It took seven miles to pull over a Tesla with a seemingly asleep driver
arstechnica.com
arstechnica.com
Fortunately he was headed north on US 101, not south. Seven miles ending in Palo Alto means he went through the 101-85 interchange where a Tesla hit the divider barrier and killed the driver. But that shallow-angle split is southbound only. Northbound, you can go seven miles to Palo Alto without any unusual offramps that would confuse Tesla's lane-follower.
Tesla's lane follower is pretty good. That's what creates the illusion that it can self-drive. Car-ahead detection, reasonably good. Obstacle detection, not so good. So the CHP did the right thing, boxing it in with their patrol cars and slowing down. It would have been much worse if they'd set up a roadblock.
I’m pretty sure it would be some emergency stop button in the cabbin at first but then someone somewhere would want to pass a law that would enable the police to remotely initiate it.
Audi already has a backseat traffic driver mode where you can get a foot rub while the car drives in traffic.
This will eventually require a solution.
Regulatory requirement is per jurisdiction, isn’t it.
Also I don’t think manufacturers would want to develop 500 system so when the first demand comes they’ll likely implement it across the board.
I would also not be surprised if they find a way (or at least try too) to charge LEO’s for this “services” similarly to how cellphone companies charge for wiretaps, call logs and telemetry data.
No more Botts dots.
The article states: "The vehicle was traveling southbound on Highway 101 in Palo Alto."
The interchanges for seven miles north of Palo Alto are all rather vanilla, too. 85 is the only big, complicated interchange near there.
and i don't like the overly positive implication that auto-pilot saved this person's life. no, in fact, its existence, as described in a separate incident in the exact same article, likely encouraged the drunk driving in the first place. a car just free running on the road with no one behind the wheel is not a good thing. seven miles is a long time to stop the car and endangered many other people's lives just as an active drunk driver would. auto-pilot did not improve the situation as the article implies.
autopilot is the lane guidance of the skies that maintains basic stability not a level 5 system. people hear autopilot and think it means it does "everything."
further muddying the waters, tesla maintains their car has all the hardware, sensors and computation to support fully autonomous driving. it's just that pesky software missing.
FWIW that's not necessarily true either, autopilots run the entire gamut from "maintains trajectory and altitude" to "automated flight from takeoff to landing" (only the taxiing is left to the pilot).
Military aircraft go a step further and avoid hills. https://en.m.wikipedia.org/wiki/Terrain-following_radar
Remember, the aircraft needs to avoid everything in a line in front of it. But, they also want to stay as low as possible to avoid enemy radar, so it also wants steer steer around hills and even trees vs simply going up.
i dont think most airplanes even have active radar, let alone cameras/lidar and randomly moving obstacle avoidance at milisecond latency, road sign recognition, etc.
comparatively boring at 35k ft. many orders of magnitude complexity difference.
An aircraft autopilot can be quite dumb and remain safe.
2. it's not communicationally correct, most drivers are not pilot, and when you talk about "autopilot" to a layperson they expect an automated driving system (SAE3 or more), and this false impression is exactly why Tesla's marketing named their driving assistance the way they did
3. and it's not really technically correct either, there are autopilots which can do everything between takeoff and taxiing (including landing), tesla's driving assistance can do no such thing
The only thing an autopilot, both the tesla's and the airplane's, is good for is reducing the cognitive load of the operator.
I disagree, based on my Model S orientation from my delivery specialist, as well as the very clear warnings on the MCU display regarding the driver needing to maintain control at all times (which is also presented on the dashboard display above the steering wheel each and every time autosteer is enabled). It’s made abundantly clear what the limitations are.
Stupid people will do stupid things. Safeguards have diminishing returns, and does not absolve an operator of personal responsibility.
To even bring up aircrafts (or boats or whatever) is incredible disingenuous and not comparable in any practical sense what so ever.
The issue isn't autopilot at all—it's all still the driver. In this case the autopilot prevented his death, even if you may root for it.
Blame, whether or legal or moral, isn't zero (or fixed) sum; greater blameworthiness for one party need not be offset by lesser for another.
The liability of an accessory to a crime, for instance, doesn't reduce the liability of principals in that crime.
When people hear autopilot, they think of planes where autopilot means you can literally not be in the cockpit for 10 minutes and still be fine. That leads to Tesla drivers making unwise decisions - like drunk driving - with the assumption that their car can handle itself on its own.
People will always fall for false claims which is why misleading advertising is illegal. How Tesla still manages to do it without being able to provide evidence to support the claims is beyond me. Probably some legal loophole.
P.S. Although the wording was slightly changed, Tesla still insists their cars have fully self driving capabilities on www.tesla.com/autopilot. A claim that’s obviously not supported as of yet.
No they don't. They say they have the hardware for it, but the functionality will only become available once it passes software validation and regulatory approval.
>Full Self-Driving Hardware on All Cars
As a massive headline and
> 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.
As the rest of what I see on one screen- means that it's self driving and safer than a human.
More from the page
> Full Self-Driving Capability
And only towards the end does it say
> Please note that Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction. It is not possible to know exactly when each element of the functionality described above will be available
Which even then isn't clear if it's self driving or not yet.
Frankly, if you sell me something you say is capable of full self driving, I expect it to be able to do that. I'm sure they're technically correct but I think they're wildly misleading.
> if you sell me something you say is capable of full self driving
As I pointed out, they haven't said that. They've said it's got the hardware needed for it.
Compare:
This iPhone is capable of 16-way FaceTime calls.
…to:
This iPhone has the hardware necessary for 16-way FaceTime calls.
When iOS 12 was released, one was true but the other wasn't. It wasn't until Apple released iOS 12.1 that iPhones were capable of making 16-way FaceTime calls.
Hardware support is necessary but not sufficient. What Tesla are saying on that page is that when you buy the car, you've got all the hardware needed, but the software that is needed is yet to come, at an uncertain future date.
Right now, Tesla are saying that you've got all the hardware, you just need the software, and the software isn't ready yet. This isn't at all the same thing as saying that the car has this capability right now. Those are two quite distinct claims and I think that page is clear on saying the former not the latter.
If you're saying that Tesla claiming hardware support is tantamount to claiming full functionality, then how do you suggest they describe the situation? Do you think they should be banned from mentioning the hardware support? Does this apply to other manufacturers like Apple as well?
Less dishonestly.
Your entire argument boils down to "people should be expected to parse Tesla's (intentionally, I might add) obfuscated language". There's a very good reason why the language used in marketing is regulated in just about every country on the planet.
No, my argument is as follows:
1. It is reasonable for Tesla to distinguish between having hardware capability and actual functionality.
2. It's necessary for them to make this distinction in order to offer the hardware upgrades for pre-sale.
3. Other vendors do this without people complaining.
Do you disagree with any of those?
It's completely irrelevant whether or not it's reasonable for Tesla to distinguish between having hardware capability and actual functionality, the way they are describing it is demonstrably and indisputably leading consumers to believe their cars are actively capable of full self-driving, which is absolutely not the case.
Tesla are misleading consumers, their intention is thoroughly irrelevant.
I understand that, but I get the impression that no way of saying it will satisfy you. This is why I asked how you think they should describe it. This is why I asked if you actually disagree with any of those points. Do you?
> the way they are describing it is demonstrably and indisputably leading consumers to believe their cars are actively capable of full self-driving
Is it? I mean, there's obviously going to be idiot outliers for any product. A drunk person using awful judgement does not qualify as "demonstrably and indisputably" in my view. Has anybody actually demonstrated this is actually misleading to the average person?
Well, the people at Tesla who assumed a Vulcan like understanding of their marketing language from buyers are either malicious marketeers or idiots themselves.
No, it's not. Not in any way that can be misconstrued, and not in marketing copy for existing products, for which the "hardware capability" alone means shit.
In fact, the very distinction is BS. Either the car can autopilot or not. Whether the hardware can but needs the software which is not available is of interest to noone.
It might even be totally wrong as a claim too (e.g. if the software is eventually finalized they might very well discover it actually needs some more sensors to be safe, which means the hardware wasn't ready actually).
>It's necessary for them to make this distinction in order to offer the hardware upgrades for pre-sale.
They shouldn't be preseling updates of hardware that supports something for which they don't have the software yet.
>3. Other vendors do this without people complaining.
No excuse, if they do those other vendors are bad actors too.
Tesla says “Full Self-Driving Hardware on All Cars”. The implication is clearly that the cars are perfectly capable of fully self driving. Hence the misleading advertisement.
The claim is not substantiated in any way. Simply the assumption that technically a few cameras, sensors, and a not very powerful processing unit are enough to enable FSD. To this point there is no actual evidence that the statement is true.
My phone has all the hardware needed to crack any non quantum encryption. Also to solve the teleportation problem, world hunger, and is able to cut any cake in perfectly equal slices. It just needs the right software and perhaps infinite retries.
All Tesla has to do is provide in eqully large typeface and placement evidence that the HW indeed can do it but not at this time. And in case you’re wondering yes, all manufacturers should advertise clearly what the product can and cannot do, and when. Whataboutism is not an argument that the problem isn’t there, just that we should ignore it because that other guy does it.
[1] Proof only available under certain conditions and subject to local regulatory approval.
What they have said explicitly is that they cars are not capable of fully self driving until the software and regulatory approvals are ready. Explicit trumps whatever you think is implicit.
A Tesla car is not just hardware. It's software as well. You are equating claims that are explicitly about the hardware as if they are about the car as a whole, when the page in question clearly makes the distinction between them.
How do you suggest they describe the situation? Do you think they should be banned from mentioning the hardware support?
> Whataboutism is not an argument that the problem isn’t thee, just that we should ignore it because that other guy does it.
So do you think that Apple were wrong to talk about 16-way FaceTime before iOS 12.1 was released? If this is so unreasonable, why didn't anybody complain that Apple were lying about iPhone functionality?
My argument is not that we should ignore wrongdoing because other people do it. My argument is that the reason people do ignore it for other vendors is because it's not wrongdoing.
Are you telling me I’m wrong or what I say is false and misleading, against all you’ve stated in your comments so far (that there’s no wrongdoing in making unsubstantiated claims)?
That's what every scamster who puts 100pt text "This product does X" and them tells in 7pt fine print "fat chance you'll see it doing X" would say.
For one, the "software" might never be ready, it's a hard problem. What timeline they have? What are their guarantees they'll get it? How do they know if the "full self-driving" problem is solved eventually in software, it won't require extra hardware (sensors and such) to operate?
They don't, they just depend on idiots buying their cars thinking a) the autopilot is here, b) the autopilot is near, it's just a matter of shipping the software which they'll release soon.
Should they have been able to make the same claim about the iPhone 4? I mean, it's got the theoretically necessary camera and audio hardware, if you overlook minor details like Apple not actually introducing 16 way calls until that device was basically obsolete, and then not in a software update that would actually run on it.
Even then it's fairly clear doing the 16 way call wasn't a fundamental challenge for Apple and they did it fairly short order. Things are not so cut and dry with self driving technology.
It is quite clear from the marketing material that Tesla knows that average people don't understand plane Autopilot. And the name gives people illusion of more capability than for real and at the same time allows them to point that plane autopilot is fairly limited so they are not misleading people.
Their original language was way worse and it is clear from the changes they are trying to make sure they can point out to the fine print to say we didn't really say it can self drive.
Please note that Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction. It is not possible to know exactly when each element of the functionality described above will be available, as this is highly dependent on local regulatory approval.
This is clearly meant to suggest to the unsuspecting customer that they have the whole thing ready and just waiting for pesky regulatory approvals. This is beyond shameful specially when they are doing this literally for years.
However, I think you're making two mistakes. One, I think to most people they absolutely are making that claim. Two, as I say they may technically be the accurate claims but they're extremely misleading and designed to hide the fact that the cars do not have this ability right now in a usable form.
> They aren't saying the car has that capability.
They have this as a headline: "Full Self-Driving Capability". Under it it says "Build upon Enhanced Autopilot and order Full Self-Driving Capability on your Tesla." I am not cutting sentences in half or anything, they very explicitly say that I can order full self-driving capability on my tesla.
So I can order full self driving capability on a Tesla. They say I can buy a car with that capability. They say parts of it may not be available, depending where I am (though I'm on a country specific page where they know full well that it is not usable).
One of the first things on the page is a video showing self driving!
I feel if I list a series of functionalities on a product I'm selling then say "We can't say exactly when each of these will be available in your area" while I know that definitely none of them are available then I am being either misleading or completely dishonest.
> When iOS 12 was released, one was true but the other wasn't. It wasn't until Apple released iOS 12.1 that iPhones were capable of making 16-way FaceTime calls.
And I'd say they were dishonest if they had headlines of "16 way FaceTime call capability" under which it says "Order the 16 way capable version today!"
> If you're saying that Tesla claiming hardware support is tantamount to claiming full functionality, then how do you suggest they describe the situation?
Quite simple - make it extremely obvious that it's not yet available. Say "Future self driving capabilities". Make it clear that it may never be usable where you live. Maybe have a clear note on the video that the car you buy won't do that when you get it.
As a side, I do not understand their claim either - they say all cars produced have the necessary hardware but then later say it's my choice to order as an extra because the basic version doesn't have it.
Maybe they are to blame. But for practical reason, I don't care much. I interpret Tesla's communication as highly misleading, for the very simple reasons that it could be written in a waaaaay less misleading way, and also because there is absolutely no consensus in the field that their "hardware" will actually be capable one day to fully self drive.
Plus asking the users to distinguish between hardware and (vapor) software is disingenuous and misleading.
Sounds like a problem with people.
autopilot is not a real autopilot, in this case marketing names thing as they want since there are not laws what you can call autopilot
The "autopilot" is designed to stop if something happens with the driver not to continue driving, so it OBVIOUS something went very wrong and we don't need to debate that the autopilot saved his life, it could as well killed many other people.
I'm a bit curious what you think a real autopilot is and how it works. Do you imagine a bunch of planes flying around with pilots asleep or away from the cockpit?
I think autopilot is probably a surprisingly appropriate and accurate name, but most people don't have any real idea what it means so we see a lot of surprising assumptions being brought to bear.
I do not understand the point of trying to change the meaning of autopilot to mean lane keeping, or car chasing but with 100% attention still needed, then how will you name the real autopilot ?
In the article it mentioned that a drunk guy said it is OK to drive drunk because of autopilot , do you think he would also be confused if marketing would have been more honest/modest ?
Again, I ask you what you think an autopilot function in a plane does?
> do you think he would also be confused if marketing would have been more honest/modest ?
I'm merely noting that you said "autopilot is not a real autopilot", and I think that name is actually accurate, but unfortunately the public has a skewed perception of what autopilot means. Specifically, you can read the first paragraph of Wikipedia's entry for it[1] and it should be fairly clear where the public understanding (and possibly your interpretation) of what autopilot is and what it is and is meant for in reality. To cut to the chase, since you seem to have ignored most my hints to look into how autopilot works, it does not automate every portion of flying a place, but just allows pilots to focus on other aspects of operating the craft. A pilots should not ignore the operation of the plane just because of autopilot any more than a driver should ignore operation of a vehicle with some driving assist technology. In both cases, they are assistive technologies, not replacement for drivers.
If you want to make some case about what Tesla specifically said in addition to the naming, feel free to do so, but I'm not interested in arguing that. It wasn't what you initially argued, and it wasn't why I replied (I'm not interested in defending Tesla, just pointing out the naming may not be quite as bad as you think).
#1: Autopilot is mainly two features (Adaptive cruse and auto-steer) that must be enabled in the settings screen, and the driver must click “I agree” on the screen listing the limitations of the feature in reasonably large print. #2: When engaged, the car reminds the driver to pay attention to the road. To engage “autopilot” the driver turns on cruse control, then turns on auto-steer.
Probably when the car can do level 4 driving they will let the driver just engage autopilot, but until then, they should keep having the driver jump through a few hoops to help remind them it’s not complete.
Autopilot as it stands is on the blurry line between the two sides. Just changing the name I don’t think will fix people’s behavior.
It’s hard to get the point across that if you completely rely on autopilot/supercruse you will die and endanger everyone else without sounding like your full of FUD. There is a small subset of the population that will always think ‘I turned it on and it didn’t immediately kill me, so it must work perfectly in all situations’. As an example, just look at the number of videos of people ghost riding, where they assume that if they can take their hands of the wheel for a few seconds, they can also climb onto the roof of thier car and start dancing while its in motion.
You know those extra thick milkshakes you get in burger franchises? Try not to buy them from the cheaper places.
People drink and drive. It’s a huge problem in this country. If Tesla autopilot made him drive drunk then needle exchanges promote drug use. You can’t have it both way folks. It’s either a public safety issue or it’s not.
I mean, when I catch a taxi home after I’ve been drinking, I’m not the driver, so your point stands.
But if the car really could self drive then it could be a selfdriving taxi?
The system was rapidly scraped when people realise someone else could easily blow in.
I guess the down voters are probably wondering what would have happened if nobody tried to stop the vehicle. The obvious thing to do is slow down and stop in front of it, unless the car wound then change lanes out of the way. That’s a worrying thought.
I tend to agree with your sentiment, at least in the future tense: self driving cars could, at least theoretically, drive sleeping / inebriated / injured people.
We’re probably not there yet, and I don’t believe Tesla advertise this as a feature.
Based on the report and the description of how Tesla's auto-pilot functionality is supposed to work, it sounds like a bug or unintended activation of the auto-pilot prevented an initial tragedy, and the Highway Patrol luckily developed a smart way of safely stopping the car. As was commented elsewhere, the story would have been way different had any number of factors, human or AI, been slightly different.
I really hope that a proper review on what engaged the auto-pilot here is performed so there is a clearer understanding.
But as for the GP, I personally find their post to be contentless, even if their reaction is one that many people might have.
a) not have saved his life
b) not encouraged him to rely on them to make up for his current state
Now a basic Corolla will detect that you’re sleepy and encourage you to pull over if it has to intervene often enough. Autopilot is a step backwards from systems like this.
Lack of information in these still-developing stories makes for some entertaining discussion. We can conjure up all sorts of theories to explain the unexpected phenomena.
I think the sleeping driver was slouched against the wheel and engaging the two-hand steering touch sensor.
https://mv-voice.com/news/2018/11/30/los-altos-planning-comm...
However this would have ended more dramatically if the driver had a car from another brand (except a Waymo, in which case he wouldn't be at the driver's seat)
Technology like that will certainly be abused.