Elon Musk Says Tesla New Autopilot Features Would Have Prevented Recent Death (fortune.com)

An anonymous reader writes:Tesla Motors Chief Executive Elon Musk said on Sunday the automaker was updating its semi-autonomous driving system Autopilot with new limits on hands-off driving and other improvements that likely would have prevented a fatality in May. Musk said the update, which will be available within a week or two through an "over-the-air" software update, would rely foremost on radar to give Tesla's electric luxury cars a better sense of what is around them and when to brake. New restrictions of Autopilot 8.0 are a nod to widespread concerns that the system lulled users into a false sense of security through its "hands-off" driving capability. The updated system now will temporarily prevent drivers from using the system if they do not respond to audible warnings to take back control of the car. Musk said it was "very likely" the improved Autopilot would have prevented the death of Brown, whose car sped into the trailer of a truck crossing a highway, but he cautioned that the update "doesn't mean perfect safety."

Comments

So what was the prior feature?

[Anonymous Coward]

So he admits from his own mouth that the previous technology is a killer?

Re:So what was the prior feature?

[110010001000]

Exactly. Lawyers are going to love that comment.

Re:So what was the prior feature?

[Anonymous Coward]

No, all he said is that he expects that the new software will keep a few more Darwin award candidates in the gene pool. There still is such a thing as personal responsibility. Even in the USA.

Re:So what was the prior feature?

[K. S. Kyosuke]

The notion that newer car technology is safer than older car technology will completely blow their minds!

Re:So what was the prior feature?

[Anonymous Coward]

They most definitely do not. Just ask any vegetable or the cattle!

Re:So what was the prior feature?

[Junta]

This is less like something like adding airbags and more like when airbags explode with shrapnel.

The former is an improvement that's seen as natural evolution of things. The second is a safety problem where a functionality is likely to do more harm than good and comes with liability issues.

Here, Tesla may be found to have been irresponsible by calling the feature 'autopilot' with a bunch of connotations in the minds of the users causing them to watch DVDs instead of driving. Additionally, 'beta testing' highly dangerous functionality is not something car companies generally get to do. When it comes to anything resembling autonomous vehicle operation, you can point at *any* other company and how extremely careful and conservative they are being. Even with similar 'lane assist' technologies that are in production vehicles, they *already* were being far more strict about monitoring driver attention than Tesla was.

So here we have Tesla being more aggressive about how 'automatic' things are, taking less measures for safety than the rest of the market offering equivalent feature today, and not being as conservative as the efforts that are what Tesla purports this technology to be.

I know there is a desire to bow down and really kiss up to Tesla, but they need to be held to the same standards as their competitors. They are not holy saviors of our society. They aren't even the only electric car company. They are certainly not the most accessible/affordable things. They have done nothing to earn having a blind eye turned to their mistakes.

Re:So what was the prior feature?

[110010001000]

Because airplanes have actual highly trained "drivers" in them, and know what an autopilot of capable of, and cars are driven by stupid fuckwits like you. The average driver doesn't know that autonomous cars are a pipe dream at this point, especially with supposedly knowledgeable tech people claiming it is right around the corner.

Re:So what was the prior feature?

[K. S. Kyosuke]

The airbag comparison sucks because an explosion of the airbag wouldn't be preventable with your behavior in the car, whereas the recent death was Darwinian in nature. The explosion would make the manufacturer liable, but the late driver was most certainly breaking the ways in which the feature was supposed to be used, such as paying attention to the road. At best, one could argue that it shouldn't be offered but then you have a chicken and an egg problem of sorts. As to the "autopilot" label, see below.

Re:So what was the prior feature?

[NatasRevol]

You fail at history. Civilization has always had a few gladiator sports.

Re:So what was the prior feature?

[K. S. Kyosuke]

Yes, and cars don't have actual highly trained drivers, obviously. ;) They just let anyone drive these days, even people calling others "stupid fuckwits".

Re:So what was the prior feature?

[K. S. Kyosuke]

That even you have a driver's license in the US?

Re:So what was the prior feature?

[110010001000]

Exactly. If they allow morons like me to drive, then they allow anyone. I don't know why you are attacking me, anyway. Space and AI nutters are the worst.

Re:So what was the prior feature?

[K. S. Kyosuke]

By the way, there's one more thing... Apparently, the current average vehicular death rate in the US hovers around one death per one hundred million miles driven. All Teslas together have recently crossed the one billion miles mark. That would imply ten deaths if Teslas were average. Yet we have only one death linked with autopilot usage (no matter how flawed said usage was). That doesn't sound very bad for Tesla unless there were nine more deaths associated with it.

Re:So what was the prior feature?

[K. S. Kyosuke]

Well, you need your daily dose of righteous anger! :D

Re:So what was the prior feature?

[Anubis IV]

So he admits from his own mouth that the previous technology is a killer?

That's faulty logic. If a hospital upgrades to some just-released equipment and is able to save more people as a result, that doesn't mean that they killed the people who could have been saved by that equipment, had it been available earlier. The state of the art is constantly getting better. Admitting that the newer stuff is more safe than the older stuff doesn't mean that the older stuff was killing people. Quite the contrary, since in many cases that older stuff saved a number of lives that would have been lost had people relied on the alternatives that were otherwise available at the time. Saying that the newer stuff is even more safe just means that we have something even better now.

Of course, I say all of this to point out the fault in your logic, not to suggest that I think Autopilot is ready for primetime already. Because it's not.

Re:So what was the prior feature?

[110010001000]

I just am always shocked at the stupidity and naiveté displayed by Space and AI nutters like you. You have no grasp on reality. I should let you guys wallow in your ignorance, but thats no fun.

Re:So what was the prior feature?

[fluffernutter]

Many people here think the lives that autonomy end today don't matter based on the dream that it will ever be in the hands of enough people to make a statistical difference in the future.

Re:So what was the prior feature?

[fluffernutter]

If you think flying in the open sky is anywhere as complicated as driving on a public highway then you're an idiot.

Re:So what was the prior feature?

[DrXym]

If they're allowing the user to take their hands off the wheel for up to a minute then they're still not taking safety seriously. Maybe the car is better than it was before, but it'd be better again with an attentive driver instead of one playing with their web browser because the car lets them.

Re:So what was the prior feature?

[110010001000]

I can confirm: he is an idiot. Just kidding. He is just ignorant.

Re:So what was the prior feature?

[imgod2u]

Current cars allow you to take your hands off the wheels for way longer than a minute. You may not live that long depending on the road you're driving but it's an option.

I fail to see how it's somehow required that an assist feature also make you do something a car without the assist feature doesn't make you do.

Re:So what was the prior feature?

[imgod2u]

Autonomy has not ended any lives. The auto-feature didn't make someone take their hands and eyes off the wheel and road. I can take my hands and eyes off the road right now in my car (without any lane assist) and it'd be nobody's fault but mine.

Re:So what was the prior feature?

[fluffernutter]

You're right.. Human nature did all that. The fault of autonomy is that it is not designed with human nature in mind.

Re:So what was the prior feature?

[Kjella]

It's easy enough if is an improvement across the board like our new vaccine is more effective with no new side effects. Statistically though, often your new way will end up killing different people which makes it hard to swallow at an individual level. For example, say you have an injury with 5% chance of dying without surgery but you're in a dirty field hospital with a 2% chance that infections and there's very little correlation between one and the other. Are your relatives going to happy that you got a three percent better chance or are they going to blame the hospital for killing you? I think the latter. I'm not sure if Tesla does better or worse than humans overall, but I'm pretty sure it will kill different people and by that I include idiots who over-rely on the Autopilot. The real fireworks won't begin until an innocent bystander is killed though.

Re:So what was the prior feature?

[ravenshrike]

How long till someone comes up with a fake 'hand' that tricks the system into believing you maintain control of the car?

Re:So what was the prior feature?

[ThatsNotPudding]

I know there is a desire to bow down and really kiss up to Tesla, but they need to be held to the same standards as their competitors.

Such as the sky-high moral standards of GM https://en.wikipedia.org/wiki/General_Motors_ignition_switch_recalls/?

Re:So what was the prior feature?

[ColdWetDog]

OK, let's use boats as an example. For non commercial use, any old idiot can drive a boat. And many do. Boats have autopilots. They're quite common these days and actually have more routing options than a Tesla AP. I can program my boat to autonomously steer its way until the tanks run dry. Of course, that is a remarkably stupid thing to do but it happens.

And AP equipped boats crash all of the time. So far there hasn't been a giant wave of insurance companies insisting that we ditch the things (nice puns, eh?)

Progress, and human stupidity, trundle on.

Re:So what was the prior feature?

[MattskEE]

Your statement is not backed up by data. Fatal Tesla crashes make the news far more readily than any other vehicle which leads you to falsely conclude that they are less safe than other vehicles, a fact not born out by data. Nearly 100 people die every day driving on US roads.

Further, multiple drivers have falsely claimed that they had Autopilot engaged during a crash. But due to our rapid news cycle people are far more likely to read the first article that says Autopilot caused a crash and skip over the followup article where it explains that the driver was driving unsafely and lying about Autpilot being engaged.

You cannot rely on your impressions to make an informed judgement about this topic, only data matters. The NHTSA is looking into the matter (as they should) and will let you know if there really is a safety problem.

Re: So what was the prior feature?

[Junta]

It is the average user understanding that must be kept in mind when doing branding/naming exercises like calling something 'autopilot', unless you have a very thorough licensing process that explicitly covers what autopilot is or is not.

Here it's a marketing gimmick to sound cool. There are repurcussions.

Also, Musk was still saying 'hey it's still safer than driving yourself', which again reinforces the common imagined image of autopilot.

Re:So what was the prior feature?

[Junta]

Simple. In the unaided case, it's unbelievably obvious what will happen immediately. There's no perception that the car will save you. Driver's ed required pretty well covers what will happen if you aren't paying attention.

The problem is the assist technologies can let you get away with it for extended periods of time, the majority of the time. So car has to nag the user to remind them that it is *not* safe to remove human attention as well. It's mitigating risk and doing a deceptively good job at it when it doesn't nag the user. Also not helping when Musk talks about how safe it is, abusing statistics in an unfair way that further encourages risky behavior for the sake of making Tesla sound like magic.

Also, here, the hands can be off the steering wheel for a minute in the old scheme. This tells us that they *explicitly* put in a mechanism to measure driver attention and explicitly chose to let them neglect the vehicle for a full minute.

Re:So what was the prior feature?

[Junta]

A soda can worked for some mercedes S class. However that is a lot more blatantly obvious that a user is doing unreasonable stuff to bypass safety mechanisms, so vendor can reasonably be considered less responsible if the user is having to get so 'imaginitive' to be jackasses.

Re:So what was the prior feature?

[Junta]

GM was crucified in the media for that. No one took GM's side on that. That is how reaction *should* be.

With Tesla, tons of folks are white knighting for a luxury car brand. It's insane.

Re:So what was the prior feature?

[Junta]

Also, the 'all inclusive' number includes all ages and variety of vehicles. Tall SUVs, 15 year old cars with fewer safety features. Cars with much older pieces and different levels of maintenance.

It would be interesting to compare Tesla autopilot specifically to good-condition highway miles driven of 2014+ model year vehicles with forward collision alert type systems. That would be a more fair comparison.

Re:So what was the prior feature?

[Junta]

This may be true, but competitors all were more strict about lane assist and more reluctant to use the term 'auto' in any way associated with the technologies. Google has avoided anything other than trained professional testers to use their system, and has publicly stated their opinion that a system that's 90% there is more dangerous than one not there at all, because user expectations are problematic.

Again, people always say how people who know anything about piloting aircraft know that autopilot is far more limited than the public imagination thinks. The problem being the market is not people who know anything about aircraft, but about common drivers. In their world, they have the 'auto' transmission. In that case, they say 'go forward' and don't have to think about it as it completely takes care of itself. That is the level of expectation set for automobile operators.

If you get your pilot license, you know full well the limitations of the technologies as it's part of the required training. Before you can make a misconception and fail with it, you have gone through personal training, hands on with an experienced pilot. In the Tesla case, there's a click through EULA that is probably not read and not given much weight, because we are just inundated with EULAs. This competes with Tesla execs going around praising how much safer the system is than the human operators, and tossing out apples to orange comparisons to make the case that the user is better of surrendering their attention to the system (the 'deaths per driven miles' statistic).

Re:So what was the prior feature?

[CWCheese]

Fatal driver-piloted automobile crashes make the news every single day of the year, usually several times each day. No matter what software update is disseminated by Tesla Motors, there will always be a bug or fault that can and will lead to another fatal crash. Just a matter of time.

Re:So what was the prior feature?

[Anubis IV]

You don't need Tesla Autopilot to live (drive the car) compared to an dialysis machine.

If you are going to die unless technology intervenes, then, yes, the technology is needed for your survival. Period. It makes no difference when or even if you recognize your need. The need exists, regardless of your perception. A Tesla driver suffering from Imminent Fatal Crash Syndrome who could have been saved by Autopilot needs Autopilot in exactly the same way that a patient suffering from kidney failure who could have been saved by dialysis treatment needs dialysis treatment.

The reason your statement appears to hold water is because you're comparing the general population on one side against the small group of people whose need we recognize on the other side, rather than comparing like with like. What you should have been comparing was the people on each side whose lives would actually be saved by the relevant technologies.

More or less, you're getting hung up on your perception of the situation. You knew about the dialysis patient's need in advance, so you think that they have a need where Tesla driver's don't, but the fact is, most people will never have to face either situation, so your logic that says Tesla drivers "don't need" Autopilot would also tell us that most people "don't need" dialysis. In both cases, however, there are people who will die unless the technology intervenes. They both need the technology to survive.

What you're defending [...]

You do realize that people can scroll up just a hair to read the last line of my previous post and see that I said nothing of the sort, right? I actually said the opposite of what you claim I did. The only thing I was defending was the application of logic and critical thinking in our discourse, regardless of which side you take.

[...] half-baked technology that can kill multiple people being marketed as fully baked

Pretend for a moment that Autopilot was replacing a previous automated system that was exactly as safe as human drivers are today (i.e. 30,000 fatalities per year in the US). If we could say that Autopilot was 5% safer than the earlier system, we wouldn't hesitate to replace the earlier system, would we? It'd mean saving 1500 lives each year in the US alone, while still allowing 28,500 deaths. It wouldn't be perfect, but it'd still be a large improvement over the status quo, so we'd consider it well worth it, right?

But because Autopilot is taking control from humans, rather than from an automated system, we (myself included, I'll admit) won't consider it fully-baked until it is SIGNIFICANTLY beyond a 5% improvement. Even though Tesla is claiming that it's already nearly twice as safe as the average human driver, we still won't trust it. We each like to think that we're above-average drivers (even though half of us statistically aren't) and we fear ceding our lives to systems outside of our control, even if they're demonstrably safer than keeping control to ourselves. It's a matter of perception for us, rather than fact. Our perception tells us that the handful of deaths attributable to Autopilot are a huge cause for concern, regardless of the fact that they may have come at the benefit of many more lives saved.

Again, as I already said, I don't think that Autopilot is ready for primetime just yet, and I'm not saying that I buy into Tesla's line that Autopilot is already safer than humans (to be clear, I absolutely do not buy into that line, based on firsthand accounts I've heard of how it performs), but I also want us to do a better job of separating perception from reality. We got security theater at airports the last time we let perception shape our reality when it comes to the topic of our safety. I don't want to see a repeat.

Re:So what was the prior feature?

[AmiMoJo]

The most irresponsible thing was ignoring the vast amount of research into human attention spans. People simply can't concentrate on doing nothing for hours on end, ready to take over control of the vehicle at a moment's notice.

People will get complacent, take a nap, watch a DVD or just zone out. It's human nature. NASA even warned them about it.

Re: So what was the prior feature?

[dave420]

The users are required to confirm that they read about it before they can use it, apparently. It explains precisely what it is and isn't, in no uncertain terms. It doesn't stop people misusing it, but blaming the tool seems a bit bizarre.

Re:So what was the prior feature?

[DrXym]

I should think the problem is extremely obvious. A normal car will crash itself if you take your hands off the wheel. A car with autopilot will continue to steer itself and will appear to do so quite well for the most part. A consequence of that is drivers WILL become inattentive and WILL do things other than paying attention to the road, hazards, other vehicles etc. This is entirely forseeable and obvious.

And for the most part (that part where the car seems to be doing okay) maybe it doesn't matter if a driver is inattentive. But it sure as fuck matters when some kid runs out into the road, or a truck in front jack knifes, or the car has a "moment" and fails to see some obvious hazard. At that moment, the driver's attention might mean the difference between a serious accident and avoidance. Perhaps the car is quite good in normal situations but it's still not as good as a car AND a driver.

Allowing a driver to spend up to a minute messing around with his / her phone or whatever else is simply inexcusable. The car should force the driver to hold the wheel and only allow them to remove their hands for a moment. And if they do for longer, ignoring warnings then the car should pull over, stop, and disable autopilot for some period of time commensurate to making them use the mode properly.

The problem here is Tesla wants to have their cake and eat it all. They say their car is safe because of this-or-that feature and then they shove in things which are inherently unsafe. I wonder what Volvo does with their self drive functionality because I bet they take safety a hell of a lot more serious than Tesla does.

Re:So what was the prior feature?

[DrXym]

I'm sure there are people who spoof their car's safety belt sensors too. You can only do so much to prevent determined idiocy. It doesn't mean you shouldn't enforce safety at all for the majority of people who aren't idiots.

Re:So what was the prior feature?

[elgaard]

You are comparing apple and oranges here.

There have been other fatal Tesla accidents that was not caused ty autopilot. And plenty of the fatal accidents in average car were not caused by bad driving (i.e., you cannot assume that autopilot would have prevented them).

Re: So what was the prior feature?

[Junta]

A click through EULA. I'm sure that was thoroughly reviewed and taken seriously.

The wider industry has several exmaples of being more careful. Tesla is *now* being more careful, like their competition, which is a good thing. I don't know what it buys to continue to defend their previous behavior, which they themselves have realized was wrong.

Re:So what was the prior feature?

[david_thornley]

For non commercial use, any old idiot can drive a boat.

I used to sail on rented sloops about 36' long fairly frequently, around Wisconsin's Apostle Islands. On one trip, we had a blind guy take the helm for a while, and also a seriously developmentally disabled woman with sensory problems. Neither of them were ever going to be able to drive a car.

Just emphasizing your "any old idiot".

Re:So what was the prior feature?

[david_thornley]

Sure. That doesn't mean we should do everything possible to extend every single life. People who do dumb things in potentially dangerous situations are going to die at a higher rate than people who don't, and there's really nothing we can do about that.

Re: So what was the prior feature?

[Coren22]

but it won't stop you from plowing into another plane or a mountain.

Actually, yes it will. automatic collision avoidance systems (both ground and airborne) lare quite common in aircraft these days.

No, it is quite rare, as it is on one model of airplane. Also, it is a backup, the pilot is supposed to do the navigation/altitude changes, but it will do the changes when the pilot doesn't respond.

so with an auto drive car what happens when the

[Joe_Dragon]

so with an auto drive car what happens when the software is at fault with the payouts?

Upgrades!

[Anonymous Coward]

...better than educating generation after generation of human driver and relying on them to have their faculties intact every time they're behind the wheel.

"the system will temporarily shut off"

[fustakrakich]

What does that mean? Will the vehicle quickly slow to a stop? Will it veer off a cliff or into a building or "let go of the wheel" and start swerving to scare the driver into grabbing hold? The statement doesn't make much sense.

Re:"the system will temporarily shut off"

[Luthair]

Does it matter? The crash won't be autopilots fault since it wasn't in control anymore!

Re:"the system will temporarily shut off"

[fustakrakich]

Well, the thing should at least scream over the speakers, *Hey dumbass! You're all gonna die if you don't grab the wheel and hit the brakes!* to get the guy's attention. A little flashing light and chime won't cut it. More than one airliner has crashed because the pilot didn't know the autopilot disengaged.

Re:"the system will temporarily shut off"

[michelcolman]

And not just in English, but in Mandarin too!

Re:"the system will temporarily shut off"

[CaptnCrud]

Then you get people sueing because the alarm was to disruptive and scared them into a wreck....

The only way to win this battle is to literally sign a waiver for some new class of driver's license. Basically stating you understand and are aware of all the risks and best handling practices when owning/operating a car with an autopilot system.

Re:"the system will temporarily shut off"

[wonkey_monkey]

I'm guessing it gives the warning, and if you don't take heed, it won't work the next time you try to engage it (but will continue to work until you do retake control this time).

So you'll just need to find a convenient roundabout so you can dive out and leave the car running until you're ready to hop back in again the next day.

Re:"the system will temporarily shut off"

[Anonymous Coward]

I prefer the computer should say "Take the wheel or I swear I'll drive this car into a tree!" And or "Don't make me pull this car over!" and "Don't make me come back there!"

Add in an occasional dad joke (like any time some one says "I'm Hungry" replying with "Hi Hun Gry, can I just call you Hun?" and provide answers to the question of "are we there yet?" such as "we'll get there when we get there and not a minute sooner!" and you've got the perfect auto pilot we all recall from childhood family road trips.

Re:"the system will temporarily shut off"

[YrWrstNtmr]

Then you get people sueing because the alarm was to disruptive and scared them into a wreck....

To prevent that, you first broadcast a 'trigger word' warning.
(over the car speakers, in a very calm voice) "Hey, bro...I'm about to yell at you quite loudly, because there is a very scary moment coming up, and I don't know how to handle it. We'll need you to go ahead and take the wh..." (crunch)

Re:"the system will temporarily shut off"

[Anonymous Coward]

Well! Happy Monday to you too! Why are you being such a dick? Haven't had your coffee or beat the wife and kid yet? Where's your priorities? First things first, man!

Re:"the system will temporarily shut off"

[PopeRatzo]

and the safety briefing drivers are required to receive before Tesla will turn it on.

Is it a safety briefing or a EULA?

Re:"the system will temporarily shut off"

[fustakrakich]

Not too comforting to the station wagon full of nuns in the next lane over. The problem with things like this, it has to be all or nothing. Either it works or it doesn't. On the other hand auto parking isn't such a bad idea. At least that could get you within walking distance to the curb without playing bumper cars.

Re:"the system will temporarily shut off"

[CaptnCrud]

Not any better than a drivers license either though, anyone can steal a pair of keys and just drive a car today without any knowledge of how to drive. The point was the driver would be liable and not the auto-pilot, or at least the norm would be assume driver error should they actually be using the feature during a wreck.

I guess to me this isn't any different from someone trying to use cruise control, you should still be ready to use the pedal at all times no matter what.

Re:"the system will temporarily shut off"

[ColdWetDog]

"Terrain, Terrain, Terrain. Pull Up. Terrain. Terrain. Terrain. Pull Up."

That's what planes do.

No sense getting personal.

Re:"the system will temporarily shut off"

[ColdWetDog]

"I'm sorry Dave, I can't do that."

That would be so cool.

Re:"the system will temporarily shut off"

[CWCheese]

The analogous message would then be:
Semi, Semi, Semi. Engage brakes. Semi, Semi, Semi. Engage brakes.

Would that be effective enough?

Re:"the system will temporarily shut off"

[thegarbz]

What does that mean?

Same thing it means now. Slowly bring the car to a stop. This facility already is in place.

Re:Take back control of the car...

[Luthair]

I recall some automotive journalist saying in the earlier days it would just bail, even if you were say in the middle of a corner...

Elon Musk

[Anonymous Coward]

I was getting worried there for awhile when I hadn't seen an Elon Musk story for like 9 hours. I was beginning to fear the worst, but I think I'm OK now.

Phew!

Re:Let's talk about the name!

[phantomfive]

Let's be honest, if the driver had thought of it as 'assist' rather than 'auto', and had been paying attention, he wouldn't have died.

Re:Let's talk about the name!

[clickety6]

Introducing Tesla's New and Improved Hands-Free Driving System*



*NOTE: Hands-Free Driving System requires hands to be holding the steering wheel at all times.

Re:Let's talk about the name!

[taiwanjohn]

Obviously just a typo... they clearly meant "Otto Pilot".

Re:Let's talk about the name!

[MrLogic17]

That's a mighty big assumption.
That and you're not factoring in that Tesla's system in an opt-in beta feature with a big disclaimer you must agree to. Changing one word in the description of the feature isn't going to negate all those hurdles to just get it turned on...

SubjectsInCommentsAreStupidCauseTheSubjectIsTFA

[lesincompetent]

Elon Musk implies next autopilot version would have prevented death(s) caused by current autopilot version.

Re:Let's talk about the name!

[phantomfive]

with a big disclaimer you must agree to.

You're a moron if you think people read those.

Elon Shrugged

[PopeRatzo]

What would the world do without him? Now he's claiming his rocket blew the fuck up because it was hit by a UFO.

I assume the release notes

[mrprogrammerman]

-Fixed bug where the car would crash into things if the sun was too bright It looks like that Bill Gates car industry joke is coming true.

Re:Sigh.

[Anonymous Coward]

They're not whitelisting any/all objects in a particular location.
They are whitelisting a specific radar signature in a specific location.

Re:Let's talk about the name!

[Junta]

Note that people treat 'assist' as 'auto' as well. However no doubt Tesla exacerbates the issue by calling it that.

A bigger thing is that in addition to calling it 'lane assist', competitors *also* more aggressively monitored user attention.

For example, Mercedes drivers taped a can to trick the sensor:
http://www.roadandtrack.com/ca...

Though in that case, I think anyone can argue that a person getting killed doing that was *really* going out of their way to act against the designer's intent for the vehicle. Contrasted with Tesla which had messaging *and* implementation that encouraged people to be reckless.

Re:Let's talk about the name!

[dcollins]

"Cue the endless discussion on the "Autopilot" name, rather than any discussions of the technical merits of the system or its implementation."

The user interface, the presentation to the user, including how it is described by its name, is a big part of the implementation.

Frank, but ambulance chasers will rejoice!

[140Mandak262Jamuna]

Elon spoke frankly and with candor about the technology. Most other manufacturers don't do that. They hide behind bland statements and corporate spokesman. They have learnt it the hard way, being frank and open leads to law suits. So they hide behind these bland useless press releases. That makes them look cagey and shifting, and when something actually comes out, it is seen a lot more harshly it deserves. Let us see what happens in this crash.

Doesn't sound very safe to me

[DrXym]

The driver can still take their hands off the wheel for up to a minute. During which time they could be turned around fetching something out of the back, eating a sandwich, playing with Twitter. Anything except actually paying attention to the vehicle they are inside - a 2 ton vehicle hurtling down the road at 70mph.

So yeah, maybe the new software makes the car better at not crashing into trucks. It sure as hell isn't better than if the car AND the driver were both attentive to the road. Humans are excellent at negating safety features and any system which doesn't take this into account is dooming people (and not necessarily the driver) to an early grave.

Re:Doesn't sound very safe to me

[DrXym]

Correction - 1 ton

Re:Doesn't sound very safe to me

[fnj]

Where do you find a car weighing 1000 kg nowadays? A Mini was 600-700 kg in 1960. A VW bug was under 1000 kg. Today a Mini is 1200-1400 kg, a VW Golf is 1300-1400 kg, a Prius is 1400 kg, a Tesla S is 2000-2200 kg, a Ford Explorer is 2000-2200 kg, a Fiat 500 is 1100 kg, and a Yaris is 1100 kg

Cars nowadays, even tiny ones, are bricks.

Re:Doesn't sound very safe to me

[DrXym]

I was right the first time. Somehow I thought I'd read 2000 pounds.

Goofy

[fluffernutter]

When I first read about this comment, all that ran through my mind was the voice of Goofy saying, "Heeeyuk well I guess THAT didn't work, lets try something different!"

Re:Goofy

[110010001000]

That is essentially how people in the IT/software industry think. It is easy to change strategy and approach with software. You just change the code. However that doesn't fly in other fields. You can't just say "shit well they died so lets try again with version 2.1 beta and see if it kills fewer people." Well actually I guess you CAN say that.

Re:Goofy

[fluffernutter]

I would hope people who write software in life or death situations would approach the problem with greater foresight.

Re:Goofy

[cellocgw]

However that doesn't fly in other fields. You can't just say "shit well they died so lets try again with version 2.1 beta and see if it kills fewer people." Well actually I guess you CAN say that.

That's exactly how it goes, and there's no other option. You could do nothing, and guess what the death rate wouldn't change. Or you could do something which directly addresses the root cause of the problem, and see if the death rate decreases. In all the cases I know of - checklists for pilots, drug dosage level changes, etc. -- by the time the new protocol is released for use on humans it's pretty much guaranteed not to make things worse.

Re:Goofy

[fluffernutter]

Usually in cases of life or death, there is standardized testing done before products are released to the market. Reference FDA testing standards for drugs. Reference flight tests by professionals before a prototype airplane is mass produced. Tesla has designed a new airplane here and basically brought it from blueprint to production in a single step.

Re:Goofy

[fluffernutter]

I can't imagine that they would have released software that could drive into a trailer if they had tested thoroughly. ESPECIALLY after the case a month before where a person pocket dialed their car and it drove into the back of a trailer on its own. No excuse for them to only be closing that gap now.

Re:Goofy

[fluffernutter]

Sure.. you have one case with a trailer, think of all other situations that can happen with a trailer... This stuff isn't hard. You just have to have a mindset that you need to prepare for anything. If you don't feel you are prepared for everything that can possibly happen in the physical world then you haven't tested enough.

Re:Goofy

[david_thornley]

In that case, you have never tested enough. Nobody will ever think of all the possible things to test in any real-world scenario. If you listed all the possible situations with a trailer, there's a good chance I could look through it and find something you'd missed.

Re:Sigh.

[Joe_Dragon]

So now you can have gps accuracy issues and DB update issues that can lead to a big mess.

Re:This seems like such a trivial problem

[fluffernutter]

Still won't work. You would have to reproduce any stupid event that can happen in real life:
- Kid running out into road 20' in front of car
- Kid running out into road 10' in front of car
- Kid running out into road 5' in front of car
- Person dropping rock off bridge at same increments
- Sandbag in road on bend
- Sandbag in road on straightaway
- Large obstacle overhanging left six feet of road (on straightaway / on bend)
- Large obstacle overhanging left ten feet of road (on straightaway / on bend)
- Shopping bag blowing across road
- Heavy object the size of a shopping bag swinging from bridge


If engineers were able to think of all this stuff then they would design the car to handle it in the first place.

Revenge of the Developer

[Etcetera]

Next week: "If only they'd stayed on HEAD, we wouldn't be having these problems. Also, your kernel is old and you're not running the latest version of systemd. Why don't you just hook into our Jenkins server at http://carautopilot.github.io/ so you can get the latest nightly before you head out on the road each morning?"

My Dad's Subaru

[wiredog]

has an almost-self-driving capability when the lane departure assist is activated. But when driving on the freeway in heavy snow last winter, as soon as the optical system couldn't see the road because of the snow building up, all the automation shut down. Wth lots of visual and auditory warnings to let me know it was shutting down.

I imagine this is the same sort of thing. Auditory and visual warnings to let the driver know the system is switching to fully manual operation.

Re:Sigh.

[ColdWetDog]

I'd suggest not flying, taking a modern boat or even a bus if you're worried about stuff like this.

Just walk along side of the freeway - you'll be fine.

Re:This seems like such a trivial problem

[ColdWetDog]

If engineers were able to think of all this stuff then they would design the car to handle it in the first place.

We've got you covered.

I won't trust a Telsa autopilot for a long time.

[superdave80]

When I found out the cause of the fatal wreck (the sky was cloudy white, and so was the trailer!) I couldn't believe it. You would think that the FIRST thing you do with an autopilot program is to make sure it can see properly in front of the vehicle. Or combine it with radar. Or... something. But an autopilot that runs full speed into a giant truck/trailer without even realizing that it is even there is a complete and utter failure. What would happen if there was a blue trailer that was a similar color of the blue sky? Or an empty flatbed trailer. Would it run into those as well?

Re:This seems like such a trivial problem

[fluffernutter]

Perhaps armor is the most reasonable solution. I keep hearing about all the power of the electric motor, it would have enough to carry it.

Re:Let's talk about the name!

[Junta]

I'm saying that, in the Mercedes article, people are still going to be jackasses and tape soda cans to the steering wheel. It's still going to happen.

However, in the Mercedes case they call it assisst and require folks to fool the sensors to be unsafe, so that's a bit excusable.

Mercedes even had to pull an ad that indirectly implied that maybe, possibly the lane assist feature was in the same ballpark of an autonomous car.

So Tesla bad: using autopilot and not being strict about user attentiveness, Mercedes better by using 'assist', being very careful about messaging, and being strict about user attention.

Re:I won't trust a Telsa autopilot for a long time

[thegarbz]

You would think that the FIRST thing you do with an autopilot program is to make sure it can see properly in front of the vehicle.

Define see. You're assuming that the human would have made a better decision. Given how many accidents happen on a daily basis with humans in control I don't think this is a conclusion you can make. But there's one amazing thing here: Iteration.

A human has an accident you can't prevent it from happening again. If you run a red light and t-bone someone else the entire world can't learn from it. Yet here we have a case where in the future this accident won't likely happen. I will happily tolerate many deaths providing we maintain an excellent record of staying ahead of crappy human drivers and continuous improvements that make driving ever safer. Assuming any system is perfect on day one is asinine.

Re:I won't trust a Telsa autopilot for a long time

[superdave80]

You're assuming that the human would have made a better decision.

I think most humans would have NOT kept driving at full speed into a giant tractor trailer. But that's just a guess.

Now, don't think that I am against self driving cars. My post specifically mentioned Tesla's autopilot. In the long term, I think self driving cars will prevent many, many more deaths than they might cause. My issue is that the very first thing any self driving car should be able to do is know if something is blocking it's path. Literally the first thing it should 'learn' to do. Given that it failed this test (spectacularly, I might add), I wouldn't touch a Tesla autopilot system for a long time.

Yet here we have a case where in the future this accident won't likely happen.

Really? Why would you think that? Tesla didn't properly program/test this scenario the first time. Why do you think they would get it right the next time? See my examples of other, similar modes of failure that could cause this.

Assuming any system is perfect on day one is asinine.

I'm not asking for perfection. I'm asking that an autopilot system deployed to consumers can do the first, basic task of any autopilot system: Know if something is in your way, and stop if there is. If it can't do that bare minimum task, I won't use it.

I will happily tolerate many deaths...

Thanks for offering to Beta Test the Tesla autopilot system :-)

Re:I won't trust a Telsa autopilot for a long time

[thegarbz]

I think most humans would have NOT kept driving at full speed into a giant tractor trailer. But that's just a guess.

And yet it happens on a daily basis without something as stupid as the trailer being the same colour as the sky.

Really? Why would you think that?

Re-read my post.

If my computer calculates 1+1=5 and I figure out why and correct it to 1+1=2 then the problem is resolved. That doesn't mean 2+2 won't =5 in the future, but the 1+1 situation has been corrected.

Now compare that to a human. It's hard enough teaching one person something, it's not possible to correct it for everyone.

Thanks for offering to Beta Test the Tesla autopilot system

If I had a Tesla you bet your arse I would be using the autopilot system.