Federal investigators blast Tesla, call for stricter safety standards
arstechnica.com
arstechnica.com
One thing to keep in mind, is thats its not just the feds, as you will notice testing is done in certain states that have less regulation (read: governors hands got greased)
Ill leave you with one insight into the sausage factory to contemplate. I told a c-level we needed a code review, and he balked, saying something to the effect of "Do you know how much that costs, how long it takes, and how hard it is?! No" This company has vehicles on public roads already!...
"Do you know how much that costs, how long it takes, and how hard it is?! No" This company has vehicles on public roads already!...
Safe "AI" driving is in principle unsolvable.
Blow the whistle.
The car is decidedly paranoid when on full autopilot to the point of overly cautious at times. It certainly won't let you have your hands off the wheel nor with a light touch suffice. I tried timing the systems "apply light pressure" warnings interval but its very random and definitely more aggressive when on a curve as its very easy then for it to recognize you aren't holding the wheel.
There are warnings when you use autopilot that flat out say you have to pay attention and demands are constant that you show you are by causing real resistance on the steering wheel.
On anything but freeways it won't even go over five miles over the speed limit. Straight cruise control does not have that limit but once AP is engaged you are restricted to +5. It will slow down automatically for detected speed limit signs with a lower speed but not auto increase for higher limits. It will not go through a green light unless car in front does first without you indicating its safe.
Now interestingly enough Tesla has kicked people out of beta of the newest software for not paying enough attention to the car either through use of the inside camera; the beta uses it; and steering wheel resistance.
So perhaps if you fail enough maybe the car should say you don't pay enough attention to use the feature and must wait 24 to 48 hours before you try again and then be overly strict on driver detection?
It certainly is not perfect, current general availability version, but it is amazing for what it can do especially at night or poor weather.
https://www.thedrive.com/tech/39688/drivers-new-to-automated...
We all know that there are car accidents every year, and it is no secret that driver inattention is the root cause for a very large percentage of them.
But what can be done about them? The NTSB is focusing almost exclusively on driver monitoring while the driver assistance is turned on.
But what about when its turned off? And more generally, how do these systems compare against driving without such systems at all? Its clear that the current state is far from perfect, but an overly heavy focus only on failure cases will lead to misleading conclusions.
Auto pilot vs human drivers is not apples to apples because of the culpability question imo.
There are always 2 highly trained professionals in a cockpit and both are familiar with emergency situations because they regularly train for it. They are in constant contact with ATC that keeps other planes far away from them. In case of a crash or even a deviation there are entire agencies all over it with sometimes international investigators present. All communication is recorded.
A driver is not only alone, they can have very poor skills in general which were last tested 40 years ago. They are on the road with dozens of vehicles around them, tightly packed and often separated by less than a vehicle width between them. If a crash occured there may be a few police officers present that pick up the pieces and fill out forms to satisfy the insurance.
Nup. Sometimes it's just one barely trained amateur.
Lack of adequate recurring certifications or poor standards does not absolve the driver of his responsibily towards himself and more importantly to others on and near the road, same as the pilot , inadequate oversight is not an excuse.
driving a 2+ton vehicle that at the typical speeds can be fatal or cause major injury you are responsible period. Same for any other equipment or a pet
Do you still believe that the driver is responsible under those circumstances?
Anything L2 or under is a “driver assistance system”, and the driver maintains control and responsibility at all times. All of the systems in recreational or commercial aircraft today (autopilot, auto-land) are L2 or less.
This is why it’s very critical that automakers not blur the lines to drivers. We are approaching an inflection point as to who is responsible behind the wheel, the car or the AI.
Granted in the late 90s the self driving technology was mostly focused on closed roads and using modified systems (eg magnetic pins in the road) to help cars self locate
But it came down to: if there’s an accident who’s responsible? The driver? The car manufacturer? The road designer? The road builder? The government?
For this to all work, companies have to take responsibility (not legal, but much more ethical than Tesla has shown) for their technology and you probably need a government system like the vaccine adverse event reporting system with associated payouts.
My snarky answer says “who wrote and who is interpreting the licensing agreement”
I remember the big problem with them were stray soda cans triggering stop.
I don't see how current technological advances can any improve the local environment awareness of the car until the actual cognition element is there.
Fairly soon autopilot will be safer than new drivers and the elderly who are aging out of the ability to drive safely. I hope we have the political will to potentially save my parents' and child's life by allowing them to have an AI driver even if it's not 100% safe.
To be it must navigate any situation without any driver intervention:
- parking in tight spaces
- follow the assigned turn lanes in an intersection
- follow the local laws on turn on red
- navigating streets that are less than a meter wider than the vehicle and have vehicles parked alternating between right and left but are still both ways (very common in Europe)
- nighttime driving in heavier rain
- following diversions that are not published, badly signposted, and were set up quickly by the city due to sudden a hole or similar in the street
- follow streets without any or even wrong markings on them
Most importantly it must not drive people into dividers, well-lit emergency equipment, and/or crossing trailers.
The fun thing about this is: we both won't ever be able to prove our theories right because neither enough elderly nor enough new drivers will own a Tesla with "Autopilot". They are just too expensive.
This assumes that our hill climbing hasn't hit the top of the hill in this area. I submit that there's almost no evidence of recent continuous growth in this technology.
Alternately, your own auto liability insurance will include coverage for your car’s automaton, and again it will be priced based on risk. For most people, the premium will be lower than if they drive themselves, as the risk will be lower.
Want to start a class action lawsuit? Go for it. The company might say you've waived that right, but any such clause is null and void.
If autonomous vehicles are safer than human drivers, then ultimately they are going to be safer to ensure.
The closest direct equivalent with driverless is that the vehicle may lose its right to operate until the problem has been corrected. I'd actually expect this to happen somewhat frequently in the early days.
Another case in point: F-35.
Please swipe up and to the left to apply the brakes or draw a peace symbol if you'd like to eject instead of dying. - megacorp UX of the future
My car's "infotainment" system is so bad I've given up on it. When working with a phone, the sound is good - but rattling thru menus via buttons on multiple surfaces, never quite knowing what some of the terms mean (because I'm focused on the road, not on sorting out "MEDIA SOURCE" vs "INPUT SOURCE" in context of connecting Bluetooth or whatever), and having to literally alternate between sound unit face and the "OK" button on the steering wheel, well just forget it we're using the phone's own speakers (relatively lousy but they work). There's a speech recognition system in there too, but that goes nowhere (first, memorize all the possible commands it recognizes; next, say the phrases just right in a noisy & distracting environment...).
Yeah, physical buttons/knobs are distracting too. At least a touchscreen system can better present/organize information, and likely comes with a viable speech recognition system. And can sanely interface to a phone via Bluetooth.
If you hand misses a physical button, you easily orient yourself by recognizing the shape of the and relative location of the button by touch, whereas a mistouch on screen might have activated some functionality, or at the very least is harder to orient without looking because there are no physical features.
While OEM infotainment systems universally suck, it’s a non sequitur to say that a touchscreen fixes any of these issues, in particular your Bluetooth and voice problems. My Acura has a touchscreen and voice and it sucks. The voice is slow to respond with constant boilerplate announcements, and the typing addresses into the gps is painful, even when using autocomplete (it animates the typing, and then forces you to reconfirm the completed text.)
Touch and voice control can be improved with an overnight software update.
My current car, a random 5 year old sedan has about 4-5 knobs on the console I use- volume, heated seats, ac mode, temperature, and nav knob that rotates, pushes in, and pushed in 4 directions.
I can do all this without taking my eyes off the road.
I also have an 8 inch touchscreen that I never use because it’s too hard while driving.
Voice controls are terrible and I won't be convinced otherwise.
[0] “It has a streaming service like Pandora!” “Is it Pandora?” “It’s Sketcher. It’s like Pandora!” “So it’s not Pandora.” “No.”
[1] “This car has full self driving.” “So is it going to turn here?” “No. It has FSD capability, but it’s not on because of regulatory approval. It should be here in a year to two.” (That was four years ago.)
Is there any evidence at all that this is the impairment of safety opponents claim? Model 3s have been sold in large numbers since 2018. Are drivers crashing into other cars at higher rates? Or are we just going to keep beating this dead horse?
https://www.tesladeaths.com/#FAQ
Note that I say "they might" because tesladeaths.com is probably overcounting slightly. It's created by a group of short sellers, and the IIHS may count fewer incidents for other cars.
Until the IIHS releases properly comparable numbers themselves, it's hard to say for sure.
More than slightly, and that includes far more than just the ones where the touchscreen replaced nearly all controls not on the steering pedestal (Model 3, Y).
I don't know how many years have to go by without supporting evidence before we can move on from these claims. More than three, apparently, and despite other manufacturers making the same shift toward touchscreens.
It was tired when the car first came out, but at least it was a valid concern. Now it just feels like an unwillingness to accept reality.
If the thread is about safety, yes.
This came out this week:
> A study measuring the affects of driver distraction suggests using a dashboard touchscreen is up to four times worse than being at the UK’s drink-drive alcohol limit.
https://www.tu-auto.com/touchscreen-infotainment-four-times-...
Every Model 3 has one, and they’ve been made in large numbers since 2018. It seems like we’d have the numbers by now if it were the threat it’s claimed to be.
I’m pretty sure you can operate it with voice commands as well, and the click wheel in the 3?
That said, why would you not leave it on auto? I own a normal car and don’t touch the dashboard at all during 98% of drives, everything is set to auto (wipers, lights, AC).
Also, It’s not the click wheel, it’s the left stalk in the 3/Y. And touching the button on the stalk brings up the part of the UI that controls the wipers. I can reach it with my pinky while my hand is on the wheel.
To your point, I’ve never taken it off auto. I just hit the button on the stalk every now and again to get an extra wipe/clean.
The wipers engage automatically by default. If you disable that, you can push the button at the end of the stalk to wipe once, or adjust via voice commands. But yes, you can also use the touchscreen.
On the other hand: calling something “Full Self-Driving”, when it is in fact definitely not fully self-driving, kind of flies in the face of the obvious plain text reading of the phrase “Full Self-Driving”.
- Edward S. Herman
Decades of experience has repeatedly taught that people are not perfect robots that always read manuals, and if you design your system under the assumption that they are then it is you that is at fault, not them.
There's a reason we have safety interlocks and dead man switches and seat belts and helmets and ... It's no use having a manual that says "do not open the microwave door while it is running".
That’s exactly how the manual/pilot operating handbook and supplements for most airplane autopilots (and airplanes) read.
There's the classic aphorism about tools like lathes: "This machine has no brain, please use your own." But that's usually in an industrial setting with OSHA, mandated safety trainings, inspections, a culture of reporting unsafe practices without judgement, etc.
As CNC machines get cheaper and smaller, we are starting to see things like laser cutters sold to consumers with a pair of glasses and a manual rather than an enclosure. Hope you remember to lock the door if you have kids or pets...
It is a worrying trend which seems to be accelerating in the US due to crumbling and senescent regulatory institutions.
A much better solution is to add a physical safety interlock which prevents you from starting the lathe without removing the chuck key, which is exactly what some lathes do!
I agree the safety level of a lot of Chinese machines is very poor.
Also, aren't chainsaws nowadays designed so that if the user doesn't have two hands on the chainsaw grabbing the triggers, the chainsaw automatically cuts off? IIRC that's a thing.
Edit:there's the grip-safety ("safety throttle") for the trigger finger and there's the chain break for the forward hand, which will push forward and activate the brakes if the hand let's go or if recoil pulls the hand forward.
The user also needs training as there's still no way to guard the exposed blade when in use nor to keep it out of proximity of humans.
On some earnings call some analyst asked Elon if he would add LIDAR to the cars even if it costed literally nothing. "Nah, hmm, no. We would not.".
I for one, would be more than happy to prolong my commute, on a dedicated, reduced speed lane, if it meant I could sit in the back seat and work. The belief that we’ll incrementally get to level 5 besides human drivers on the extant infrastructure (lanes) is very misplaced.
Fingers crossed that the feds can develop a roadmap (sorry!) to support the transition from level 2 to level 5. It will not happen on open roadways—lidar or not!
Hence, levels. Besides, we're in a transitional period on this tech, so I think there's value in defining "how" automatic the experience is.
The binary statement youre looking for comes between L2 and L3. A L2 system is a driver assist system, the driver is in command at all times. An L3 system can drive itself (driver napping or otherwise occupied) under specific well-defined conditions.
You are. Pretty much every other manufacturer offers adaptive cruise control, lane keeping, and the like. They're just honest about their capabilities.
edit: I'm pretty anti-Tesla (strongly dislike the focus on the central screen and other driver-hostile 'features', don't think "full self driving" should be advertised as such, etc). But I still think their capabilities are significantly above the other car manufacturers enough that they shouldn't be directly compared in that way. But I guess readers like the fact that calling them "adaptive cruise control" and "active lanekeeping" is more accurate and much less bullshitty than Tesla.
If it's a matter of occasionally going over or under the speed limit (or if the speed limit on your route has sudden changes), you may want to look into getting a vehicle with a manual speed limiter. That, along with something like Waze, can be a big help.
Losing the ability to change the limit or (dis)engage at will makes the feature practically useless too, considering the various speed limit zones one goes through in one drive -- not to mention dynamic speed limits based on weather and other conditions.
I don't think it's unreasonable to suggest part of the problem here is a 60 mile commute twice a day. That's an absurd distance to drive for a job.
Like most Money Mustache article’s the analysis looks sound because “numbers” but the numbers have a bunch of fucked assumptions to reach them.
> When I was at google
So it's fatally flawed if you work at Google (or another high-paying tech company) as a SWE in Silicon Valley - which is a minuscule fraction of the working population of software engineers. It may not be the assumptions in the article that are "fucked". You might just be an edge case.
> The housing cost difference was obscene and would easily offset the cost of a new car every other year.
It's not just the cost of a new car. It's also the time and energy lost to the commute, the increased risk of a serious accident, the opportunity cost of all that time spent in traffic.
As a SWE working at a big tech company in Silicon Valley, I've done pretty well financially by avoiding long commutes - and had lots more spare time. There's more than one way to skin a cat.
Yeah, aka the assumption is fucked. This dude cranks out shit on his blog as if it’s based on some objective stuff but realistically it’s a bunch of evidence based on his lifestyle.
> It's not just the cost of a new car. It's also the time and energy lost to the commute, the increased risk of a serious accident, the opportunity cost of all that time spent in traffic.
That’s the flaw again. “Time and energy” is not lost in a commute unless you waste it. If you are a competent SWE, you should be able to spend that time thinking about architecture, design plans, algorithms, data flows, UX, and on and on. Not to mention you can keep up on industry podcasts, industry news, and a bunch of other audio only shit.
BTW, “increased risk of a serious accident” is yet another misunderstanding of a stat. “Deaths per million miles” in the US is a general stat averaged across commuters, drunk drivers, poor driving conditions, old cars, and bad roads.
I’d posit that my chance of death putting along on the 101 at 20-30 mph during rush hour in a car with modern safety features for an entire year was lower than on a single night on a two lane road up in Sonoma on the weekend.
All driving is not equal, and that’s just one of the many mistakes the money mustache blog makes. I’m all for promoting objective analysis of costs, but it needs to be done on a case-by-case basis. Trying to make general financial advice based on averages is as much folly as trying to make a shirt that fits everyone by sizing it to the average.
I do all of that during my workouts and walks, or while cooking my family a healthy dinner. All of which I have time to do because I'm not spending 2 hours/day "putting along on the 101 at 20-30mph". On top of saving a ton of money and helping the environment, it's better for my health and my stress levels, because I'm not swearing at the 10th person to jump the queue at the exit. Before WFH started I was in far better shape than my colleagues who commuted from 1.5 hours away because they could get more house for less money.
Long commutes are a destroyer of health, wealth, happiness, and the environment. They steal time from our lives, with our children and families and friends - the only non-renewable asset we have.
The differences for FAANG incomes make this simpler for you to understand. At Google you can clear $500k/year pretty easily (this was common L5 TC when I left) spending about 6 hours in the office. Let’s assume you spend 2 hours each way and it additionally costs $25/hour in driving.
235 working days in 2020 (assuming you do 0 WFH days and take all PTO).
2354$25=$23,500 in cost ($500k-23,500)/235=$2027 per 10 hours of work (6 in office, 4 commuting).
You work from home for 6 hours a day for a non-FAANG and clear... maybe $150k?
$150k/235=$638 per 6 hours of work.
Commuter hourly after commute dilution: $202 At home hourly: $106
If you’re rich enough to get a house for a big family close to your office, good for you. That shit just doesn’t apply for people that make these trade-offs.
> They steal time from our lives, with our children and families and friends - the only non-renewable asset we have.
Truly spoken with the privilege of being rich enough with a non-commuting job to not have money struggles. Everyone I know putting up with these commutes has to give up a huge chunk of income that will make finances a burden or abandon their community for a worse quality of life near the job.
Having worked for a FAANG myself, that assumption is not incorrect at all. I know what I'm talking about.
You may not find a house with a yard or two car garage within an appropriate commute distance of the FAANG campuses in Menlo Park, Palo Alto, Mountain View, Sunnyvale, Redwood City, or Los Gatos. You may have to settle for a townhouse, or a condo, or even gasp, shudder an apartment. But there are ways to make it work.
Everyone has to decide what they value more. Big house for their family, but you give up 2 hours/day that you could spend with that family. Or lots more time with the family, good health, and less stress, but you give up some space and/or comfort. For me it's a no-brainer.
We could have a similar train network if we wanted one, we just don’t want one enough to pay for it.
Thanks for pointing that out. I meant 50, just edited
I have it in a less than optimal car (Honda Fit) and it has transformed how I drive.
I don’t think there is enough evidence that the current ADAS systems are better or worse than human drivers at this point. I’d say the best way to improve the situation is not to have heavy handed regulation at this stage and let the technology play out.
Tesla cars in Autopilot mode are roughly 8x less likely to get into an accident compared with the average number of reported accidents in the US.
Note: There is a bias about when Autopilot is enabled (i.e. on highways). There is a bias in the type of person that can buy a Tesla car.
That said: Tesla is legally obligated to report all Autopilot accidents. Most accidents are not reported, only the major accidents are reported. So it’s likely that Tesla car’s accident rate are even better than the data suggests.
This is such an enormous bias, it invalidates the rest of the comment.
Yes. This makes any extrapolation to the population as a whole impossible without a lot of assumptions. In particular, things like this are meaningless because of sampling bias:
> Tesla cars in Autopilot mode are roughly 8x less likely to get into an accident compared with the average number of reported accidents in the US.
I don't understand how this is a controversial opinion. The moment self-driving cars is mentioned, safety is the first thing that comes to mind. As a parent comment stated, Tesla is legally obligated to report every accident. And if a huge accident occurs, it will be in the news. Meanwhile it takes a 100 car pileup for any other vehicular accident to make it into a tiny sliver of the news cycle.
I don't own a Tesla or any of their stock, but the crux of the argument is that companies developing driving automation are already heavily regulated and will be put to the fire if mishaps do occur. We desperately need this automation if we do claim to not accept thousands of vehicle-related fatalities per year in our society, since better machine vision is always possible while better human judgement and reaction times are not.
Individually, no one ever wants to get into a car accident. Collectively, the desired number of accidents each year is zero. We’re not there yet because it’s hard to do, not because of acceptance.
So any new mechanism for accidents is going to come in for very heavy scrutiny. And by mechanism I don’t mean solely technology, but also human factors, which are huge issues when it comes to the current state of “self driving” cars.
Safety regulations is good but too much safety regulation is bad.
As individual I accept some level risk, I accept that there is some level of risk that I will be in accident, the trade off of 0 risk are too much.
Citation sorely needed. The few self driving cars currently on the road either require active human supervision or only operate in the absolute best circumstances so can’t be compared to human drivers who operate in all conditions and all types of vehicles.
If it's proven that a system performs significantly worse than a human then regulate it further.
It's not that crashes with only humans behind the wheel are uncommon. They're common as dirt! But they're not "high profile." So arguably NTSB's perspective (since they only do "high profile" crashes) is heavily skewed toward dramatic anecdotes compared to NHTSA who have to take into account all the crashes, high profile or not.
Data vs anecdote.
Critical thinking far too often goes out the window.
https://medium.com/@MidwesternHedgi/teslas-driver-fatality-r...
So you should compare against overall cars, not just "high end" luxury cars. Being electric has made them more expensive, but they're not really true luxury cars for the most part and are not driven by the same people or under the same conditions or under the same driving style as luxury cars.
Tesla is right up there but there are many snapping at its heels, especially from China.
I am hoping my next car will be a Chinese electric sports car. But, unlike a Tesla, one I own.
Is it the Porsche Taycan? Or the Porsche Taycan Cross Turismo? Or do you mean the Audi e-tron GT? Or perhaps the Lucid Air? Or maybe the Mercedes EQS?
https://www.caranddriver.com/reviews/a30874032/porsche-tayca...
https://www.cnet.com/roadshow/news/2021-porsche-taycan-epa-r...
We had gone a long time without any new successful car companies, and now maybe we can have two launched this century? Maybe even a few little boutique makers can make a go of it too... I really what Aptera is doing right now.
From the IIHS:
"The overall driver death rate for all 2017 and equivalent models during 2015-18 was 36 deaths per million registered vehicle years." (https://www.iihs.org/ratings/driver-death-rates-by-make-and-...)
Tesla's rate is the same as a Mazda 6 or Toyota Camry hybrid.
164 deaths, but only 6 on autopilot. While it might be safe to make statistical inference with 164 observations, no good statistician would make strong claims based on just 6. There is no evidence one way or the other for the impact of autopilot on fatalities.
I'm curious to know if "autopilot cut off automatically a few seconds before" counts in the first or second category.
Compare it to a Ford Mustang. I bet you won’t. ;)
Putting it another way, is there evidence that some ADAS are better than others? Because that's another way to choose regulations.
The guy he plows into can’t.
Tesla’s driving technology seems good enough, so I can picture guys whom are tired, or have been drinking, thinking I’ll take a calculated risk tonight.
Once a week I think about that tech guy whom plowed into a bollard on 101, or the guy sleeping coming home from Vegas.
I’m glad the NTSB is being very prudent. This is coming from a guy whom hates most laws, and believe society would be better off without so many regulations.
But as a software developer myself..there is no WAY I'd let the car drive itself. I'm a pretty decent developer, and maybe Tesla's developers are better, but knowing how many mistakes I make in code...just hell no.
Just don't be one of the free beta testers :) When a new version of anything gets released, I give it a year or two. By that time, stackoverflow answers get magically generated, bugs fixed, articles written and trade-offs articulated.
The currently enabled features require active driver supervision and do not make the vehicle autonomous. The activation and use of these features are dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience, as well as regulatory approval, which may take longer in some jurisdictions. As these self-driving features evolve, your car will be continuously upgraded through over-the-air software updates.
Edit - Here's a link to the old FSD fine print. They say the system is designed to take trips with no input from the driver, but they don't guarantee that, which to implies FSD is just meant to be good level 2 driver assistance.
https://web.archive.org/web/20180806143159/https://www.tesla...
To be fair, Tesla could consider stuff like TACC to be level 3+ (no beta status, no disclaimer about always paying attention, etc), but it's definitely not level 4 because they require driver input and nag then disable without it.
The robotaxi/Tesla network idea was separate from EAP/AP and FSD and was only mentioned on the Tesla website for a short period of time with even more fine print/legalese (it was pie in the sky, which I imagine is why they took it down).
You can't say "the sky is green[1]" and then in footnote [1] say "by green, we mean blue." That's not allowed, and the FTC is supposed to prosecute this. You can say "the sky is cerulean[1]" and then note that although cerulean can be categorized as either green or blue, in this context you mean blue. But calling the feature "full self driving" when it is not full self driving is more akin to outright calling the sky green.
https://www.forbes.com/sites/ericgoldman/2016/02/29/why-the-...
Now Tesla on the other hand, absolutely falls into the realm of material harms done…
I didn't realize the settlement was thrown out...or that a settlement can even be thrown out!
It's not supposed to be precise down to the millimeter, but it's quite literal.
If you're making fun of the idea, I still don't get it, because it's not like subway was trying to make them not be 12 inches.
When did they make that claim, since it's not in either of the linked articles as far as I can tell?
And that was immediately followed up by a sentence talking very reasonably about slight loaf-to-loaf variation, which is the real argument being made there and elsewhere and in court.
> What more are you expecting out of this?
I expect that "legal position" means something put out by a lawyer, not a facebook post.
I expect that the actions the company takes and the things their lawyers say take precedence over a facebook post.
I expect that if it's presented as an important reason subway gives then it shouldn't be taken out of context.
The judge ruled on the settlement, nor the actual merits of the case.
I only called it their legal argument as you did not understand what I was saying, I probably could have worded that better. It was their stated position on the issue, and is very similar to the "fully self driving" being discussed.
I am done with this, if you want to think I oversold Subway's claim, I disagree but fine.
I think if people were pulling "fully self driving" out of some random facebook post that discusses the capabilities in the next sentence, then the complaint would be something to sigh at and ignore.
Maybe we need to stop getting mad at the method and start getting mad at the companies doing the lying.
I've said repeatedly I was ready to leave this country, but only in the last 2 years have I begun actually making plans to leave.
I'm voting with my feet, but where to go? Shouldn't take years of careful planning to just up and go somewhere else.
Or as George Carlin called it: bullshit.
"The success of these AVs depends on the driver completing a monitoring task that requires sustained attention; however, humans generally perform poorly in the role of monitor. Also, if the automated control system behaves consistently and reliably for prolonged periods, the user of that system can become complacent about its operation and may not respond appropriately when a situation requires him or her to act."
The NTSB people see this clearly because they investigate aviation accidents. Over reliance on automation is a known problem in aviation. The aviation systems are pretty good, but they are not yet good enough to be the pilot in command. Pilot training covers this in detail. Drivers are not trained that way.
[1] https://www.cnbc.com/2020/07/14/tesla-autopilot-self-driving...
I would have a hard time arguing against requiring a few hours of class time learning about exactly what the systems do and what pitfalls there are because those things are not at all like what anyone has ever been taught in drivers ed before.
There's a certain similarity to Tesla autopilot crashes. Everything was going just great until Tesla's automated system failed to detect a big solid obstacle. Tesla drivers need to be very aware of that failure mode.
It's amazing that Tesla is allowed to ship that. It's Musk's "we don't need LIDAR, we'll just kill people instead".
> The person in the driver’s seat is only there for legal reasons. He is not doing anything. The car is driving itself.
[1] https://www.tesla.com/videos/autopilot-self-driving-hardware...
I mean, they are not completely wrong. Although not perfect. FSD Beta is still pretty impressive. There's no denying that:
"FSD Beta," which incidentally they told CA DMV was emphatically not full self-driving[0], hasn't been around for the vast majority of time that that video has been online so using it as an excuse is disingenuous. Plus Tesla's own words in their emails disprove this anyway.
[0] https://www.plainsite.org/documents/242a2g/california-dmv-te...
The videos that I linked are also from Dec 2020 and Feb 2021 respectively. Not from "from the last six months".
Here's one from 5 days ago (Mar 8, 2021): https://www.youtube.com/watch?v=zafOEZrKZ1c
I linked to a raw PDF inside a viewer containing Tesla/CA DMV's emails directly from a FOIA request-response. Are you claiming the site forged these emails?
> The videos that I linked are also from Dec 2020 and Feb 2021 respectively. Not from "from the last six months".
That's literally within the last six months..? Pointing out that they're even newer only strengthens my case that comparing them to a 2016 video is disingenuous.
Have you read the actual definition of what constitutes as an autonomous mode/autonomous vehicles. As defined by California law?
From CA SB-1298 [1]:
38750. (a) For purposes of this division, the following definitions apply:
(1) “Autonomous technology” means technology that has the capability to drive a vehicle without the active physical control or monitoring by a human operator.
Autopilot and FSD Beta requires a human to supervise it. In fact, it requires you to respond to the prompts and will actually disable itself for the rest of the drive if you keep ignoring it.
Here is Tesla's response to the CA DMV's query:
"Currently neither Autopilot nor FSD Capability is an autonomous system, and currently no comprising feature, whether singularly or collectively, is autonomous or makes our vehicles autonomous."
From what I can interpret. It doesn't fit CA DMV's own definition. Making Tesla's statement factually correct.
If you don't think that's the case. How so? I'm curious how you see it differently.
[1] https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml...
While I would agree the FSD path is a dangerous one, if they crack it they also instantly justify their share price.
An example of this would be the "obvious" superiority of micro-kernel OS designs compared to monoliths. Which is not at all how things actually panned out with Linux - but the problem was never about the architectural choice, it was about the fact that OS designs are hard - and you need to do the work to solve all the little problems which crop up with your big idea.
FSD is essentially going through the same thing, but the stakes are higher - FSD can kill people as a basic part of its function. We were always in a countdown till we started getting the first self-driving casualties.
This is not of course to suggest that what Tesla is doing right now is okay, but I'd sure be cautious trying to read the winner of the next 20 years into it.
I agree though. At the moment I’d rather just have a solid car with good adaptive cruise control.
How many elderly folks care about sub 4 second 0-60s times?
Well, this is, actually, quite understandable. Tesla and many investors in the company, especially Cathy Wood and her ARK Invest, assume high correlation between Tesla's future success or even survival with the success of the company's autonomous driving strategy, which, they argue, could win the lion's share of the global robo-taxi, robo-delivery and ride-hailing markets (with selling a relevant subscription service to future owners of FSD-enabled Tesla vehicles being IMO a secondary and insignificant revenue stream). This can be easily illustrated by reviewing some of the relevant ARK's open research materials[1-4]. Elsewhere on the Internet, a notable coverage of FSD that I'm aware of (though I haven't yet had a chance to review) is provided by Tesla investor and popular YouTube blogger Dave Lee, who offers multiple - as of today, I've counted five of them - detailed video interviews on Tesla's FSD on his YouTube channel "Dave Lee on Investing"[5].
[1] https://ark-invest.com/articles/analyst-research/tesla-ride-...
[2] https://ark-invest.com/articles/analyst-research/ride-hailin...
[3] https://ark-invest.com/articles/analyst-research/autonomous-...
[4] https://ark-invest.com/articles/analyst-research/tesla-price...
[5] https://www.youtube.com/channel/UCi2G5aQTN0YwGIfnt1u7Nbg
Except Waymo has better tech, is working toward a similar goal, and has a stronger balance sheet. Google has a history of dropping the ball on opportunities, but unlike cloud computing, they're already a leader in this space.
People want Tesla to succeed and like the company's story, but it basically requires every other automaker to fail in order to justify its valuation. Or Bitcoin to hit $500k.
Do I think people need to step in and force Tesla to dial back their advertisement, eh... idk, maybe, maybe yes?
To me, it’s pretty clear what “FSD” gets you these days (spoilers: not a lot). I suspect the average Tesla customer can also read. I also think reckless drivers will be reckless regardless of law or guidelines.
I think the increasing price for “FSD” and calling it fsd is all marketing designed to get people to FOMO into it (knowing that it might take a while).
What I would like to know is, if I pay “10k for FSD Someday” will this be applied to a future Tesla I buy that will actually be capable of FSD?
Peter Thiel's support for Donald Trump, alongside the army of Musketeers, continues to pay dividends, I guess.
All other questions are second order, and almost besides the point unless we have reason to suspect regulation of self-driving will be more efficacious than other safety regs.
Now, I've seen stats either way, but this article doesn't even address it. So what are people taking away from it?
If the same amount of people die except it’s nuns and orphan busses instead of drunks and other unsafe drivers then the “good” drivers odds just went the wrong way.
I will be less sceptical about fsd when I see elon musk taking a blind fold amd his kids on the rear seats of a fsd Tesla in snow covered new york heavy duty traffic or an unmarked, slippery and snow covered road in a storm with 20feet visibility taking a 5 hr ride. Should not be a big deal for someone who announced a coast to coast trip without intervention. What happened to that anyway..?
As someone not in a Tesla that’s what’s important to me.
As to the actual drivers/passengers that’s a more complex question. It’s likely going to continue killing some people who would otherwise survive, but nobody is forcing them to use the system while otherwise safe to drive. On the other hand others will only turn it on when tired etc which is likely where the net benefit comes from.
Can you substantiate this statement with discernible data? We recently did a road trip in Oregon. Even with an older version of autopilot (AP1) on our Model X which only has one front facing camera. The drive with having it engaged while closely supervising was safer.
There was really dense fog with long stretches of windy roads. I was having such a hard time that I had to slow down considerably. I was probably were pissing off local drivers behind me.
Meanwhile autopilot made quick work of the task without any disengagements. The latest version (AP3) is already considerably better and actually provides practical value.
I have questions. Were the NHTSA's policies regarding ADAS different under Obama than Trump? Are they different under Biden than Trump? If so, what changed? If not, why mention it?
More hate, more clicks. People hate Trump. Therefore editors want more clicks.
Their driver assist and safety systems have rarely worked and will always trigger false alerts.
Tesla on the other hand have actually saved me from a few potential frontal, rear and side swipe accidents. Before I even realized what was happening.
I wasn't referring to "self driving" capabilities. I was mainly referencing DISTRONIC PLUS®, Lane keeping assist(LKAS), Automatic emergency braking (AEB) and collision avoidance.
We owned 2 MB sedans and an SUV and all had the same issues.
> I got Mercedes because when there’s a serious accident, I have a very low risk of death
I definitely get where you're coming from. There just isn't enough data available for Tesla yet from the IIHS. At least from their latest report dated May 28, 2020. I hope it comes out soon.
What I do know is that the Tesla Model 3 scored high (if not the highest) marks as tested by multiple orgs such as NHTSA [1], Euro NCAP [2,3], IIHS [4].
[1] https://www.nhtsa.gov/vehicle/2020/TESLA/MODEL%2525203/4%252...
[2] https://youtu.be/z3cqj5AAP5I?t=27
[3] https://www.euroncap.com/en/press-media/press-releases/euro-...
[4] https://www.iihs.org/ratings/vehicle/tesla/model-3-4-door-se...
The Tesla Model X was also "The First SUV To Receive A Perfect Crash Test Rating"
https://mashable.com/2017/06/13/tesla-model-x-safety-rating/
https://www.tesla.com/blog/model-3-lowest-probability-injury... (first party, but sourced with links)
Having owned 3 different models now, I would disagree.
Regardless, even without autopilot. The Automatic emergency braking, Collision Avoidance, and Obstacle Aware Acceleration saved me countless times. Those are also all features that come standard with every car.
Meanwhile, other manufacturers will charge money for advanced safety features. Even basic autopilot made my drive along windy Oregon roads with dense fog much safer.
Autopilot also pretty much eliminates driver fatigue on long road trips.
What number of deaths will be due to slower progress due to needing to comply with whatever the NTSB believes will save lives?
But the end to Telsa throwing out unreliable Not-Even-Close-to-FULL-SELF-DRIVING is to sell cars--not to usher in a glorious era without traffic collisions.
I find this hard to square with a large segment of society just matter-of-factly espousing "equity", which is a less obvious way of saying what was expressed in the past as "equality of outcome".
Of course, through societies of the past experimenting, it was largely recognized as only possible through systems that end up being miserable and authoritarian.
So it's easier to push it again to a new generation under a vague, benign-sounding new term.
For example an equitable approach to criminal prosecution is that people are exonerated or convicted based on the best available evidence, with no variance attributable to their ability to afford a lawyer (an equitable approach enshrined in our Constitution but still a challenge for society).
And an equitable approach of salary is that people are paid in proportion to their contributions to the business, with no variance attributable to the color of their skin (for example).
We have to measure inputs and outcomes to infer the state of equity because we don’t have a reliable deterministic model of society.
4 = 4
If one team wins the Super Bowl, and one team loses the Super Bowl, then those teams are not equal; that is not equality. But if the game was contested fairly, with no cheating or corruption, then the outcome is equitable, even though it’s not equal.
In the U.S. what we generally strive for is equality as an input and equity as an output. If you hire two people, each should have an equal chance of contributing to the business, i.e. neither should be inhibited by irrelevant factors. But it’s expected and ok if one contributes more and advances farther. If it’s based on a fair measurement of their performance, that unequal outcome will still be equitable.
That’s why you’ll often see folks talk about “equality of opportunity.” Not equality in general; equality as an input.
No I think you're loading the meaning of equal. No one would say "you and I are equal" and expect that we're clones.
> That’s why you’ll often see folks talk about “equality of opportunity.” Not equality in general; equality as an input.
People talk about equality of opportunity because that's the circumstances that people concluded recognize human dignity and autonomy.
As soon as you begin the path of equal outcomes, you're talking an expansive micro-managing state optimizing for variables they scarcely understand.
And fundamentally there's a lack of faith in people as capable human beings that can manage the affairs of their own life and work to lead the life they want by their own endeavor. It's paternalism.
In other life and death fields, like medicine, the promise of saving lives is made all the time, but even for great benefit we do not cut corners to prove benefit more quickly.
Given the way that automobiles have little regard for life, I would rather that we take our queue from fields like medicine that have bothered to deal with the ethics and the cost benefit analysis.
Similarly, your claims about COVID reactions not weighing the cost benefit analysis are unfounded, at least that I have seen. I have seen many self-motivated parties making ridiculous, absurd, and harmful suggestions, but nothing that could be taken intellectually seriously. It's just people scamming for their own personal gain at the expense of others' lives.
Can you give one example of a government or government agency that has justified whether to shut down based on expected value of quality adjusted life years?
As if there’s a simple “spend lives now to save more lives later” knob.
But when it comes to driver alertness monitoring they pick the easiest to fool solution. I suspect it's intentional. Eye tracking would prevent their users from misusing autopilot. It would kill the hype.
They don't have LiDAR. So this statement is clearly false.
This separation of responsibilities has contributed to a culture gap between the agencies. As the agency responsible for writing regulations, NHTSA has to trade safety off against other considerations like economic costs, the lobbying clout of automakers, and the risk of consumer backlash. In contrast, NTSB's rulings are purely advisory, which frees the agency to doggedly advocate strong safety measures.
(NTSB is the one “blasting” Tesla). It’s hard to be completely sympathetic to a system-safety-at-all-costs viewpoint, in that slowing down the development of actual full self driving with superhuman performance by even a year due to increased friction on making changes has severe hidden costs - 38,000 are killed every year in the US alone in car-related incidents, and many more are severely injured with long lasting repercussions. And many of those are in the prime of their lives, so it’s worse than an apples to apples comparison with death counts from many ailments would suggest.
I think the NHTSA letting things develop on their own for a bit is actually the smart move here, all things considered.
That said, Tesla’s marketing strategy is pretty borderline.
I’m not sure if this is it, but I wouldn’t be surprised to see this administration try to reign him in. And when it happens, I wonder how news media and society will perceive it. Will they compare it to the relationship between China and Jack Ma? Or will they view these actions as virtuous, moral, and therefore justified?
No. It is distrust of such a erratic self loving obsessive fool.
Yes he is very smart, and is capable of assembling teams of fantastic engineers. But he cannot shut up. If it is tweeting jokes about taking Tesla private, slandering peopel who are doing the dangerous work and do not want his foolish interference (pedo guy), advocating (Dog help us!) Bitcoin. The lawsuits against whistle blowers. And here: his tolerance of fatal failures in his technology in favour of marketing and cash flow.
"aggressive disregard for the norms of society": If it were thought through and considered it would not be such a problem. If he were less intelligent, less driven, it would be less of a problem.
He is a fool. But a fool with a big stick. And that is a very dangerous thing
[1] https://www.the-sun.com/news/126445/who-is-vernon-unsworth-t...
I think we can acknowledge that people are complex animals and that someone can be amazing at some things and terrible at others.
I think Elon regularly fails to meet my expectations of how we should treat other people. He regularly exceeds my expectations as the CEO of a launch services company.
I said Elon was "just a mean, vindictive person", but I think that's far broader than I should have said. I think that sentence would more fairly read "capable of being a mean, vindictive person".
It wasn't an isolated incident. Musk called the man a pedophile several times over the course of weeks and months. Millions of people's only exposure to the person who helped save kids' lives was seeing Musk call them a pedophile, over and over again, reinforcing the accusation.
Also, I have never smeared someone like that, and would never consider doing so at all, let alone just because they hurt my ego.
He was enjoying the limelight and was pretty hostile to anyone/anything that was taking the attention away from him. He also wanted to make money from it.
Most shockingly. He said "I will make the divers sugger [sic] for leaving us out of the loop "
Source (Exhibit 7, Page 49): https://www.courtlistener.com/recap/gov.uscourts.cacd.723137...
When he heard that Elon was offering SpaceX engineering resources to the cause. He said on a video interview to take the submarine and shove it where the "sun doesn't shine" while maniacally laughing.
These are all facts that came to light from his lawsuit. Which by the way he lost:
https://www.wired.com/story/elon-musk-wins-defamation-suit-b...
https://www.theguardian.com/technology/2019/dec/06/elon-musk...
https://www.courtlistener.com/docket/7887513/vernon-unsworth...
https://law.justia.com/cases/federal/district-courts/califor...
You must be responding to a strawman argument that I never made. Please show me where I claimed anything other than "he helped save children's lives".
> He was enjoying the limelight and was pretty hostile to anyone/anything that was taking the attention away from him. He also wanted to make money from the fame.
Considering that this is exactly what Musk did, I find it pretty ironic that his fans have no problem projecting that upon people that Musk smears.
> He said on a video interview to take the submarine and shove it where the "sun doesn't shine" while maniacally laughing.
Musk forced himself into a high stakes situation where literal children's lives were at risk. Instead of listening to the experts and officials executing the rescue effort, Musk turned the tragedy into a PR event and injected himself and his companies into a situation that he both wasn't wanted in and was asked not to participate in.
When celebrities get involved with disaster scenarios, especially when they are not experts in themselves, they interfere with rescue efforts because officials often trust celebrity advice and interference over that of experts[1].
Had officials delegated to Musk's celebrity and his impractical "submarine", the children would have died. Why? Because the submarine was physically unable to navigate the sharp turns[2] that were required to reach the children and get them out.
Unsworth, an expert and participant in the effort, used much kinder words than I would have had Musk tried to make a disaster that affected me into a PR spectacle.
[1] https://www.bbc.com/news/technology-44779998
[2] https://ichef.bbci.co.uk/news/976/cpsprodpb/10C6A/production...
A bit ad hominem isn't it? How do you know I am a fan? Everything I said are facts from court documents. Unsworth was even hostile to actual rescuers/divers... Are they fans of Musk too?
How do you address this statement by Unsworth: "I will make the divers sugger [sic] for leaving us out of the loop"
I think it speaks volumes about what kind of person he is and his moral values. Trying to capitalize on a "high stakes situation" per your own words. He is not exactly the saint/hero you're claiming. Based on court documents. Not my own opinion.
edit: It would be nice if you indicated that you edited your above post after I had replied to it. Thanks.
edit: edit: Given that you edited your post, I will have to reply via edit.
> A bit ad hominem isn't it? How do you know I am a fan
You're aware that HN posts aren't deleted, right? I know because of your own posts[1]. Of 338 total comments, 140 of them are defending Tesla, which is what 41.42% of your total posts are about.
For someone who claims they aren't a fan, you sure have a strange fixation with defending the company, even making it a significant chunk of your HN career.
I implore anyone who thinks that's an "ad hominem" to point this out to click on this link[1].
[1] https://hn.algolia.com/?query=xedeon%20tesla&dateRange=all&p...
You are essentially discounting the judge/jury decision on this matter. If the facts were on Unsworth side. It would have been an ironclad case. I am operating from a factual standpoint. You seem to be on the emotional side because you keep going on this tirade.
I love how you keep steering the conversation away from facts and directing it towards me. How exactly are you providing value on this discussion? Also, hang on... There are people who actually consider HN as a "career"? How do I get compensated?
I don't matter. Facts do. If you can correct me where I said something that is not factual on this thread. Please feel free to do so.
> You're aware that HN posts aren't deleted, right?
Of course! This is the internet after all. You're digging through my profile to make a point? Now that's what you call infantile. Why don't you address what I wrote with your counterarguments? You yourself seem to be on an emotional Tesla tirade [1]. Don't act all holier than thou.
[1] https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...
> I implore anyone who thinks that's an "ad hominem" to point this out to click on this link[1].
You just proved my point! What your doing is a sad attempt to rally the mob towards me. If people disagree, they can just downvote my comments. But I do try to be factual as best that I can.
Zip2 Corporation
SpaceX
Tesla Motors
SolarCity
The Boring Company
Hyperloop
Open AI
Neuralink
Musk is not your real-life Tony Stark, he's a real life Hugo Drax.
Why such contempt? To infer that Starlink, Tesla and SpaceX are not positive achievements for humanity is a bit cretinous.
More importantly. What positive achievements have you done? The likely answer is that Musk or any founder which company is publicly traded (not including social media ones for obv reasons) are a net positive for the world compared to you.
I find it fascinating how people watching from the sidelines love to dunk on entrepreneurs. They also seem to be always angry at the world.
The quote "Comparison is the thief of joy" by FDR has never ringed more true.
If you think the only motivations are envy/political/ideological, you are just underestimating the number of people who don’t think like you.
Yeah I definitely agree, some of the hate for Musk comes from hate of capitalism and of the wealthy. I think that some of it also probably comes from his flirting with 4chan-esque trolling, like his talking about the red pill, etc. I think it's funny and refreshing but some cultural leftists get triggered by it.
I think you can tell I'm not exactly an Elon Musk hater, but that said... let's get real, the guy is a con artist. He has accomplished a lot of genuine things but he is also a con artist.
>I think the emphasis on his character flaws - which we all have - is disproportionate and frankly totally irrelevant considering the scale of his achievements.
He sells driving assistance technology as "Full Self-Driving". I give him full credit for his genuine achievements but I'm not going to ignore that he is also a con artist.
No. That's a false narrative. He arrived in North America with $2,000 worth of savings a backpack and a suitcase full of books. He also paid his way through college.
According to Ashlee Vance's book: "Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future"
However, his father didn't like him wanting to go to the US so he didn't pay for it. Musk had to work threw university.
Later his father invested some money into Zip2.
> He sells driving assistance technology as "Full Self-Driving". I give him full credit for his genuine achievements but I'm not going to ignore that he is also a con artist.
If you actually buy the car you see an exact description of what the feature does currently, what they think it will do in a couple of months and that you will get updates in order to get full self driving.
The money is not counted as revenue for the company because they have not delivered the features.
See from the website:
Full Self-Driving Capability
$10,000
Navigate on Autopilot
Auto Lane Change
Autopark
Summon
Full Self-Driving Computer
Traffic Light and Stop Sign Control
Autosteer on city streets
To me this sounds like an unproven and spurious claim that regulation will offer more benefit relative to letting a company like Tesla innovate freely. I don’t buy it, and it feels like they’re trying to fix a problem that doesn’t exist.
Do you have a citation for that?