You mean like every other car company, and the exact opposite of tesla's entire marketing strategy?
Citation needed. Do you have any examples of Tesla’s marketing strategy indicating that “you can sleep at the wheel”? (Not what they predict of the future of the system, but about it’s current state)
The term autopilot comes from the avionics industry. Correct me if I am wrong, but I am pretty sure a pilot cannot go to sleep while auto-pilot is engaged either.
Elon musk takes hands off the wheel. https://www.businessinsider.co.za/elon-musk-breaks-tesla-aut...
Tesla very much makes it sound like that you just need to sit back and enjoy a coffee while your car drives. Many imagine that the car will give them 5-10 seconds of heads up before takeover is needed.
> Tesla very much makes it sound like that you just need to sit back and enjoy a coffee while your car drives
Again citation needed. Although this might sound like a fact in the HN forums, it is far from the reality.
Coast to coast autonomous drive, cross country summon mode, full self driving on highways from entry to exit, the fsd fake demo video with "driver is there only for legal reasons", million robotaxis that earn you 30k$ a year from 2020, lidar is a fools errand etc etc.
You need more? Here's Elon musk in 2016 proclaiming self driving to be easy and solved. Showing entire lack of understanding what driving is like and why computers fail at it. https://youtu.be/wsixsRI-Sz4?t=1h17m57s
> Coast to coast autonomous drive, cross country summon mode
They have not claimed that they have achieved that yet.
> full self driving on highways from entry to exit
This is from the Tesla website when I search for the above: “The currently enabled features require active driver supervision and do not make the vehicle autonomous” [1]
> the fsd fake demo video with "driver is there only for legal reasons"
Based on what evidence do you claim it is “fake”? That video is to demonstrate FSD and show the future value of the system. It does not say here is how you should operate AP.
> million robotaxis that earn you 30k$ a year from 2020
Oh shit, is it 2020 already?
> lidar is a fools errand
How is that even relevant to this thread?
> You need more?
More than the zero citations you have provided so far? Yes. Yes I do need more.
> Here's Elon musk in 2016 proclaiming self driving to be easy and solved. Showing entire lack of understanding what driving is like and why computers fail at it.
Again not relevant here since we are not talking about future promises here. The discussion is on how Tesla encourages its drivers to use the car today. If they say they are working on a flying car would anyone try to drive their car of a cliff?
Again, if you think any of the above are not accurate please provide citations for your claim. (I’m starting to think folks around here don’t know what citation means)
[1] https://www.tesla.com/support/full-self-driving-capability-t...
Second, regarding your link, it also states you can
>[...] summon your car to come find you.
How can it do this but _also_ "require active driver supervision"? This statement implies that it can, without you in the car, drive to come find you, and it _definitely_ implies the vehicle is autonomous.
Yes, they also say like you said, but these are conflicting statements, and it's VERY easy between this, average Joe's understanding of autonomous cars, and Elon's behavior on twitter to see why so many people think that the Tesla is fully autonomous, and thus get into accidents like they have by simply not paying attention. You can link stuff from Tesla all you want showing how they do a good job covering their ass, but if people are still doing this, then Tesla isn't doing a good job of making this clear. They simply hide it inside the legal documents so that when people _do_ treat the car as autonomous, they can't be sued for it.
Sure it is a fact that Elon Musk predicts they will have robotaxis in a year. It is not however a fact that Elon Musk advises the drivers to drive carelessly.
> How can it do this but __also__ "require active driver supervision"? This statement implies ....
Yes it does. In a parking lot. At parking lot speed. While you are not in the car.
> and it __definitly__ implies the vehicle is autonomous.
If by autonomous you mean it can automatically come to you in a parking lot then yes. If by autonomous you mean you can sleep behind the wheel while it is driving in the streets then it __definitly__ does not imply that.
> But these are conflicting statements...
No they are not. One is about operation in a parking lot. One is about future functionality/value of the product. And one is about the current functionality of the product.
> people think that Tesla is fully autonomous and thus get into accidents like this
Citation needed. I'd say it is more likely that they are not using the car as intended, not that they mistakenly think their car is fully autonomous.
> ... covering their ass
Not sure what your point is here. So they are damned if they do damned if they don't?
> but if people are still doing this then Tesla isn't doing a good job of making this clear
People don't always use things as designed. This is not exclusive to Tesla cars.
> They simply hide it inside the legal documents
Not true. It is on the first page when you search for Tesla FSD. And I am pretty sure the warning is shown to the users at different steps before AP can be activated.
What does the header say right above that video? In a really big headline font, that’s really hard to miss?
“Future of Driving”
Tesla (and more specifically, Elon) seem to continually go out of their way to make that less clear, not more.
A better angle would be that, some people who were gonna fall asleep anyways, had their lives saved.
Warning labels are not an effective way of communicating to people what the limits on a product are. Especially so when the company's CEO heavily implies that those warning labels are exclusively there for regulatory reasons.
On the one hand you have a boiler plate warning, on the other hand you have the mulit-billionare genius CEO of Tesla doing interviews with 60 minutes making a point of taking his hands off the wheel completely and bragging about it.[1]
>A better angle would be that, some people who were gonna fall asleep anyways, had their lives saved.
I don't buy this, people are more likely to not pay attention when they think they've been told they can not pay attention.
[1] - https://www.cbsnews.com/video/tesla-ceo-elon-musk-the-2018-6...
emphasis added
Maybe. Here's the thing. The feature, by its very nature, does not punish inattention like normal driving. If you let go of the wheel and focus on your phone -- even using your knee to hold the wheel, chances are the longer you are distracted the more likely you are to have a minor wakeup that says "don't do that." Maybe your car starts to hit the bumps between lanes. Maybe your peripheral vision picks up something and you get an adrenalin jolt as you refocus attention on the road.
AP is not like that. 99.9% of the time it does a pretty good job on the expressway. (95% if you are in traffic and have defensive driving tendencies where you are projecting ahead 10-15 seconds and moving lanes accordingly.) So you literally can look at your phone for 15 seconds look up and find your car at reasonable distances from other cars and properly in the lane. All of the normal "don't do that" feedback loops for inattention are removed.
But then there's that unusual case, like this one, where AP gets it wrong and the consequences are extreme. Is this different from normal inattention? I think yes because normal inattention often leaves you feeling "I just took a risk" due to the natural, and usually small, consequences.
My point is this: It may not be that the labeling of the feature is dangerous. It may be that the feature itself, by its very nature of removing the less serious consequences of inattention, is dangerous.
This reminds me of sports car reviews where they talk about how the combination of extremely grippy modern tires and traction/stability control mean that there is no warning before losing control, whereas older cars had gradual feedback.
Ultimately, I'm skeptical these are a good idea regardless of messaging.
I can see the appeal. Drivers get to experience cool new tech. Tesla/bmw/etc get a ramp that lets them develop autonomy gradually.
But at least in my uneducated opinion, autonomy-with-human-on-standby is a bad idea. Uncanny valley and whatnot.
Human in the loop is a terrible idea. We love doing make-up, texting, driving drunk, etc. We love road rage and silly egotistical lane changes or blocks and reduce efficiency of the roadways. It was fine at speeds of horses but not anymore. Especially with smartphones in the loop.
Partial autonomy and then transitioning to full autonomy is the future of transportation.
A car should be more like an elevator in functionality. Press a button, and take me there.
What? Because Elon wasn't able to achieve a milestone you consider it not possible at all? That's just absurd.
> Still a long way from full autonomous driving.
Yeah, still a long way from full autonomous driving, but having that training data make it that much closer. Theses cars have driven more miles than they ever could on test roads in a century.
The question is, does the path to full autonomy necessarily lie through partial autonomy? And related to that - is partial autonomy actually more dangerous than no autonomy at all?
Personally I think that full autonomy might be easier to hit from scratch on prepared roads, with sensors built into the environment, and the roads/lanes not shared with pedestrians/animals/animals driving cars.
I would consider making the equivalent of e.g. the London low emissions zone into an autonomous only zone, with the roads, environment and static cameras providing assistance to the cars, possibly centrally guided, and extending it out from there.
Not on my 2015 S 80D. There it is a small stalk on the left of the steering column. Up to go faster down to go slower. Double pull toward driver to enable Autosteer.
I am surprised there are not laws to have cars alert you to the fact your speeding for the systems that can read speed limit signs in addition to have current maps.
Well, there are systems that do this. Many new cars have speed limit sign recognition built in and they will display the current limit on the dashboard(and sometimes show you an icon indicating that the vehicle thinks you are speeding).
The reason why this is not more aggressive is because it's only seemingly a simple problem - in reality it's far far from it. In my experience of driving around the EU - I had a Mercedes CLS(brand new 2019 model) with adaptive cruise control and speed limit signs recognition, and while amazing when it worked, it worked maybe 50% of the time. Issues I noticed:
1) it kept detecting speed limit signs on roads parallel to the one I was on [0]
2) it kept detecting speed limit signs on side roads - leading to actual dangerous situations, because I was on a 50mph road, going past a 20mph side street and the car would suddenly hit the brakes because it detected a 20mph speed limit. That's 10000% unnacceptable in a production vehicle. [1]
3) it kept reading speed limit signs on the slip roads leaving the motorway - so you're doing 70mph on the motorway, and it suddenly reads the 50mph speed limit that's for the motorway exit. Same problem as with point 2.
4) In situations where there are no speed limit signs, it would use the map data - and even though this was a brand new, March 2019 vehicle, it had no idea that a stretch of motorway next to my house is not limited to 50mph anymore, and hasn't been for the last 2 years.
In practice, I had to keep overriding the speed limit chosen by the car constantly, and after few days I switched it fully off as it led to me nearly shitting my pants as the car hit the brakes for seemingly no reason whatsoever few times.
And then I have concerns about countries where the speed limit is only valid to the nearest intersection - and the definition of an intersection varies. There are paved roads merging with the main road which are not intersections because the side road leads to private property, and then there are dirt tracks which are intersections because technically it's a public road open to motor traffic. There are no signs to indicate either - the driver is just "supposed to know". It's idiotic, I know. And in some countries(Poland) the speed limit can change as often as 4-5 times in a span of a single kilometre. What's even worse is that there are situations when the speed limit can actually increase(!!!!) the speed limit set in a certain zone(built-up zones have a 50km/h speed limit, but this can be increased by a regular speed limit sign - and then yes, it goes back down(!!!) to 50km/h after an intersection). All of this stuff would be incredibly hard to code into a system like this, and it's one of the reasons why I suspect the proposal made by EU to have automatic speed limiters built into cars will either fail, or people will just keep disabling it altogether, making the whole idea pointless.
[0] Here it would think the side road has a speed limit of 50mph, because it would read the wrong sign:
https://www.google.pl/maps/@54.9929676,-1.5788994,3a,75y,78....
[1] Here for example the 20mph sign is placed at an angle to the road, so the car would read it going past it - and set the adaptive cruise control to 20mph which is incorrect for this street:
https://www.google.pl/maps/@55.0074774,-1.615172,3a,41.3y,90...
Same here:
https://www.google.pl/maps/@55.0071611,-1.6214135,3a,75y,240...
My sister has an insurance policy with a tracker(because she's quite young and there was no other option for her) and one day we noticed that her insurance score has fallen down through the floor, hugely negative points for....speeding. Of course the website doesn't tell you where or when, just that you're being penalized for being naughty. Ok, so we call them up, and they say they will send us the exact coordinates where the "offence" happened. Ok, cool - and turns out that they thought she was driving 70mph on a road going next to the motorway - which had a 20mph speed limit. Which was literally physically impossible, as previous GPS points were showing her on the motorway and there was no direct exit to that smaller 20mph road. So the GPS was off by maybe 10-20 metres at most, but the insurance has automatically penalized her massively for speeding. What's stopping the same system from failing when used for automatic speed limit adjustment? Especially in some extremely dense cities where you literally have roads of varying speed limits overlapping each other.
The insurer did correct that score penalty manually after I complained. But it's still a joke that it could make that mistake in the first place, especially since they can literally cancel your insurance due to speeding.
Google maps was (dangerously in the winter) directing people to an old road, into mountains after a new one had been built. It took national news, and several months of attempts to get them to update it.
The just build a big very accurate database approach is the one I would absolutely guarantee will never work.
If there is something out of date, hit the report button and it's typically acknowledged immediately, confirmed in a few hours, and public within a few days.
If it's a proposal by the government, they could do better: they could mandate not only automatic speed limiters, but also something embedded in the roads like balises (https://en.wikipedia.org/wiki/Balise) to reliably indicate to the car the correct speed limit.
Tesla Model 3 has this feature and, while autopilot is engaged, will decrease the speed of the car when the speed limit decreases.
The core idea is to continue to grant enough agency to the operator that they are held responsible as long as the vehicle faithfully executed the intent.
The tongue and cheek bloodsport throttle in the real world would instead be a proactive user choice of how much human guidance and interaction is required (solving the handoff problem, for example, 1 minute automatic, 1 minute manual or an automatic system that you'll need to irritatingly press a button every few minutes to stay engaged).
If a user options for "zero engagement" (let's pretend they're drunk or just got out of surgery), then there has to be a visceral experience of the underlying risks; a constant beeping and flashing, some real sense of urgency and warning.
I know it's a luxury product but cars are machines of carnage where death is never more than a mistake and mere seconds away. That reality should supercede.
This may be eventually removed, but a system like this would honestly solve all the current issues
Just like trains! There is an alarm system just as you describe. After a certain amount of time, there are audio and visual warnings for the engineer. The engineer must engage the switch to "snooze" the alarm, and in the event that the engineer does not engage, the train automatically cuts power to throttle to brake.
I'm not sure flashing lights and a klaxon are what we would want in a car, but maybe beeping akin to lane assist that quickly increases in volume + a light on the dash would be a good start.
Europe is mandating this (combined with limiting) for new cars soon. It will be possible to override 'temporarily'.
1) I don't believe we'll go "autonomous cars only" for predictable future. I can see that happening on limited set of streets(say, strict city centres), but not overall. Mostly because I think requiring everyone to buy an autonomous car is ridiculous, and then there are cars made even right now that I can totally see people driving in 20/30/50 years from now. We'll always allow manual cars on the road, just like we still allow horses - it just became so rare that in practice it's not an issue to have an animal with no safety features on the road next to modern motor vehicles.
2) there will always be users of the road that cannot be made automous - motorcycles. Unless you believe that motorcycles will be banned, but I don't think that will happen.
So yes, I think any speed limits of the future will be set for humans first and foremost. I do look forward to those speed limits to be revised up though - cars are far safer and more efficient at higher speeds than they were few decades ago, and yet speed limits have never been revised to reflect that. US is particularly bad at this, with empty stretches of straight road having a speed limit of 50mph for instance. That should have been fixed years ago.
Where anyone who actually drives a Tesla would get the idea that you could sleep while the car drives I can't imagine. The configuration screen where you enable it makes it perfectly clear that this is a driver assistance function that requires the driver to be able to take control. And every time you enable the function with the steering column stalk it reminds you in the dash to keep your hands on the wheel.
Yachtsmen sleep while on autopilot but that kind of autopilot is considerably less capable that Tesla's and is only safe while in the open sea away from busy shipping lanes, ditto aircraft. As far as I know commercial aircraft require at least one pilot to be awake and in the seat in the cockpit at all times whether or not autopilot is on.
Only a few short weeks ago Elon proclaimed Tesla will have 1 million fully-autonomous, value-appreciating robo-taxis driving the country within a year. Way too late for them to admit their technology is nothing but "assisted driver".
Perhaps they should rename it to assisted driving or something like that. I think I saw something about someone sleeping while on autopilot. There's only so much parenting the manufacturer of the car can do. The ethics of all this are much more complex of course - but perhaps the messaging needs to be clearer that autopilot is more akin to advanced cruise control as opposed to you can go to sleep, the car can drive itself.