It wouldn't necessarily be cause to disengage the autopilot. Plenty of things could cause negative acceleration besides a collision. A hill, a patch of mud, etc. And presumably they didn't anticipate this kind of situation, so didn't include moderate slowing due to collision as a code path.
The car should still have started emergency braking at some close distance before the obstacle (when the radar data got very clear about that not being an overhead sign). Braking in case of an inevitable crash is useful to reduce severity of the crash. And it's a bad idea to release brakes after the crash either. So it still seems like wrong behavior to me.
Well of course it's wrong behaviour! But in the article itself it says the car never braked, as it did not see the trailer. Presumably it did detect some deceleration and compensated for it. I'm just saying that apparently it wasn't programmed to interpret sudden deceleration absent any other indicators as a possible collision. I agree that perhaps it should be.
"Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied". Did the model not have a forward looking RADAR?
The main problem with all of this, is that its not a true autopilot. In fact its the worst kind of autopilot. one that bails out at the last minute leaving you fuck all time to take action.
The thing I fear now is the toxic mix of fanboyism and excessive PR management reminiscent of the worst excesses of Detroit. The only way safety is improved is when hubris and bluster is put to one side.
The thing I fear now, is fearmongering by people who dont drive Teslas, havent spent time with Autosteer/autopilot and ignore that this was a freak accident, that probably would have killed the driver regardless of using autopilot.
> There's no reason why an entire sedan should fit under the trailer of a truck.
There is, it's called capitalism, and a safety skirt costs money.
The fix is, as in many cases, regulations. However, I believe that even the shown safety skirt leaves too much space below, and especially exposed tires, which is still dangerous for pedestrians and bikers.
A lot of those skirts also aren't quite as strong as they should be. Here's a comparison between one that fails testing and one that doesn't: https://youtu.be/gL8DQIAqZms?t=151
As for pedestrians and bikers, I really really like Europe's solution to that problem: Pedestrian and Bikers are illegal on the freeway. Big trucks are made strongly and forcefully impractical in cities, towns, and villages.
It's crazy that I live in downtown SF and there are huge freight trucks driving past my bedroom window every day.
With the pedestrians and bikers, I mainly meant when large trucks are driving in the cities to serve e.g. supermarkets - one I worked at, right in the center of Munich, had a huge 40-ton truck serving it each day.
It's dangerous as hell, but impractical to do any other way...
Maybe it was just my neck of Europe then. Trucks would only go to supermarkets on the big roads where there's strict physical separation between traffic and everyone else. They're also position on the edge of the city next to the ring freeway so that trucks have a short path from freeway to supermarket.
Deliveries inside the city are done with smaller lorries which are a bit safer maybe. If not safer, definitely more practical on tight inner-city streets.
I don't think it's practical for trucks to be designed as pedestrian safe as modern city buses are, so it's better to keep them away from pedestrians as much as possible.
Capitalism is the issue because the money is being saved by cutting corners on safety. In case of non-profits and the government people might actually sue for wasting of money because money is spent on stuff that is not required by law.
Capitalism is the issue because the money is being saved by cutting corners on safety.
Saving resources by skipping on safety features is not exclusive to capitalism. If anything, the correlation seems to run backwards - capitalist countries are generally safer (mostly, I'd say, because they're rich enough to spend extra resources on safety).
In case of non-profits and the government people might actually sue for wasting of money because money is spent on stuff that is not required by law.
Can you give an example of such a lawsuit ever having succeeded?
Good point. But I think it is too early to call this a 'freak accident', because we don't know enough.
Tesla have statistics on their side so far, but they still have not done a fraction of kilometres travelled by conventional means.
I hope the technology moves forward, but the technology is not there yet and won't be appearing on a majority of new cars for decades.
I predict - moreso than fear - that as more of these incidents occur (and I am confident there will be more), more people will lose their trust in self-driving cars.
Consider the following: The safest form of transportation is air travel. Yet despite the overwhelming evidence in air transport's favour, a significant portion of people are afraid to get on an aeroplane. Indeed, an aeroplane can practical fly itself (and have been able to for decades), yet few of us would feel safe if there wasn't a pilot on board.
Surely the same fears will soon apply to self-driving cars?
Part of the reason why people are afraid of flying is because you are locked inside an aluminium container flying 10km above ground. Sitting in a self-driving car is psychologically quite different.
True, but that's those who are simply afraid of flying. It still doesn't necessarily explain those people who would be uncomfortable flying without a pilot, but perfectly fine with one.
I also hear people telling me that they would not be comfortable with a passenger on a self-driving bus. Anecdotal evidence, I know. And since they haven't actually experienced one, I still think the jury is out on this one; but it still strike me as curious that rational people are not always swayed by statistics.
It depends on people's preconceptions. I live in a country where taxi drivers race like in that Taxi movie, and often remove seat belts from passenger seats for macho reasons, so self-driving taxis would be something that would allow me to ride in a taxi without fear.
Why do you think that this is a "freak accident"? The situation doesn't seem to be very unusual or unlikely. The sensors should definitely detect a truck that is crossing your lane.
Thank you for illustrating my point so clearly. There is no evidence that this is freak accident. There may be a systematic flaw in the firmware, or not using sensors, or any number of things that only come to light in situations like this.
I don't think you can statistically make that claim on this N=1 data point given that "This is the first known fatality in just over 130 million miles where Autopilot was activated. Among all vehicles in the US, there is a fatality every 94 million miles. Worldwide, there is a fatality approximately every 60 million miles."
They should not have mentioned that global number. I doubt Teslas drive much on the really dangerous roads in third-world countries that make that number so high, certainly not on autopilot (Aside, those numbers make it look like US roads are safe. They are relative to third-world countries, but not relative to Western Europe)
More importantly, autopilot bails out when it thinks it cannot do a reliable job. I think those times correlate strongly with times when it gets dangerous.
There also is the effect of car quality. It is harder to get killed in a Tesla or Volvo than in a 1970's Volkswagen Beetle.
It probably is impossible to make a fair comparison, but I think the better comparison would be with modern and fairly expensive cars driving on US highways.
I'm not even sure it counts as level 3, it's more like a high-end Level 2 system.
Level 3 requires the car to detect "all" conditions when something is going wrong, and provide the driver "sufficiently comfortable transition time".
If there is any class of event (such as a 1m overhang) which the vehicle can't safely cope with (perhaps coming to a complete stop before requiring the driver to take over), then the driver really has to be paying attention all the time, making it a Level 2 system.
And there is basically no way to deliver on that promise on today's highways without a suite of sensors and computers that are capable enough to do full Level 4.
I think Tesla have a lot to answer for here. They are pushing the limits of a technology which Google has said, "needs to be 100% autonomous". No steering wheel. Tesla knows this, but branded their car as 'auto-pilot' anyway. When it became clear from early youtube videos that people were not using the technology 'correctly' Tesla did nothing more than issuing a statement.
They should be considering a recall. It is not good enough that Tesla PR is 'sad' that someone died. What are they doing about it?
They don't even have to do a recall; All Teslas can be updated (or in this case, downgraded) over the air. All they need to do is either learn from this and fix the 'white vehicle against a white sky' problem, or remove the 'auto-pilot' feature completely, until they actually develop proper auto-pilot.
Paraphrasing from JFK: We choose to make auto-driving cars, not because it is easy, but because it is hard.
There will be casulties along the way. Deaths are not a reason to stifle development. If we'd done that, we'd never have spaceflight or Vaccines (or Nuclear Weapons for that matter...) .
Tesla actually use the correct term for it, the fundamental problem is a mass misunderstanding of what aircraft autopilot does.
Autopilot in an airliner is capable of flying the plane on a set course in ideal conditions. This means good weather, no other air traffic in the vicinity and full feedback from its various sensors. As soon as the plane enters a scenario outside of which it's designed to operate, control is reverted to the human pilots.
That's exactly how Tesla's system is designed. So long as the computers can account for the current scenario, the car drives itself, but it explicitly falls back on the human (who, like a pilot, must be ready to take command at any time).
A trucker pulling out on a driver and blind-siding them falls outside the definition of 'ideal condition' in my book. I live in Europe where underrun guards are mandatory on all trailers. Sounds like one wasn't fitted here. That's likely more to blame for the fatality than Tesla's software or a distracted driver - let's face it, anyone could suffer a lapse in concentration at a critical moment on the road.
On the surface, yes, but what actually differentiates one large angular piece of metal from another to radar? Not much. The best option, as others have suggested, may well be for Tesla to incorporate LIDAR, but that means hardware changes to all its cars.
I expect those would have helped in this situation, as the car would have impacted the guards (and thus it would be a normal collision), rather than having its top sliced off.
The description of the accident surprised me as well, but your comment makes sense. Vehicles getting under trailers is something I've only seen in (US) movies.
There is also an exemption for historic trucks as well as for trucks, which are mainly used in terrain. That applies to all the Expedition trucks, which are used by world travelers.
What's interesting, I didn't know what exactly a tipper truck is, and I did a Google image search for that query. All of the trucks shown to me had highly visible underrun guards on their sides.
Underrun guards are a good idea but not yet practical for trailers in the US. The reason is that loading dock approaches have far too much variance in ground height, and the empty space under the trailer is essential to have. If guards were in place, they would come into regular contact with the ground and either damage the trailer or the truck would be 'high centered' and stuck, resting on the guards. I suppose the guards could be made to fold up out of the way but this would add weight and complexity, and probably affect strength. Trucking companies in the US do things like use 'super single' tires that are extra wide to span the width of dual tires. This saves the weight of two unneeded sidewalls and save a few pounds weight. Because truck design is so influenced by large fleet purchases, even very tiny changes to weight are avoided. Lighter > safer.
I don't think it's a question of willpower. The contiguous US is 48 states, each with the autonomy to calculate and set max axle weight / distance rules and many do. Sliding tandem axles under a 53-foot trailer are not something you just re-architect overnight.
The span under the trailer behind the rear drive axle(s) and the forward trailer axle is much longer than you find on European trucks. That means you need a much longer underride guard. You also have to convince trucking company execs that all that extra weight is a good thing even though they know it's cheaper to pay the occasional settlement after a gruesome accident.
Also, putting the trailer axles further forward to shorten the gap is not possible until Indiana changes their minimum axle length regulation (40 feet trailer kingpin to rear axle centerline) because I believe that while they are the only state to have such a rule, you can't do much interstate commerce with a fleet of trucks and avoid travelling through that state with its central location.
Trucking in the US is massively regulated and bureaucratic and you are going to need a lot more than will to get it to change. I'd love to see safer trucks too but I am not optimistic about seeing underride guards in the US anytime soon.
Given all the EU bashing recently: That's something I really like. We share the driving licenses (classes, as in 'B' for cars) even.
You're only touching axle weight and distance, but I wonder if the max length (truck + trailer) is clearly defined? Are you allowed to have more than one trailer (here: Nope unless you're driving a tractor most likely)?
I have a BCE ('40 tonnes', biggest you can get here) license, although it is currently inactive¹. Trucks continue to amaze me and I'd like to understand how different the world is over there.
①: Truck licenses have a renewal date here in DE. For people far above my age (my dad) it's something like "Every five years after you're 60" or something like that. For me it was "Every 5 years" from the get go and I haven't renewed it for quite some time. I still could get it back, but would need to visit a doctor (general health), optometrist (field of view test) and maybe redo a first aid course.
"Will" in this case, though, is basically a euphemism for "cost of (retro)fitting trucks and loading docks with guardrails" vs "cost of lives lost by failing to do so." I'm not sure how those two numbers would compare.
The dock approach wouldn't matter with side guards and those could've prevented the death in this accident, if I'm reading Tesla's description of events correctly.
The guards are mostly just thin aluminium panels on steel supports, so if the car impacted them front first, going at highway speed(lets assume 50-60mph?) they would have made no difference whatsoever.
I am getting so disappointed by how widespread this wasteful use of the English language has become. The phrase is just filler and means nothing, yet it is used everywhere, nearly every HN thread has someone who uses it. If you need to note something, then note it. By putting it into words, you have said "this is important enough to actually say it." Further saying that it is important is a crutch.
This is probably why copy editors get so frustrated when technical people write things, because the idiom is particularly pervasive in the tech blogosphere.
We all need to do a service to the value of our words and see [0 ... 3]
I don't think this has anything to due with legality. A legal contract, for example, doesn't need this phrase anywhere for the words to be legally binding: after all, they are in the contract itself, therefore don't need any special modifier.
Some clauses and phrases are required to have extra attention called to them to be binding, which is why you see some paragraphs bolded, or all capitals.
I kind of agree but I think there's a few subtle changes as to how you phrase it.
I sometimes type "Please note <something important>" to make sure the recipient pays extra attention to this bit and actually notices it, because people have a tendency to read through email very quickly and skip vital parts. I know this for a fact based on stupid responses I've gotten.
"Please note" never implies it should actually be noted on a piece of paper, just notice it and ready it thoroughly.
"It should be noted" - however, could be interpreted as you want the recipient to actually write it on a note.
Edit: And yes, Tesla's use of "It is important to note" kind of falls in the gray zone for me... but I've never even considered this an issue
Not everything that is important enough to write down has the same importance. While the phrase may be over-used, it is merely one in a long range of methods we use to draw more attention to statements we believe have particular importance at a specific point.
If the statement is not genuinely something that needs particular attention, then I agree that such phrases should be avoided.
However Wikipedia's policies in are not particularly relevant here, as in the case of an encyclopaedia which pieces of information are important depends entirely on what the reader is looking for at any given moment - the purpose is to make the information available, not to direct the reader to pay special attention to particular bits of information.
In many other contexts, the purpose of the text is not to merely make information available, but to influence the reader, and in many cases to specifically draw their attention to specific pieces of information. E.g. consider safety instructions or documentation of something where there are serious caveats.
In those situations various forms of emphasis are important, and using specific phrases to convey that a given statement is more important can be one valid way of creating emphasis.
On the other hand, dashing off a quick comment isn't exactly the pinnacle of writing. "It is important to note" is lazy writing. So many of my comments here have the terrible sins of passive voice. I think; in my opinion; just; very; i throw in useless phrases all the time.
Forgive us for writing a long letter, we didn't have the time to write a short one.
> By putting it into words, you have said "this is important enough to actually say it."
But you have not said this is important enough for you to read, right now, nor especially important enough to remember every detail. If everything written were important to read, we'd spend our lifetimes failing to work through the backlog - of every word ever written since the dawn of writing. There are key points to remember, and then supporting facts to be forgotten after they've done their job of arguing for those key points, and this language emphasizes their key points (in defense of Tesla, naturally, rather than the mundane statistics.)
> Further saying that it is important is a crutch.
Perhaps. But remember - crutches are legitimate medical devices built to serve a need, they have purpose! I tend to use the terser "IMPORTANT:", "NOTE WELL:", "N.B.:"[1], "NOTE:", etc. to call attention to high priority stuff. I'll especially use this style of emphasis in code, where I can't make my text bold, underlined, large-fonted, red, blinking for emphasis, embedded into a modal dialog you can't dismiss for the first 10 seconds, just to make sure you read the damn thing. These are the notes I deem to have e.g. a 100:1 time saved/spent ratio, where I need my fellow programmers to stop skimming as a thousand inane comments before have perhaps trained them to do, and to instead save us all a lot of time and energy by reading the important thing.
But being that terse in a press release like this would come across as incredibly tone deaf and insensitive. Try replacing "It is important to emphasize that" with "NOTE:" and "It is important to note that" with "IMPORTANT:". The paragraphs themselves might read fine, but in the broader context of the press release it'd be cringe worthy.
> Try replacing "It is important to emphasize that" with "NOTE:" and "It is important to note that" with "IMPORTANT:".
That would be an equal problem as the original and not how a good writer should solve the wording.
Compare this to, for example, the press releases by a major news organization when it admits it got something important quite wrong. There are many ways to use our language as it was designed without getting meta.
Compare these two versions of the first 2 paragraphs. The second version actually emphasizes the point better, without saying "it is important to emphasize."
Original:
We learned yesterday evening that NHTSA is opening a preliminary evaluation into the performance of Autopilot during a recent fatal crash that occurred in a Model S. This is the first known fatality in just over 130 million miles where Autopilot was activated. Among all vehicles in the US, there is a fatality every 94 million miles. Worldwide, there is a fatality approximately every 60 million miles. It is important to emphasize that the NHTSA action is simply a preliminary evaluation to determine whether the system worked according to expectations.
Following our standard practice, Tesla informed NHTSA about the incident immediately after it occurred. What we know is that the vehicle was on a divided highway with Autopilot engaged when a tractor trailer drove across the highway perpendicular to the Model S. Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied. The high ride height of the trailer combined with its positioning across the road and the extremely rare circumstances of the impact caused the Model S to pass under the trailer, with the bottom of the trailer impacting the windshield of the Model S. Had the Model S impacted the front or rear of the trailer, even at high speed, its advanced crash safety system would likely have prevented serious injury as it has in numerous other similar incidents.
Simple edit to one sentence:
We learned yesterday evening that NHTSA is opening a preliminary evaluation into the performance of Autopilot during a recent fatal crash that occurred in a Model S. This is the first known fatality in just over 130 million miles where Autopilot was activated. Among all vehicles in the US, there is a fatality every 94 million miles. Worldwide, there is a fatality approximately every 60 million miles.
The NHTSA action is simply a preliminary evaluation to determine whether the system worked according to expectations.
Following our standard practice, Tesla informed NHTSA about the incident immediately after it occurred. What we know is that the vehicle was on a divided highway with Autopilot engaged when a tractor trailer drove across the highway perpendicular to the Model S. Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied. The high ride height of the trailer combined with its positioning across the road and the extremely rare circumstances of the impact caused the Model S to pass under the trailer, with the bottom of the trailer impacting the windshield of the Model S. Had the Model S impacted the front or rear of the trailer, even at high speed, its advanced crash safety system would likely have prevented serious injury as it has in numerous other similar incidents.
Well this is 9 days old so I doubt you or anyone will see it. But I really want to say that this is just how language evolves. Phrases appear, become common, and even change meaning, subtly or entirely, from the original.
I don't think it's a waste of space. I think it conveys some information. If you remove it from a paragraph, the paragraph feels different. The meaning slightly changes. In a literal sense it may not mean anything, but I feel it does convey some intention from the speaker.
According to the numbers there's a 25% probability it would have taken more than 130 million miles before the first fatality, assuming an average of 1 fatality every 94 million miles.
That's ignoring the fact that Tesla's autopilot might only be used in situations with a relatively low average number of fatalities per million miles.
This is just beyond me. How did the radars not see the front portion of the truck, the front wheels, the underbody of the container isn't that high to let the entire car go through?
This is what I was wondering too. Which is also why I predict (although basically speculating!) that an alert driver would have noticed the actual lorry itself and made the assumption that it carried a trailer, and at the very least slowed down.
It's a narrow beam radar (about 200m range) which checks the current lane is clear, it probably assumed the other parts of the truck were some obstacle on the side of the road (or in another lane)
I'm not even sure the radar has a height dimension, I suspect they just focused it at bumper height to catch most cars.
The car also has a single camera in the front to see the lane markers, and a short range ultrasonic system to give it a 360 degree view, out to a few meters.
For the amount of confidence drivers are putting into the system, Tesla don't have anywhere enough sensor data.
It was explained elsewhere that the radar does not scan "up" to avoid collision warning for overpasses. Actually I remember reading a reddit thread months ago saying that a Tesla in Autopilot could, theoretically, hit a steel beam that was sticking out onto the road at around the roof height, because the radar wouldn't see it.
And I absolutely agree with your last sentence - these cars need more sensors.
I understand they're trying to protect their reputation, but the tone of the post is really defensive and hand-wavy which really sucks the compassion out of it.
Tesla Autopilot: 1 fatality per 130 million miles
All US vehicles: 1 fatality per 94 million miles
Worldwide vehicles: 1 fatality per 60 million miles
If the statistics above are accurate, logically there should be no reason for distress, worry, accusations, or mistrust. Or at least much less so than for ordinary driving. It's like shark attacks: No matter how remote it is, there's going to be big debates about it, investigations, TV specials, lawsuits, hand wringing, and new laws.
Well, if you read those vehicle statistics as "the richer the people, the less fatalities", then those Tesla Autopilot scores may be worse than you might expect.
If all the world drove cars whose passive safety was as good as Tesla Model S, number of miles driven per fatality worldwide would be more than 130 million miles, I think. An average car around the globe is quite a junk. I think we should compare the Tesla Autopilot numbers only with non-Autopilot Tesla numbers.
The dupe detector needs to be improved so that submissions cannot be posted twice. It works when I try to submit something which just has been posted the same day, but there must be some logic that allows it to be reposted and also the URL check isn't sufficient. Alternatively a merge of posts by moderators would be okay, too.
What strikes me most is that the first paragraph is only dedicated to the benefit of Tesla, trying to downplay the event by comparing it to human drivers.
After three more paragraphs to deflect damage from Tesla the last paragraph - like an afterthought - is about the tragedy of the death of a person.
Perhaps this is just me.
I at least would have wished the first paragraph would have followed the headline of "A tragic loss".
There's a 10% of abandonement for each additional paragraph in the post so you need to put the most important facts first to ensure your message is conveyed. Their priority is to restablish confidence and there is nothing that can be done to bring that person back to life. They are trying to save other people's lives by pushing their idea for a safer driving.
For me they have tarnished their brand. They are just like all the others. Puff, magic gone with one press release. All of Musks talk, in the end he is not different than profit-above-all GM executives - a little more brilliant perhaps.
93 comments
[ 4.7 ms ] story [ 167 ms ] threadtelsa's autopilot is level 3, which means freeway driving only: https://en.wikipedia.org/wiki/Autonomous_car#Classification
The thing I fear now is the toxic mix of fanboyism and excessive PR management reminiscent of the worst excesses of Detroit. The only way safety is improved is when hubris and bluster is put to one side.
This is the future of roadsafety.
Even if it's statistically safer than the alternative.
There's no reason why an entire sedan should fit under the trailer of a truck.
This is what a truck designed for safety looks like: https://upload.wikimedia.org/wikipedia/commons/3/3f/Artic.lo...
Notice how no part of it is higher off the ground than a foot or so. Way too low for a car to fit under.
This is what truck in the US looks like: http://www.center-linecurtains.com/wp-content/uploads/2015/0...
Notice how you could hide a sports car underneath.
There is, it's called capitalism, and a safety skirt costs money.
The fix is, as in many cases, regulations. However, I believe that even the shown safety skirt leaves too much space below, and especially exposed tires, which is still dangerous for pedestrians and bikers.
As for pedestrians and bikers, I really really like Europe's solution to that problem: Pedestrian and Bikers are illegal on the freeway. Big trucks are made strongly and forcefully impractical in cities, towns, and villages.
It's crazy that I live in downtown SF and there are huge freight trucks driving past my bedroom window every day.
It's dangerous as hell, but impractical to do any other way...
Deliveries inside the city are done with smaller lorries which are a bit safer maybe. If not safer, definitely more practical on tight inner-city streets.
I don't think it's practical for trucks to be designed as pedestrian safe as modern city buses are, so it's better to keep them away from pedestrians as much as possible.
I don't think capitalism is the issue.
Saving resources by skipping on safety features is not exclusive to capitalism. If anything, the correlation seems to run backwards - capitalist countries are generally safer (mostly, I'd say, because they're rich enough to spend extra resources on safety).
In case of non-profits and the government people might actually sue for wasting of money because money is spent on stuff that is not required by law.
Can you give an example of such a lawsuit ever having succeeded?
Tesla have statistics on their side so far, but they still have not done a fraction of kilometres travelled by conventional means.
I hope the technology moves forward, but the technology is not there yet and won't be appearing on a majority of new cars for decades.
I predict - moreso than fear - that as more of these incidents occur (and I am confident there will be more), more people will lose their trust in self-driving cars.
Consider the following: The safest form of transportation is air travel. Yet despite the overwhelming evidence in air transport's favour, a significant portion of people are afraid to get on an aeroplane. Indeed, an aeroplane can practical fly itself (and have been able to for decades), yet few of us would feel safe if there wasn't a pilot on board.
Surely the same fears will soon apply to self-driving cars?
I also hear people telling me that they would not be comfortable with a passenger on a self-driving bus. Anecdotal evidence, I know. And since they haven't actually experienced one, I still think the jury is out on this one; but it still strike me as curious that rational people are not always swayed by statistics.
Really speaks to the promise of autopilot.
They should not have mentioned that global number. I doubt Teslas drive much on the really dangerous roads in third-world countries that make that number so high, certainly not on autopilot (Aside, those numbers make it look like US roads are safe. They are relative to third-world countries, but not relative to Western Europe)
More importantly, autopilot bails out when it thinks it cannot do a reliable job. I think those times correlate strongly with times when it gets dangerous.
There also is the effect of car quality. It is harder to get killed in a Tesla or Volvo than in a 1970's Volkswagen Beetle.
It probably is impossible to make a fair comparison, but I think the better comparison would be with modern and fairly expensive cars driving on US highways.
Level 3 requires the car to detect "all" conditions when something is going wrong, and provide the driver "sufficiently comfortable transition time". If there is any class of event (such as a 1m overhang) which the vehicle can't safely cope with (perhaps coming to a complete stop before requiring the driver to take over), then the driver really has to be paying attention all the time, making it a Level 2 system.
And there is basically no way to deliver on that promise on today's highways without a suite of sensors and computers that are capable enough to do full Level 4.
I think Tesla have a lot to answer for here. They are pushing the limits of a technology which Google has said, "needs to be 100% autonomous". No steering wheel. Tesla knows this, but branded their car as 'auto-pilot' anyway. When it became clear from early youtube videos that people were not using the technology 'correctly' Tesla did nothing more than issuing a statement.
They should be considering a recall. It is not good enough that Tesla PR is 'sad' that someone died. What are they doing about it?
Paraphrasing from JFK: We choose to make auto-driving cars, not because it is easy, but because it is hard.
There will be casulties along the way. Deaths are not a reason to stifle development. If we'd done that, we'd never have spaceflight or Vaccines (or Nuclear Weapons for that matter...) .
Is personal responsibility something that's completely taboo nowadays?
Autopilot in an airliner is capable of flying the plane on a set course in ideal conditions. This means good weather, no other air traffic in the vicinity and full feedback from its various sensors. As soon as the plane enters a scenario outside of which it's designed to operate, control is reverted to the human pilots.
That's exactly how Tesla's system is designed. So long as the computers can account for the current scenario, the car drives itself, but it explicitly falls back on the human (who, like a pilot, must be ready to take command at any time).
A trucker pulling out on a driver and blind-siding them falls outside the definition of 'ideal condition' in my book. I live in Europe where underrun guards are mandatory on all trailers. Sounds like one wasn't fitted here. That's likely more to blame for the fatality than Tesla's software or a distracted driver - let's face it, anyone could suffer a lapse in concentration at a critical moment on the road.
http://www.transportsfriend.org/road/guards.html
I expect those would have helped in this situation, as the car would have impacted the guards (and thus it would be a normal collision), rather than having its top sliced off.
There was an attempt to introduce legislation in the US in the 90's, I believe, but it was killed.
It's all about will.
The span under the trailer behind the rear drive axle(s) and the forward trailer axle is much longer than you find on European trucks. That means you need a much longer underride guard. You also have to convince trucking company execs that all that extra weight is a good thing even though they know it's cheaper to pay the occasional settlement after a gruesome accident.
Also, putting the trailer axles further forward to shorten the gap is not possible until Indiana changes their minimum axle length regulation (40 feet trailer kingpin to rear axle centerline) because I believe that while they are the only state to have such a rule, you can't do much interstate commerce with a fleet of trucks and avoid travelling through that state with its central location.
Trucking in the US is massively regulated and bureaucratic and you are going to need a lot more than will to get it to change. I'd love to see safer trucks too but I am not optimistic about seeing underride guards in the US anytime soon.
You're only touching axle weight and distance, but I wonder if the max length (truck + trailer) is clearly defined? Are you allowed to have more than one trailer (here: Nope unless you're driving a tractor most likely)?
I have a BCE ('40 tonnes', biggest you can get here) license, although it is currently inactive¹. Trucks continue to amaze me and I'd like to understand how different the world is over there.
①: Truck licenses have a renewal date here in DE. For people far above my age (my dad) it's something like "Every five years after you're 60" or something like that. For me it was "Every 5 years" from the get go and I haven't renewed it for quite some time. I still could get it back, but would need to visit a doctor (general health), optometrist (field of view test) and maybe redo a first aid course.
> It is important to emphasize
I am getting so disappointed by how widespread this wasteful use of the English language has become. The phrase is just filler and means nothing, yet it is used everywhere, nearly every HN thread has someone who uses it. If you need to note something, then note it. By putting it into words, you have said "this is important enough to actually say it." Further saying that it is important is a crutch.
This is probably why copy editors get so frustrated when technical people write things, because the idiom is particularly pervasive in the tech blogosphere.
We all need to do a service to the value of our words and see [0 ... 3]
[0] https://en.wikipedia.org/wiki/Wikipedia:It_should_be_noted
[1] http://womeninwetlands.blogspot.nl/2013/12/it-is-important-t...
[2] https://en.wikipedia.org/wiki/Wikipedia:%22Note_that%22_is_u...
[3] http://english.stackexchange.com/questions/266529/similar-ph...
There is no overlap.
I guess this is due to legal concerns (which also provided us with pages-long email footer legalese bullcrap).
I sometimes type "Please note <something important>" to make sure the recipient pays extra attention to this bit and actually notices it, because people have a tendency to read through email very quickly and skip vital parts. I know this for a fact based on stupid responses I've gotten.
"Please note" never implies it should actually be noted on a piece of paper, just notice it and ready it thoroughly.
"It should be noted" - however, could be interpreted as you want the recipient to actually write it on a note.
Edit: And yes, Tesla's use of "It is important to note" kind of falls in the gray zone for me... but I've never even considered this an issue
If the statement is not genuinely something that needs particular attention, then I agree that such phrases should be avoided.
However Wikipedia's policies in are not particularly relevant here, as in the case of an encyclopaedia which pieces of information are important depends entirely on what the reader is looking for at any given moment - the purpose is to make the information available, not to direct the reader to pay special attention to particular bits of information.
In many other contexts, the purpose of the text is not to merely make information available, but to influence the reader, and in many cases to specifically draw their attention to specific pieces of information. E.g. consider safety instructions or documentation of something where there are serious caveats.
In those situations various forms of emphasis are important, and using specific phrases to convey that a given statement is more important can be one valid way of creating emphasis.
Because, Orwell was awesome.
On the other hand, dashing off a quick comment isn't exactly the pinnacle of writing. "It is important to note" is lazy writing. So many of my comments here have the terrible sins of passive voice. I think; in my opinion; just; very; i throw in useless phrases all the time.
Forgive us for writing a long letter, we didn't have the time to write a short one.
But you have not said this is important enough for you to read, right now, nor especially important enough to remember every detail. If everything written were important to read, we'd spend our lifetimes failing to work through the backlog - of every word ever written since the dawn of writing. There are key points to remember, and then supporting facts to be forgotten after they've done their job of arguing for those key points, and this language emphasizes their key points (in defense of Tesla, naturally, rather than the mundane statistics.)
> Further saying that it is important is a crutch.
Perhaps. But remember - crutches are legitimate medical devices built to serve a need, they have purpose! I tend to use the terser "IMPORTANT:", "NOTE WELL:", "N.B.:"[1], "NOTE:", etc. to call attention to high priority stuff. I'll especially use this style of emphasis in code, where I can't make my text bold, underlined, large-fonted, red, blinking for emphasis, embedded into a modal dialog you can't dismiss for the first 10 seconds, just to make sure you read the damn thing. These are the notes I deem to have e.g. a 100:1 time saved/spent ratio, where I need my fellow programmers to stop skimming as a thousand inane comments before have perhaps trained them to do, and to instead save us all a lot of time and energy by reading the important thing.
But being that terse in a press release like this would come across as incredibly tone deaf and insensitive. Try replacing "It is important to emphasize that" with "NOTE:" and "It is important to note that" with "IMPORTANT:". The paragraphs themselves might read fine, but in the broader context of the press release it'd be cringe worthy.
[1] https://en.wikipedia.org/wiki/Nota_bene
That would be an equal problem as the original and not how a good writer should solve the wording.
Compare this to, for example, the press releases by a major news organization when it admits it got something important quite wrong. There are many ways to use our language as it was designed without getting meta.
Original:
We learned yesterday evening that NHTSA is opening a preliminary evaluation into the performance of Autopilot during a recent fatal crash that occurred in a Model S. This is the first known fatality in just over 130 million miles where Autopilot was activated. Among all vehicles in the US, there is a fatality every 94 million miles. Worldwide, there is a fatality approximately every 60 million miles. It is important to emphasize that the NHTSA action is simply a preliminary evaluation to determine whether the system worked according to expectations.
Following our standard practice, Tesla informed NHTSA about the incident immediately after it occurred. What we know is that the vehicle was on a divided highway with Autopilot engaged when a tractor trailer drove across the highway perpendicular to the Model S. Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied. The high ride height of the trailer combined with its positioning across the road and the extremely rare circumstances of the impact caused the Model S to pass under the trailer, with the bottom of the trailer impacting the windshield of the Model S. Had the Model S impacted the front or rear of the trailer, even at high speed, its advanced crash safety system would likely have prevented serious injury as it has in numerous other similar incidents.
Simple edit to one sentence:
We learned yesterday evening that NHTSA is opening a preliminary evaluation into the performance of Autopilot during a recent fatal crash that occurred in a Model S. This is the first known fatality in just over 130 million miles where Autopilot was activated. Among all vehicles in the US, there is a fatality every 94 million miles. Worldwide, there is a fatality approximately every 60 million miles.
The NHTSA action is simply a preliminary evaluation to determine whether the system worked according to expectations.
Following our standard practice, Tesla informed NHTSA about the incident immediately after it occurred. What we know is that the vehicle was on a divided highway with Autopilot engaged when a tractor trailer drove across the highway perpendicular to the Model S. Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied. The high ride height of the trailer combined with its positioning across the road and the extremely rare circumstances of the impact caused the Model S to pass under the trailer, with the bottom of the trailer impacting the windshield of the Model S. Had the Model S impacted the front or rear of the trailer, even at high speed, its advanced crash safety system would likely have prevented serious injury as it has in numerous other similar incidents.
I don't think it's a waste of space. I think it conveys some information. If you remove it from a paragraph, the paragraph feels different. The meaning slightly changes. In a literal sense it may not mean anything, but I feel it does convey some intention from the speaker.
I don't expect autonomous driving to be perfect until all variables can be accounted for and the largest being humans behind the wheels.
That's ignoring the fact that Tesla's autopilot might only be used in situations with a relatively low average number of fatalities per million miles.
https://twitter.com/elonmusk/status/748625979271045121
I'm not even sure the radar has a height dimension, I suspect they just focused it at bumper height to catch most cars.
The car also has a single camera in the front to see the lane markers, and a short range ultrasonic system to give it a 360 degree view, out to a few meters.
For the amount of confidence drivers are putting into the system, Tesla don't have anywhere enough sensor data.
And I absolutely agree with your last sentence - these cars need more sensors.
After three more paragraphs to deflect damage from Tesla the last paragraph - like an afterthought - is about the tragedy of the death of a person.
Perhaps this is just me.
I at least would have wished the first paragraph would have followed the headline of "A tragic loss".