Slashdot Mirror
Tesla Issues Strongest Statement Yet Blaming Driver For Deadly Autopilot Crash (abc7news.com)
Tesla has released its strongest statement yet blaming the driver of a Tesla Model X that crashed on Autopilot almost three weeks ago. The driver, Walter Huang, died March 23rd in Mountain View when his Model X on Autopilot crashed headfirst into the safety barrier section of a divider that separates the carpool lane from the off-ramp to the left. Huang was an Apple engineer and former EA Games employee. ABC7News reports: Tesla confirmed its data shows Walter Huang was using Autopilot at the time of the crash, but that his hands were off the wheel for six seconds right before impact. Tesla sent Dan Noyes a statement Tuesday night that reads in part, "Autopilot requires the driver to be alert and have hands on the wheel... the crash happened on a clear day with several hundred feet of visibility ahead, which means that the only way for this accident to have occurred is if Mr. Huang was not paying attention to the road." The family's lawyer believes Tesla is blaming Huang to distract from the family's concern about the car's Autopilot.
Here is the full statement from Tesla: "We are very sorry for the family's loss. According to the family, Mr. Huang was well aware that Autopilot was not perfect and, specifically, he told them it was not reliable in that exact location, yet he nonetheless engaged Autopilot at that location. The crash happened on a clear day with several hundred feet of visibility ahead, which means that the only way for this accident to have occurred is if Mr. Huang was not paying attention to the road, despite the car providing multiple warnings to do so. The fundamental premise of both moral and legal liability is a broken promise, and there was none here. Tesla is extremely clear that Autopilot requires the driver to be alert and have hands on the wheel. This reminder is made every single time Autopilot is engaged. If the system detects that hands are not on, it provides visual and auditory alerts. This happened several times on Mr. Huang's drive that day. We empathize with Mr. Huang's family, who are understandably facing loss and grief, but the false impression that Autopilot is unsafe will cause harm to others on the road. NHTSA found that even the early version of Tesla Autopilot resulted in 40% fewer crashes and it has improved substantially since then. The reason that other families are not on TV is because their loved ones are still alive."
Here is the full statement from Tesla: "We are very sorry for the family's loss. According to the family, Mr. Huang was well aware that Autopilot was not perfect and, specifically, he told them it was not reliable in that exact location, yet he nonetheless engaged Autopilot at that location. The crash happened on a clear day with several hundred feet of visibility ahead, which means that the only way for this accident to have occurred is if Mr. Huang was not paying attention to the road, despite the car providing multiple warnings to do so. The fundamental premise of both moral and legal liability is a broken promise, and there was none here. Tesla is extremely clear that Autopilot requires the driver to be alert and have hands on the wheel. This reminder is made every single time Autopilot is engaged. If the system detects that hands are not on, it provides visual and auditory alerts. This happened several times on Mr. Huang's drive that day. We empathize with Mr. Huang's family, who are understandably facing loss and grief, but the false impression that Autopilot is unsafe will cause harm to others on the road. NHTSA found that even the early version of Tesla Autopilot resulted in 40% fewer crashes and it has improved substantially since then. The reason that other families are not on TV is because their loved ones are still alive."
If it was a clear day with several hundred feet of visibility, there is no reason for Autopilot to steer the vehicle into a concrete divider. What good is it even if they say you need to keep your hands on the steering wheel? It doesn't sound very auto to me.
Tesla blames dead driver. Dead driver's family blames Tesla. Who is really at fault here?
I think the four-year-old girl is right: Why not both?
I'm less concerned whether the "autopilot" was the cause of the crash than the fact Tesla knows all about your driving habits. I think it is super creepy if the car company knows when your hands are on the wheel or not.
[Emphasis mine] Hands not on the wheel, a clear day with plenty of warnings to pay attention it's like he purposely wanted to crash.
Auto-copilot would be more appropriate for the way Tesla documents its functionality. As with any slashdot topic, naming can be the hardest part. Autopilot may sound better, but its deceptive.
People die while driving to work. Using your argument, no one would ever get into a car.
Yes, Tesla's Autopilot isn't perfect, and its capabilities may be exaggerated, but I believe that, overall, drivers using Autopilot are less likely to get into an accident. Isn't that the real measure?
Tesla's crash rate dropped 40 percent after Autopilot was installed, Feds say
Don't let the perfect be the enemy of the good.
The real "Libtards" are the Libertarians!
I don't know, I've seen Telsa autopilot reaction videos on youtube (https://www.youtube.com/watch?v=RjGe0GiiFzw), and when they're right, they're dead on right. Their instant reaction is so fast, to human perception it's as if they're predicting the future. I'm positive there are already cases where autopilot has prevented deaths. Unfortunately, I do see the argument that -- both legally and morally -- if you save 100 lives but are at fault for 1, you're still at fault for 1. Tricky situation.
Tesla should be issuing challenges and driver should respond correctly, if not it should pull the car over and stop.
If alert driver is a necessary requirement for safety, the system should check for alertness and stop the car safely if the driver is not alert. It is weaseling out if it allows the car to stay on auto pilot even after its request for manual take over is not honoured. But it knows the appeal of auto pilot will be greatly reduced if it enforces alertness rules
This is why I did not order autopilot when my Model 3 offer came through last Sunday. I am a great supporter of Tesla but the auto pilot is misnamed, and promotion of its use is not correct.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
No stats? Didn't you see the link I put in my post?
The real "Libtards" are the Libertarians!
"The crash happened on a clear day with several hundred feet of visibility ahead," ... which makes you wonder how the hell the computer missed a farking wall in the middle of the road.
Tesla blames driver for using the Autopilot in exactly the way you'd expect 90% of Autopilot users to use it.
I stole this Sig
I own an Tesla S. I use autopilot daily, itâ(TM)s awesome.
However you do have to know what it is and what it is not capable of and you do have to be attentive because it can get itself into trouble ( today, for example, it didnâ(TM)t want to let a bus into my lane - the bus came in anyway )
All that said calling the thing autopilot is what gets Tesla in trouble. Itâ(TM)s more of a âco-pilotâ(TM)
Surely if Tesla demands that drivers keep their hands on the wheel at all times that the autopilot is engaged then they should have a sensor for this and disengage the autopilot whenever the driver releases the wheel -- as a safety measure.
The fact that they don't do this is a clear indication that they really do expect people to take their hands off the wheel and use autopilot as if it were perfect. Stop passing the buck Tesla!
If the autopilot is unsafe around barriers like that, it should refuse to operate around those barriers. If it can't be made to recognize those situations, it should not be used at all.
http://michaelsmith.id.au
Does anyone know if Tesla is using a bot to write their Press Releases as well?
The following:
The reason that other families are not on TV is because their loved ones are still alive.
Does not sound like something a human PR Professional would write.
I stole this Sig
It seems to me that the only point of having an autopilot would be so that you could take your hands off the wheel and not pay attention to the road. This is sorta-kinda-an-almost-but-not-quite autopilot that works ok most of the time but has failure modes involving death and / or dismemberment. Who the hell would sell a half-assed, half-baked "feature" like this?
It seems to me that the only point of having cruise control would be so that you could take your feet off the pedals and not pay attention to your speed. This is sorta-kinda-an-almost-but-not-quite cruise control that works ok most of the time but has failure modes involving death and / or dismemberment. Who the hell would sell a half-assed, half-baked "feature" like this?
In other news - Tesla autopilot can't handle 6 seconds of autonomous driving on a clear day on a clear highway without causing fatal wreck.
I'm a pilot, been flying for 30 years, and I've flown with other pilots with varying skill and experience levels.
The most experienced pilot I've flown with never took his left hand off the control yoke. I watched him for hours while I was in the co-pilot and jump seats. He'd visit, configure radios, adjust power, but if his left hand ever came off that yoke it went right back on it as soon as the immediate task was done.
I'll drive my Tesla autopilot the same way that gray haired old pilot flew an autopilot, and with any luck I'll live to be just as old.
"Every time I see an adult on a bicycle, I no longer despair for the future of the human race." - H. G. Wells
To me it has to be either autonomous or not. If semi-autonomous driving requires you to be engaged and alert with both hands on the wheel, ready to take control at any time, then what's the point? How is it different from regular non-autonomous driving? Can anyone share their experience?
And yet your "AI" crashed and killed someone. What idiots. Enough with the AI BS. It ain't happening.
You're the only one stupid enough to all it an AI. Tesla has been very clear about their softwares abilities and limitations.
All "AUTOPILOT" does is conjure up images of planes flying themselves while pilots LEAVE THE FUCKING COCKPIT to go to the bathroom.
No pilot would ever do that exactly because the autopilot is just a simple program which only controls speed and heading. In the sky, with very few aircraft around you, it would be much safer to leave the controls than it would be in a car, on a highway, and yet aircrew always make sure that there is at least one pilot monitoring the controls at all times. If you hear "autopilot" and think "well, no humans required!" then you are badly misinformed.
That sounds stressful to me, not awesome.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
There are all kinds of issues with that stat. Look back to the last chain on Tesla, lots of problems with their test data. Anyway, it wasn't a study that was meant to show the safety of Tesla; they were investigating the first Tesla accident.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
I pointed out the problems with the name Autopilot at the time of some previous crash. If a geek like poor Mr Huang can be fatally wrong about its limitations (assuming that geeks are less likely to blindly trust technology) what chance do mere mortals have?
Totally agree that Autopilot is broken if it ignores a lane divider. If I had a Tesla I'd keep my hands on the wheel, my feet over the pedals and my eyes on the road.
Then after 10,000 miles I'd think, "Hey this Autopilot is pretty good" and trust it more and more until one day I'm doing 50, looking in the glovebox for my favourite Barry Manilow CD, well we all know what happens then.
I don't think Musk cares how ridiculous it is. He cares that it absolves him from any legal responsibility.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
Then something has been gained.
We need more autonomous car testing around critical mass rides! Particularly Uber, but all should participate, no matter how poorly funded and/or engineered. We should see at least a few pigeons driving heavy trucks.
John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
Yeah the name autopilot is one of the things that kills me about what they are truly selling. Tesla's cars are at best a level two self driving car, that is hands off only. You have to have eyes on and you have to give continual input to the system. It is adaptive cruse control, lane keeping, and auto parking. It has some guidance from GPS and on-board software, but the production car that you buy is nowhere near this crap. That video is clearly showing a level three car and the car is handling cleared intersections easily, something the current level twos would be suicide if you tried.
The autopilot is anything but. I'm totally pro-self driving cars, but folks need to know what they are buying and not have a hyped up product sold to them and they think it will do something it won't. Tesla cars are a hands off only car, period, the end. You have to keep eyes up, no matter what. Additionally, you need to know the absolute limits of camera/radar combos, that's right Teslas do not have LiDAR. Radar requires a calculation between differences in order to work. If traffic is stopping up ahead and the car in front of you that your Tesla is tracking suddenly pulls out of the lane to expose a car up ahead at a complete stop, your Tesla is going to ram full speed into that stopped car if you don't do something. That's because radar was tracking something and now it's not there. So the machine needs to recalculate everything, which if you're going highway speeds, you're going to end up dead before the car figures it out.
The cars need to see lines on the road. If the lines are iffy, you're going to end up dead. Traffic needs to follow a pace, it doesn't matter if it is start and stop, or if cars gracefully merge in and out of your lane. It just needs to follow a smooth flow to things and you slowly build up a feel for what's gradual enough and what isn't. If you don't pay attention to that, you're going to end up dead. If you are coming up on a change in the road's shape, like where two highways split off and you're in the lane closest to the split, you need to turn off autopilot and handle it yourself. Most of these kinds of things have really crappy indicators on the road that a split is happening and if you don't, you are going to end up like that dude. Dead.
Now if you think that level two automation is a half baked idea, that's cool. It sort of is, which is why everyone is aiming for that holy grail of level five. So perhaps maybe sit the sidelines till we get there? If what you are comparing to is level five, you're right, this shit is beta-level crap on crap. If you're talking about actual level two automation, the Tesla and all the other cars that offer level two are pretty solid. But people need to understand what they are getting themselves into and if that's not what you were expecting, then yeah, you shouldn't buy one. However, I also fault Tesla, since they post up videos like that one I linked and people buy their cars thinking, that's what they are getting which it isn't.
Absolutely. The system actively encourages the driver to pay less attention by its very design.
In other words:
"The victim was using cruise control. Our Tesla is not a self-driving car. Stop calling it that. Reliable self-driving cars do not exist."
Do. Not. Exist.
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
The quote from Musk and the graph both reference accident reductions with Autosteer. Neither call it Autopilot.
Tesla offers, per their manual:
Autopilot Tech Package:
• Traffic-Aware Cruise Control
• Autosteer
• Auto Lane Change
• Autopark
• Auto High Beam
Why is Musk and the graph specifically referencing Autosteer? I never trust data fully when specific and unexpected words are used. The omission of the word Autopilot implies......
"...one day I'm doing 50, looking in the glovebox for my favourite Barry Manilow CD, well we all know what happens then."
You want to kill yourself, and let the car help?
"National Security is the chief cause of national insecurity." - Celine's First Law
If you call it autopilot, it should be an autopilot. If it's a warning system, call it something else. Words mean things.
Your ad here. Ask me how!
Or go for a quickie. There was a pilot's program at my school and I knew several peoplein the program to join the mile high club. Autopilot was doing the flying.
If you hear "autopilot" and think "well, no humans required!" then you are badly misinformed.
I do think autopilot means no human intervention required, and I would be badly misinformed, but I would also be representative of the majority of people. That last point is the most important one. It's the point that makes Tesla's branding effective.
We're talking about branding and marketing. The dictionary meaning or the "real" meaning is irrelevant. Only what the majority of potential buyers think is important. That these explanations about the "real" meaning are necessary strongly indicates that the majority of people do not share these perceptions about what the word "autopilot" means.
I guess you don't know what a co-pilot or autopilot is then, since you have that the opposite way around.
And this is I think the problem, on each of these stories someone always chimes in on what autopilot on planes does and how that is very similar to what tesla's does. It should however be totally clear to tesla that what probably started out as a cute term from marketing is actually not the common understanding, they should come up with a new, clearer name.
Tesla is extremely clear that Autopilot requires the driver to be alert and have hands on the wheel.
I already have to do that. What's the point of buying this autopilot-that-isn't-really-an-autopilot?
Tesla's autopilot doesn't use human eyes. If I understand it correctly, it has a monochrome camera and a forward facing radar. If there is not enough monochromatic contrast between an object and its surrounding, the camera won't detect it, and if the object is at an angle and composition where it cannot return radar signals directly back to the car, the radar won't detect it.
How good the visibility is only affects the human, who is the one with the driver's license and the responsibility that goes with it.
That is the point.
1.25 million people die on the road every year, world wide (number from 2013, had it in my head due to unrelated research).
If Tesla has one millionth of the car market, expect on death per year. That would be statistical normality.
Assorted stuff I do sometimes: Lemuria.org
The family admits that the driver had had issues at that exact location. Why on earth would he use it there then? Why wasn't he paying attention near that spot? Why did he ignore the warnings? He was a programmer. He should have known.
This and no other is the root from which a tyrant springs; when first he appears as a protector - Plato (423 to 327 BC)
Too many people take 'Autopilot' literally.
Not an appropriate analogy. Cruise control does the one thing you command it to do, that is to keep the speed constant, and it will do it perfectly and with no exceptions, because that function is well-defined and is better done by a machine than a human. You needn't worry about maintaining the speed when cruise control is engaged.
On the other hand, you command the "autopilot" to take full control of the car without causing accidents, which is an undefined problem (and Tesla is very careful about not defining explicit boundaries), and then it might fail to do that even in the most favourable conditions, such as a straight motorway with perfect visibility. When the "autopilot" is engaged, you need to be constantly wary about the possibility of it doing an abrupt and incomprehensible manouver that might result in your death and the passengers' (this is what Tesla themselves are stating in their public statement when they blame the driver). It's not even remotely in the same domain of dangerousness as cruise control.
please cite a statistic that shows autopilot creating more accidents than there otherwise would have been. Tesla claims autopilot is statistically 40% safer than a human driver. do you have actual evidence to the contrary, or just overreaction to a hamdful of incidents (among literally millions of drivers and tens of thousands of non-headline human accidents)
i could live a little longer in this prison
non-autonomous vehicles have fatal collisions at much higher rates than autonomous vehicles do. i guess that technology is even further from being ready for prime time.
i could live a little longer in this prison
It's 2018. If you rely on your car's computer to keep you alive, you're a fucking idiot. You don't deserve to die, of course, at all, but don't be so stupid that you expect 2018 tech to work like it's 2040 already.
This is sorta-kinda-an-almost-but-not-quite cruise control that works ok most of the time but has failure modes involving death and / or dismemberment.
Well when you list the negative features and ignore the positive you could come up with that. Did you post that on your energy wasting home heating desk box, or on your radiation emitting fragile money wasting glass slab which rings during meetings?
Why would you own things like that?
That dies not stop people from thinking it. People also think hackers ad bad people. The drivers are not trained like pilots, as so many people think that autopilot means that the car needs no human, than that is what it means for non-pilots.
If it would mean the same, people using one in acar should follow similar lesssons as pilots do before using one.
Don't fight for your country, if your country does not fight for you.
Yet the environment in which an airplane autopilot operates is very different from the environment in which a car "autopilot" operates. When monitoring an aircraft autopilot and something goes wrong, you have dozens of second (if not minutes) to react. For instance, after AF447 autopilot disengaged, the fall of the aircraft lasted for about 3 minutes and 30 seconds, and the situation could have been recovered during most of them. In a car, the problem is very different - you only have a few seconds to react before you run into a wall or drive over someone. This is why comparisons with airplanes are not so much relevant.
Wasn't this one of the cases from last week where it said the autopilot was actively steering people towards concrete divider barriers? Where Tesla's response was: "There's nothing unsafe about this at all"?
It is very clearly an AI. Just not a good attempt, at least not yet.
It's not even a hands off car since they require you to hold the wheel. Otherwise it'll start yelling at you. They do market it to be capable of more than it really is. They talk about AI and self driving but it's not reliable to be marketed like that yet.
All that said calling the thing autopilot is what gets Tesla in trouble. ItÃ(TM)s more of a Ãco-pilotÃ(TM)
Why is it the Apple users who can't figure out how to get their browser to send what they type who can't figure out the difference between an autopilot and a copilot? An autopilot is an assistive system which will fly you straight into a mountain if you use it incorrectly. A copilot is an additional pilot whose responsibility is to take over for the primary pilot and actually fly the plane, preferably not into a mountain. The system is nothing like a co-pilot, and it is entirely like an autopilot. You literally could not be any more wrong.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Tesla's cars are at best a level two self driving car, that is hands off only.
Tesla cars are a hands off only car, period, the end.
"Tesla sent Dan Noyes a statement Tuesday night that reads in part, "Autopilot requires the driver to be alert and have hands on the wheel... " Not according to Tesla!
If the AI system doesn't think it can manage things anymore and the user is not responding to input, it should throw the hazard lights on and make an emergency stop.
The first problem is at the "if".
Seems that in some cases, the "Autopilot" is completely persuaded that it is on the correct course.
It genuinely thinks that "straight ahead" is the 100% correct answer to the problem.
In that case it will never fail the driver "Hey, I need help".
Again, it's an "autopilot" (see planes, boats, etc.) just a thing that automatizes some low-level work. The captain of the aiplane/boat/tesla should still keep focus and check that everything goes as it should (it's a "level 2" autonomy. The human is still constantly in charge 100% of the time. Simply the human doesn't *need* to actually interact with the controls 100% of the time. Most of the time, the vehicle could control itself on it own, BUT NOT unattended, human overwatch is mandatory).
And that's what Tesla is arguing.
Not paying attention "just because" autopilot is on, is almost Darwin-award-worthy (just as in a plane or a boat).
Though one might argue that Tesla isn't insisting clearly enough in their marketing material (cue in Elon making a presentation about dreaming that within a coujple of year you could summon your car to come to you)
and/or people make wrong assumption when they see the word "autopilot" (they don't think plane / boat with a captain still in charge, but somewhat think Knight Rider or other Sci-Fi setting).
Then the second problem :
Why the fuck didn't the car see the a huge block of concrete on its course ?
This thing should (probably have) a nice radar signature.
Most of the much more primitive FCAS currently on the street would probably see it and slowdown/stop or ring alarm/hit the break.
Some weird interaction is happening.
Some filtering gone wrong ? (radar system ignoring objects not moving relative to the street, in order to not over-react on each single guardrail ?)
Some precedence conflict ? (the camera system not seeing the lane diverging and overriding "No it's safe, I don't see an obstacle" ?)
That's an error on Tesla's side.
If this is a repeated problem, the system should disable the auto-pilot feature and refuse to let the driver use it. If they want it turned back on, they can write to Tesla and explain why they think that they should be allowed to be a colossal moron with a quarter million joules of kinetic energy.
At some point in time, we might see tiered driving license appearing, with a separate module to train drivers how to use driving assistance tools properly.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Yeah the name autopilot is one of the things that kills me about what they are truly selling. Tesla's cars are at best a level two self driving car, that is hands off only. You have to have eyes on and you have to give continual input to the system.
That sounds exactly like autopilot. If you're in a plane using an autopilot, you are legally obligated to remain alert and present. If you are on a boat using autopilot, if you don't remain alert, you will probably sail right into a rock or another vessel. What was your complaint again? Autopilot is exactly like an autopilot? Okay then! Moving on.
However, I also fault Tesla, since they post up videos like that one I linked and people buy their cars thinking, that's what they are getting which it isn't.
Should have linked it again, I'm not going back through your posting history to find it.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Not only do we make sure there is always one pilot monitoring (the Pilot Flying - or PF), if the pilot not flying (PNF) needs to exit the cockpit, a hostie with basic training will replace the PNF in the cockpit. This ensures there are always two people at the controls.
When the PNF returns, they replace the hostie.
We also have a rule that if the PF doesn't react to verbal or aural cues after 3 tries - the PNF takes over the aircraft. This includes both actual pilots, or the hostie with basic training.
Trying to compare a car to an aircraft is a bad idea - we have procedures for all of this stuff - and most procedures exist because people have died.
Sendmail is like emacs: A nice operating system, but missing an editor and a MTA.
You are quoting an invalid statistic. The 40% one has been debunked many times. The point is there are *no* statistics.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
The family admits that the driver had had issues at that exact location. Why on earth would he use it there then? Why wasn't he paying attention near that spot? Why did he ignore the warnings? He was a programmer. He should have known.
The car was on Autopilot... You know A-U-T-O-Pilot. The car should have driven itself whilst the attendant sat back watching movies on their phone.
That is the logic you can expect from end users. Warnings are just something to be ignored or at the very worst summarily dismissed. Autonomous cars are something that has been sold to them as a magic bullet to their driving woes. The end user fully believes that their time having to pay minimal attention to the road is at an end and that the car will automatically handle everything for them. Also it's going to eliminate congestion because they can go eleventy bajillion leptons per microfortnight whilst bumper to bumper and there will never, ever be any collisions.
Calling someone a "hater" only means you can not rationally rebut their argument.
I used to let buses pull out ahead of me all the time. Common courtesy for other road users.
Then councils started implementing bus lanes all over the place. So fuck 'em, no bus gets in front of me if I can help it.
But they don't have anywhere close to one millionth the car market. Furthermore, they don't have anywhere close to one millionth miles driven.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
What good is it even if they say you need to keep your hands on the steering wheel? It doesn't sound very auto to me.
So unless it completely takes over operation of the vehicle you think it is worthless? That's not what the system does.
Per wikipedia it is described as follows: "advanced driver-assistance system feature offered by Tesla that has lane centering, adaptive cruise control, self-parking, ability to automatically change lanes without requiring driver steering, and enables the car to be summoned to and from a garage or parking spot." Note that NOWHERE does it say it drives the vehicle for you or that it is a hands off system. The key words there are DRIVER ASSISTANCE. It is not a driver replacement.
Frankly I pretty much assumed that there would be some Darwin award submissions by idiots who can't figure out that they still need to drive the vehicle.
It is a totally unrealistic requirement that one be fully attentive while on autopilot. It is impossible and moots autopilot completely. This argument is just another step toward allowing machines to kill people. Musk knows this. He is not unintelligent, thus, he must be deliberately agitating for machines to be allowed to kill people so he can sell more machines. What do they call a callous, remorseless, conscious-less, manipulative person? This is the same guy who wants to send people to Mars to die for the greater glory of Elon Musk. You know what to call such a person.
E Proelio Veritas.
Comparing it to cruise control is stupid. Cruise control maintains your speed extremely well and doesn't ever fail catastrophically.
Tell you what. You get on the highway and put on a non-adaptive cruise control like the one in my car and let it maintain your speed while traffic slows in front of you. Let me know how that didn't "fail catastrophically" when you are done rear ending the car in front of you.
Every accident with the Tesla autopilot I've seen has been idiots relying too heavily on the system thinking it replaces the driver instead of assisting the driver. While it's a capable system, the person behind the wheel still has to use their brain and pay attention and take control of the car.
it should know for fucking certain that human oversight is required and safely bring the car to a halt if the human is not providing it.
So your proposed solution is that if the car detects an unattentive human it should engage automatic emergency slow-down procedure (as if there's an obstacle) ?
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
In many states, you are required to yield to buses coming back into traffic.
The other thing with aircraft auto-pilot is that you have air traffic controllers controlling the airspace. You're never close to another plane except during "critical operations" (i.e. takeoff/landing). If airplanes flew as close as cars drive, their autopilots would be borderline useless.
No! Your the fucking asswipes that don't get it. It's got AUTOPILOT, not Chauffeur. When you go into an airplane, the pilots sit in front, don't sleep, and watch the skies, the instrumentation, and the aircraft handling, the pilots are paying attention! That is how you operate with autopilot, you don't see the pilots both taking a nap or coming back to schmooze with the flight attendants.
And the number of boaters who end up in accidents with their auto-pilot is enough evidence that Tesla is being negligent by using the term.
Right so if you don't want somebody to think you are a bad person, don't introduce yourself as a hacker!
All Tesla vehicles produced in our factory, including Model 3, have the hardware needed for full self-driving capability at a safety level substantially greater than that of a human driver.
Did you read that? It says "HAVE THE HARDWARE" (emphasis mine). It doesn't say it is a functional self driving vehicle or that they have the software to make it work. Now granted someone who isn't reading carefully might be confused but that is clearly NOT a claim that they have a self driving system that is functional today. They just said they put the hardware necessary for that function into the car presumably to be enabled at a later date.
More fundamentally to me is the issue that said car should not drive straight into a wall at full speed without trying to slow down.
Wrong. The DRIVER should not drive straight into the wall without slowing down. It's not the responsibility of the car. The driver/pilot is the one responsible for its safe operation of a moving vehicle. Always. If they abdicate that role for any reason then the driver/pilot is the one who is negligent.
What blows my frigging mind though is that the car will drive into a stationary object with high contrast safety striping without attempting to brake.
Why should that blow your mind? Sensors are not infallible. Neither are people and humans do EXACTLY the same thing with some regularity.
Their "neural net" doesn't seem to be learning some important lessons quickly enough.
Here's a hard truth for you. Safety systems get improved by people getting hurt. it's unfortunate but true. Airplanes are very safe today but the reason they are safe is because a lot of people died to learn the lessons necessary to make them so. Cars are not any different in that respect. People will die to make driverless cars a reality. Many of those lessons are not obvious until after the accidents. As the saying goes the rules are written in blood. The real tragedy would be to not learn from these accidents so that we don't repeat them.
Tesla blames dead driver. Dead driver's family blames Tesla. Who is really at fault here?
Unless there are facts we are unaware of the blame is clearly the driver in this case. He failed to control his vehicle. It doesn't matter what the Autopilot system did or did not do. The driver is responsible for the safe operation of the vehicle. Tesla's autopilot system is a driver assistance system, not a driver replacement system. The difference between those is not trivial.
I'm not asking for perfect. If visibility was as good as Tesla says it was, why couldn't the car stay in its lane, and why did it steer into an obstruction?
While that is a question it is not the important question. The question is why the DRIVER didn't direct the vehicle to stay in its lane and avoid the obstruction. The tool is not at fault here and it seems clear he relied upon it inappropriately. While it seems clear that the autopilot system may have had a flaw of some description the responsibility is still on the driver to steer the vehicle in a safe manner.
It's nothing but smart cruise control and lane assist (that kills you).
Regular cruise control will kill you just as dead if you fail to operate the vehicle in a safe manner. It's a driver assistance system not a driver replacement system. It did not kill the driver. The driver's stupid and careless behavior killed the driver.
Will companies selling autonomous cars try to put the responsibility on a passenger if they happen to have a driver's license? Again, how a machine is supposed to be deemed responsible? And of course the manufacturer will always try and find ways of proving this is not an autopilot default. What does that change? This is not what responsibility is about. Who cares if you did everything "right"... when you're responsible of something, you're accountable. Period. Who will be responsible?
You can't really compare this to autopilots in commercial planes either, for several reasons. First, the pilot IS responsible in all cases, and an plane's autopilot is NOT supposed to be left unattended. Second, planes' autopilots are not designed to make planes avoid obstacles, but just to follow a programmed route and keep the plane flying at a specified speed and altitude. That's all it does. Autonomous cars have new challenges, both technically and in terms of responsibility, that no one actually knows how to deal with as of now.
Will autonomous cars' victims be considered as just collateral damage?
Look, human beings suck at vigilance tasks. "This is almost always OK, detect the one time in an hour that it's not"-- no one can muster the attention. X-ray screeners use something called the "Threat Image Protection System" which shows them pictures of bombs and guns and keeps them alert (it lets them know it's a test, but helps keep their mind in the "where's the gun in THIS one?" mode instead of "oh, look, another suitcase probably without a gun"). Even S&R dogs find trainers even in the middle of a search else they grow bored with the task.
Autopilots in transport aircraft come with a big master warning and caution system that lets you know about most of the classes of developing problems and are loud about it instead of relying on a flight crew to spot them, because even highly trained, professional flight crews are shitty at detecting changes in something that almost never ever changes. Having a system that avoids steering for the guardrail 99.99% of the time is a recipe for disaster, because it will build confidence and train people not to pay attention.
No! You're the fucking asswipes that don't get it. It's got AUTOPILOT, not Chauffeur. When you go into an airplane, the pilots sit in front, don't sleep, and watch the skies, the instrumentation, and the aircraft handling, the pilots are paying attention!
I was going to comment the exact same thing. I happen to be a pilot, albeit for fun only. When I fly my little single engine plane with auto-pilot, it keeps me straight and level, and on the heading that I choose. It does not look out for other airplanes, and I definitely don't have an Ground Proximity Warning System, or a Traffic Collision Avoidance System. So I look around and pay attention, as I should when flying VFR.
Auto-pilot is the same thing. It is a driver assistance function, not an autonomous vehicle function. As sad as this accident is, in my humble opinion the driver earned himself the Darwin Award for not paying attention at a spot where he knew that the function was unreliable AND disregarded signals from the function to take manual control.
That is, unless there were underlying medical conditions that prevented him from doing so. He wouldn't be the first "pilot" to have a heart attack behind the wheel.
I'm not a complete idiot... Some parts are missing.
The problem about the term "autopilot" as used by Tesla is that it implies something that it is not. All the verbiage in the owner's manual or displayed when you engage the feature isn't going to overcome this implication (consumers don't read all that stuff any more carefully than they read the TOS on websites they sign up for or the software license agreements).
Most people are probably familiar with term "autopilot" from its use in the context of aircraft or boats/ships. In its primitive form in those contexts, "autopilot" allows you to safely not pay close attention to maintaining heading and altitude and generally allows the human pilot to pay little attention (barring sudden failure of some sort) to what is happening in the next 10 seconds. This allows the pilot to allow themselves to be "distracted" by longer term tasks - is there turbulence ahead and should I fly around it, how's the fuel balance, what are the odds that ORD is on a weather delay, etc. This works in the the sky and water because there are not a lot solid objects, moving or stationary, nearby to run into for much of the trip and the stationary ones are on the charts and taken into account long before they are encountered.
Automobiles are quite different, in fact almost completely opposite... Virtually everything (drunk drivers, road debris, concrete barriers, stopped fire trucks, stalled cars) that the driver needs to worry about is just a few seconds away. Tesla's "autopilot" doesn't allow you to safely apply most of your attention to longer term tasks - instead, virtually all of your attention is still needed on the road to "double check" what "autopilot" is doing every 100ms. So, Tesla's "autopilot", used correctly, really doesn't allow you to do hardly anything different when it is on then when it is off - where is the "auto" in that?
Tesla: JUST STOP CALLING IT AUTOPILOT, call it 'safety assist' or 'collision avoidance' or 'deathguardian' or "grimreaper" -- almost anything but "autopilot". Also, be much more aggressive about making sure the driver IS engaged -- such as hands off the wheel for ANY period of time results in very loud warnings and almost immediate slowdown with hazards coming on and "autopilot" can then not be engaged again for 4 hours. (And maybe a small "stupid light" should be added that is visible to other drivers while you are in "safety assist" timeout to shame drivers who don't use the feature safely - social shaming goes a long ways and also warns others to be careful around your car because you're obviously an unsafe driver).
The point is that you are not lied to and told that your non-adaptive cruise control will magically slow you down if something gets in the way, so you keep watching the fucking road.
Show me where Tesla has lied about the capabilities of their system. I've never once seen them make the claim that the driver wasn't responsible for watching the road and taking control if the autopilot functions fail for any reason. If the system doesn't kick in early enough the driver should take control. Tesla has been quite unambiguous about this.
It is a fundamental flaw to have a system where you can sort of half watch the road most of the time and probably not have much of an accident unless you're unlucky.
Fortunately no such system is on the road in a production vehicle to my knowledge. The fact that Darwin Award candidates attempt to use them in such a fashion does not change that fact.
With normal driving, you basically have to concentrate 100% of the time. A self driving car would mean you had to concentrate 0% of the time. It is absurd to have something in between.
Anyone who believes these driving aids are a substitute for paying attention is mistaken. The purpose of driving aids is so you can focus your attention on what is critical to pay attention to and to help you do it better or with less effort. It also can provide an additional layer of security in case the driver misses something or becomes distracted. It is NO different than autopilot in a plane. The pilot still has to fly the plane - the autopilot system just helps them do it a little more effectively by partially automating certain tasks. But make no mistake the driver is 100% responsible for controlling their vehicle.
I mean, please, look around.
sabri writes:
Tesla however writes:
https://www.tesla.com/autopilot
"Full self-driving" sounds an awful lot like an "autonomous vehicle function" as you phrased it.
At some point, the car should have realized it was about to collide with something and applied the brakes. It did not. Fault = Tesla.
Apparently this is not on the programming flowchart: Not Crashing > Staying in Lane.
"Full self-driving" sounds an awful lot like an "autonomous vehicle function" as you phrased it.
Context: they're talking about the hardware. Software isn't there yet.
Also, "substantially greater" isn't "perfect".
I see a parallel between cigarettes and Tesla Autopilot. For both, the manufacturers know that there is a danger in the way the the products are likely to be used. Warnings are issued, and branding/advertising are used to contradict those warnings. There is no question that the consumers behaved in a way that directly led to bad health and death. The question is whether the companies have any liability in the way that they brand and market their products, or do the legal warnings absolve them of all liability.
Let's do the math. Let's focus on the USA, numbers are easier to get and Tesla has most of its business there:
Number of cars on the road:
https://www.statista.com/stati...
Number of road deaths:
https://en.wikipedia.org/wiki/...
2015 Numbers:
35,485 deaths from 263,610,220 cars.
1 death per 7.429 cars.
Did Tesla have more than 7429 cars on the road in 2016 in the USA?
According to this source:
https://cleantechnica.com/2016...
They sold (in the US) 26,566 cars in 2015 alone, and 18,480 the year before. So even by a conservative estimate, Tesla had at least 40,000 cars on the road in 2015.
Statistically speaking, 5 people should have died in Tesla cars in 2015 in the US alone.
If I remember correctly, either one person or no person died in 2015 (the first fatal crash was 2016, but I could be mistaken). In either case, that is far below the statistically expected value.
Q.E.D.
Assorted stuff I do sometimes: Lemuria.org
The hardware is there, the software isn't. Everyone has enough hardware to fly to the moon in their pocket. I doubt they have the software required though.
If you gave me a choice between a printer and a giraffe with explosive diarrhoea, i'll get my ladder and my raincoat