they've been sued for subsidy fraud in multiple countries including Canada and Germany. Much of it's "profitability" has simply been selling regulatory credits to other OEMs
- Musk using Tesla to bail out his cousin's failing solar company SolarCity and not disclosing the personal connection to shareholders
NTSB did not sue Tesla.
That is appropriate in social situations, not in situations with significant consequences. If lives are at stake (or even lots of poeple's money), the burden of proof is on the person making the claim.
For example, that is true in the legal system, in science, and when you give a presentation to your boss. Saying 'you should believe me by default' doesn't go over well in those situations.
Most data we use is derived and processed data. Much of the data we use is a statistic.
And now we have data to show it :)
At this point we still don't have any actual public evidence that driving assistance technology is even safer than driving without it. If Tesla's so confident in it they should release data because all the actual academic metastudies still have the same conclusion: we don't have any data to say anything about its safety.
Because no other car maker is mandated to report these numbers either. Tesla is not the only brand with adaptive cruise control or line assist.
(I think sharing the data would be the right thing to do though.)
What about FSD?
FWIW: the original Klippenstein story was sourced only from 2-3 sentences in the police report too (a report that was explicit about its inability to verify the claim, no less). The fact that an eyeball-driving story like this has gone this long with so much as an on the record quote from the driver...
My original intuition looking at this was "Autopilot doesn't have this behavior", and the longer this goes the more convinced I am that the driver just fibbed to the cops. This absolutely looks like a classic distracted driving accident, and if it were any other vehicle, even ones with assistive technology, no one would think otherwise.
The tech itself is impressive but not really a leap forward for humanity or transportation. More sideways meamdering than anything.
Criticism is fine. Mistrusths are not. You agree with that much, right?
You type that, which is easy. Can you address the story on its merits?
The story says the NHTSA report confirms FSD was on, but the story cites as its source and appears to have just cribbed their story from the CNN story.
The CNN story just says that the NHTSA confirms that Tesla's "controversial driver-assist software" was engaged and later says that the NHTSA has been investigating both Autopilot and Full Self Driving. There's nothing in the CNN story that tells if "controversial driver-assist software" is Autopilot or FSD.
CNN does not link to the actual NHTSA report. I could not find it with a brief search, and found on Reddit people who are much more familiar with NHTSA report procedures also failing to find out.
I wouldn't call the submitted article a "fake story" like he did. It's more a pointless story that has no good reason to exist and should be replaced with the CNN story and the CNN story's title.
Umm, this is from the CNN article linked to as a source at the bottom...
>The Tesla’s driver told authorities that the vehicle’s “full self-driving” software braked unexpectedly and triggered the pileup on Thanksgiving day. CNN Business was first to report last month the driver’s claim that “full self-driving” was active.
https://edition.cnn.com/2023/01/17/business/tesla-8-car-cras...
Edit: This is not me taking a position in this discussion, just clarifying that the article does, in fact, say that.
The driver said it was engaged but drivers are not always correct. They can confuse Autopilot with FSD (especially given Tesla's misleading naming of "Autopilot"). Drivers even occasionally will lie after an accident if they think it might in any way be their fault.
I had a someone hit me once and it was amazing how blatantly they lied to the police about it.
A friend of mine had recently moved to someplace in the hills near Los Gatos, and I was trying to drive from Cupertino to their place for the third time.
The first two times I had missed a left turn to the road they were on, and there was no place to turn around for a mile or two so missing that left turn was quite annoying. On this third visit I was being very careful to memorize the names of the crossroads and the landmarks that I would see as I neared their road so I would not miss it (this was before consumer GPS was common).
On this third trip, my trip ended with me waking up in my car stopped beside the road a little ways before the left turn, with someone standing beside my car asking me if I was alright. I apparently then lost consciousness again.
My next memory is me still in the car, and I could see another car on the other side of the road that had evidently been in an accident, and was being tended to by police and an ambulance crew. I passed out again.
Next memory is the ambulance crew putting me in the ambulance.
According to what the other driver told police they had been going the opposite way as me approaching while I waited to make my left turn. Then at the last second I suddenly started my turn and they didn't even have time to react, hitting me at 55 mph (the speed limit there), which knocked my car back the way I'd come.
I found that story totally believable. It was pretty much the only way I could imagine our cars ending up where they did, given my planned route and the utter impossibility that I could miss the left turn a third time.
However, all the witnesses told a different story. Apparently, and I still cannot believe this even though dozens of witnesses and a bunch of physical evidence say otherwise, I missed the freaking left turn again! I drove the mile or two to the first place I could turn around, came back, and was in the right lane slowing down to start my right turn when the other car came from behind, going around 75 mph, and slammed into me.
That knocked my car across both lanes in that direction, across both lanes of heavy oncoming traffic miraculously missing everyone, where it came to rest as described above.
The driver of the other car then leapt out, not checking on his passengers (one of whom was seriously injured apparently having come out of his seat and slammed into the windshield cutting up his face quite badly) or me but instead trying to hide the open bottles of beer they had been drinking behind some bushes.
The physical evidence also backed this up. When I got hit I locked the breaks (I don't know if I actually was able to consciously do that or it just happened somehow) and the police found skid marks from my car from the point of collision all the way to where my car ended up. Also, my rear window popped out at the point of collision, without shattering, and fell to the street where it then shattered leaving a nice indicator of where the collision actually happened.
I spent a night in the hospital for observation for a concussion, and spent the next week or so kind of in a fog. A couple days after I got home I got a pamphlet in the mail from the hospital called something like "What To Expect After a Minor Concussion" that among other things warned me I might get some strong feelings of déjà vu.
Someone screwed up at the hospital (or someone there has a twisted sense of humor) and they mailed me another copy of that pamphlet the next day. Reading again that I might get str...
If there's ever a fake source, this is it. If there was an NHTSA report corroborating this, then the direct link would be everywhere.
It’s really easy to step on the accelerator or brake or turn the wheel which instantly makes the car human operated again.
This gap between 'sometimes needs attention' and 'never needs attention' is hard to bridge and very dangerous, obviously
This level of apologism is shocking. “Everyone knows Teslas emergency brake to a stop in the middle of highway traffic randomly, bro. Just gotta be ready for it! No biggie!”
Yikes. Ever considered rental cars? Borrowing a friend’s car that you’re not familiar? Or just 100+ years of automotive history in which random no-input braking was considered completely unacceptable and not even a thing?
Yes, it’s all anecdotal evidence from both my time as a vehicle owning adult and in my nearly 20 years picking up people from car accidents as a firefighter.
But yeah, super easy to just zone out these days.
So even if FSD made the very poor decision to slow down or stop right there (and I will point out we have good reason to think this was not FSD, but rather autopilot or human driving), the blame still goes to the rear-ender.
No, the actual research says otherwise
> If this exact same thing had happened 20 years ago, we would all agree that the people to blame are the ones behind.
No, not if someone comes to a sudden stop in a busy lane.
IIRC police has already said that at least some cars involved in the accident did not maintain such a safe distance, and their drivers thus bear their share of responsibility for the crash.
And even if the leading car was non-malicious and merely incompetent, it's your responsibility to not hit anything in front of you. For liability, you worry about what's in front, not what's behind.
The only way you get out of being liable for rear-ending someone is if the leading car maliciously caused you to rear-end them, or if something unpredictable happened that would have made it impossible for you to avoid hitting the other car.
This is literally what happened in this situation. If you have a dashcam and something like this happens, you won't be found at fault.
If you went to court for having rear-ended someone in a similar set of circumstances, that's basically what the judge would tell you. Would that be frustrating to hear? Yes, but it's how it works.
Nothing to do with mechanical sense or athleticism, all to do with attention and surprise. Self-driving encourages people to reduce their vigilance - after all, that's what they bought it for, so they could do less driving - while at the same time it introduces a whole new category of surprises.
People are trained for normal driving. They are not trained to deal with this kind of emergency.
A self-driving system that can't be relied on and needs constant vigilance has turned the "driver" into a driving instructor who needs to do nothing except for occasionally correct the mistakes of the computer. This is a harder job than regular driving.
Not to mention as peoples skills at controlling a car atrophy, via being a passenger, we expect them to be able to immediately switch on in a high pressure high context situation and make good decisions.
I will happily use a car that drives itself. But only once I am sat facing backwards, purely a passenger.
What a naive assertion. Have you read takeover studies? Driver assistance systems put drivers in a state of inattention (universally, not just ones named “Autopilot” or “Full Self-Driving”, which undoubtedly makes it worse). Regaining attention and building a model of actors and hazards around you takes time — too much time to make the necessary split second maneuver to avoid an accident at highway speeds.
So, if you constantly have to maintain full attention even with FSD engaged, what are you really buying for $15k? You don’t get to sleep. You don’t get to watch a movie. You don’t even get to relax your awareness at all. If you do, the company that you paid $15k to for “Full Self-Driving” will say YOU were the one being irresponsible.
Basically, I see Tesla’s FSD as the following proposition: “Would you like to pay $15k to allow an overly-cautious and sometimes-erratic beginner driver to drive you around everywhere? Most of the time, this driver will drive so cautiously that it will frustrate you and confuse other drivers on the road. This driver may also get confused by shadows and abnormal visual phenomena, and this may lead to inexplicable emergency braking or incorrect path selection at any time. To help this driver, we’ll install a second steering wheel and pedals at your passenger seat so that you can take over driving at any time (which you WILL be required to do regularly and immediately). If you guys get in an accident, YOU will be at fault, no matter what.” Sooner or later, people will stop buying this garbage and just opt to drive themselves.
This is almost certainly false. This kind of highway uses Autopilot, not FSD beta. In fact in the tweet that they show that simulates that same drive you can see on the screen that it's using Autopilot, not FSD beta.
But this is hacker news so for us, here, the distinction is important.
Which is what this whole absurd subthread is about. Redox pointed out that the article attributes the issue to the wrong product and you all are saying that it doesn't matter.
Well perhaps you are not interested in the topic and it doesn't matter for you. For me it matters.
---
Imagine an article about a Google Chat issue was posted on hacker news. Redox99 points out that the issue is actually in Google Meet. And then you, bob and boucher appear and you reply that it doesn't matter which one it is, both are google chat apps. Well I believe it matters a lot.
As a consumer you probably would want to know if this is caused by the autopilot that you are using and have been using for years, or if it's the beta FSD that you have to opt in.
I actually was there at this accident and swerved just in time to miss the cars piling up (you can see my car in both videos of the accident just getting out). It was a truly scary thing.
Tesla has been promising FSD v11 which will support highways for a very long time already, but it isn't out yet.
This is almost 100% the driver just confusing the name or a slip of the tongue, which is understandable. He bought FSD, he probably has it enabled and uses it, and knows that on highways the UI changes. Just doesn't know that when the UI changes it's the old autopilot.
[1] Technically you can disable FSD beta, thus it will always use autopilot
So it's kind of expected that its users are a bit more aware of it's intricacies.
Also, even though it switches and the UI is different, I don't think it makes much of difference as a driver. Both systems are Level 2 and require constant driver supervision.
Autopilot although much simpler, is more tried and true. The delays of FSD v11 are most likely because they are having trouble achieving equal performance to Autopilot on highways.
https://www.notateslaapp.com/tesla-reference/878/tesla-fsd-b...
Twist the steering wheel against the direction the car is turning it
Tap the brakes
These all immediately drop out of AP.
The driver also has the option of pressing down on the accelerator to make the car go faster and cancel AP’s anomalous braking.
One situation I can think of is using Navigate on Autopilot and the car thinks the exit is much further along than it is, tries to take the “exit” which is actually a concrete wall, AEB kicks in due to imminent collision.
But then why haven’t other Teslas had the same issue there?
https://twitter.com/TeslaOwnersEBay/status/16140836244581253...
All these names are confusing to be honest, so it's understandable that people mix them up.