Thứ Sáu, 6 tháng 1, 2017

Model X Autopilot Problem - Pretty Scary part 2

  • Jul 11, 2016
    SR22pilot
    The only mod was to disable the need to sometimes touch the steering wheel.
  • Jul 11, 2016
    lance
    I've had my S for 2.5 years now and I use a pre-beta version of Autopilot called my knee. It's very good on straight roads, in a pinch. I told the Tesla rep about this and she admitted she uses in her Honda. So far so good. I'll post a pic of my knee if anyone needs!
  • Jul 11, 2016
    KZKZ
    You just answered your own question.

    The Hyundai in the video demonstrates what the system is capable of doing, not what Hyundai allows consumers to do due to modifications in the software for the production version. Hyundai wisely chose to dumb down the system for the general public.

    The software changes Hyundai makes on the production car changes it from a "hands free" car to a car with a lane keeping assist feature.
  • Jul 11, 2016
    KaiserSoze
    I absolutely agree.

    I'm now going to make a controversial claim (freely up for debate): it is impossible that all AP users can pay attention all of the time. To begin with, there is the simple fact that the human is doing less when AP is enabled. Hands move less. Feet move less. Doing less = less brain activity. That is indisputable. The actual amount of the reduction in brain activity is difficult to answer, but it is there.

    Further, as AP gets better and better, people will, even involuntarily and subconsciously, pay less attention, or stop paying attention. Google observed this with their self driving car trials (and decided to skip to Level 4). There seems to also be ample evidence of this phenomenon among commercial pilots as well, who are responsible for hundreds of lives and whose job, literally, is to pay attention all the time, even when autopilot is enabled. Oh, you can warn them. You can nag them. You can berate them. But it will happen regardless.

    Thus, AP is baiting the user. It's like giving the driver an enormous bed, a comforter, and a pillow, but telling them it's not OK to fall asleep. Or setting a glass of water in front of a man dying of thirst. Or building a swimming pool in your back yard and not fencing it off. The executive center only a small part of the human brain. There are limits to its power. Fundamentally, the brain is a huge cost center to the body. It burns a lot of food, and food has been traditionally scarce, and thus the brain is optimized to reduce its power consumption when not strictly necessary.

    My conclusion, then, is that until all cars are completely autonomous, there will be accidents where humans are not capable of taking over from a disengaged or malfunctioning autopilot, and those will be impossible to eliminate. Blame the victims all you want--it won't help. As humans gain confidence in AP, the rate may, in fact, increase. The only question, then, is whether that situation is a net improvement in safety (fewer maimings and fatalities) or a net loss (more fatalities). I hope that, in the future, Tesla transparently publishes all of their accident data, and not carefully selected talking points, so that this may be independently judged.
  • Jul 11, 2016
    walla2
    If you can't pay attention with or without autopilot, I would contend that you shouldn't be driving. Since you don't have to pay attention to steering and the accelerator, you can devote your attention to cars around you and situations as they develop. Less energy on driving and more energy on awareness.

    The autopilot computer isn't there to lull you into comfort and disregard. If my family is in the car . . . I pay attention or they die. Pretty damn simple.
  • Jul 11, 2016
    KaiserSoze
    This is the controversial portion of my statement: I assert that even if you want to pay attention, or think you are paying attention, or even are paying attention, objectively, you are actually paying less attention than if you were doing 100% of the driving work.

    But there are people out there who aren't good drivers, period, even when they are paying attention, and people who are driving impaired (tired, drunk, etc.) and Autopilot will definitely result in a net decrease in accidents for them.
  • Jul 11, 2016
    walla2
    If you can't pay attention, then don't turn on autopilot ever.

    If you know you are unable to function as a responsible human being, then don't turn on autopilot.
  • Jul 11, 2016
    KaiserSoze
    I agree, that would work great in theory, but in practice it does not. People have huge blind spots when it comes to their own ego and their limitations. They cannot accurately assess risk. "One in a ten chance of developing cancer in my lifetime? Won't happen to me. One in a hundred million chance of winning the lottery? Could happen!" Most of the population considers themselves better than average drivers. Most people will only realize they weren't paying "enough" attention after they've been involved in a crash.

    And that assumes that a person can assess the amount of attention they are capable of giving ahead of time. You'd never say, "I can skip wearing a seatbelt, because I am responsible driver. If you are not a responsible human being, fasten your seatbelt." Why? Because accidents are not fully under your control.

    Similarly, unless you are a Buddhist monk with years of Vipassana meditation under your belt, your own attention is not fully under your control, either. Your attention can be lost, at times, accidentally, whether you're trying to pay attention or not.

    And that is why Autopilot will (if it hasn't already) make everyone safer by reducing the frequency of accidents caused by human error. I'd actually prefer that people who aren't good at paying attention, and choose to drive anyway, use Autopilot rather than not. Autopilot will reduce the chances that they will get into an accident.

    However, for the same reason that Autopilot prevents accidents--human errors occur all the time, whether you want them to or not--humans are not always capable of rescuing Autopilot when it screws up. And, therefore, "you should have been paying attention" is not a valid response to an Autopilot crash. These crashes are inevitable and some are simply not preventable, regardless of the responsibility of the driver.
  • Jul 11, 2016
    CHG-ON
    I keep my eye on AP all the time. I have never had a potentially critical event. But I think that's because I keep my eye on as the OP did. It has done a few oddball things. But nothing crazy. I always monitor its ability to slow for cars in front of me. If I don't like the rate of braking, I take over.

    We must remember to not get lulled into a state of complacency when using AP. It is still a machine that simply does not have the ability to process all the reasoning points that we humans can and it is VERY early technology.

    How many times has Siri or OK Google given completely crazy responses? Or your map guidance been wrong? It's the same exact thing.
  • Jul 11, 2016
    SR22pilot
    Wow, no matter what Tesla does it is bad and what others do is good in your eyes. How long should you be allowed to not touch the steering wheel since you seem to have very definitive opinions on this? Tesla would not be able to do the demo in the Hyundai video without a similar modification since the car would eventually slow down and come to a stop if the steering wheel wasn't touched. Deactivating the requirement to occasionally touch the steering wheel was the one mode Hyundai did. How about the Infiniti video? It could just as well be a Tesla video.
  • Jul 11, 2016
    stopcrazypp
    Actually, Tesla can't do the demonstration at all since it uses the seat sensor to disable autopilot (after someone pulled the empty seat stunt shortly after autopilot release). AFAIK none of the other automakers does this, but you don't see people going on their forums to take issue with that. Same with the advertising (for example Mercedes advertising their car as "self-driving"). People seem to only take issue if it is Tesla.
  • Jul 11, 2016
    SR22pilot
    First and foremost it is the double standard that irks me. It makes having an intellectual debate difficult.
  • Jul 11, 2016
    woof
    I dispute it. Less physical activity doesn't mean less brain activity...examples: Air Traffic Controllers and Dr. Stephen Hawking . Contra example, Professional Football players.

    I've no experience with Tesla's AP, but with ACC I find I am much more keenly aware of my situational awareness, as I can safely spend a few more moments looking in the mirrors, turning my head all the way left, as I can trust ACC to look ahead at the lead car during those short moments, and as I am looking far ahead beyond the lead car when looking forward the chances of an ACC failure are minimized. So in my particular case, I find I can make up for less driving load with MORE brain activity.

    The combination of AP covering for driver inattention, and the driver covering for AP inattention, makes for a system that is safer than either individually.
  • Jul 11, 2016
    cb32000
    As self driving software goes from 0% reliable to 100% reliable, the amount of people trusting the software to not fail will also go from 0% to 100%

    There will always be people who will become distracted more than they should and that is a problem, Google figured this type of natural human behavior out early and went for full autonomy.

    If autopilot becomes 95% reliable how many people will become distracted, how about 99%, 99.9%?

    It is currently the ultimate distracted driver thing out there when someone trusts it and stops paying attention
  • Jul 11, 2016
    SSD420
    I just used autopilot all the way to my cottage 220km. I only grabbed the wheel twice when my MX told me to, and each time was for a mere few seconds on big curves. The 3rd I felt a bit uncomfortable as I felt I was a bit too close for comfort to a transport truck.

    I no longer hover my hands up but I am very alert when I'm driving. I know that it's not fool proof but Autopilot + a competent driver (me!) = a pretty safe drive!

    I've also started filming the other drivers back when they film me! Or when there are kids in their passenger seat to the left of me, I tend to cover my eyes and spook them! (I always make sure that I've covered my eyes loosely, so I can still see the road. I'm still aware that our cars are in "beta" and I get a little nervous when the silhouette on my dash is only a car when it's actually a truck.
  • Jul 11, 2016
    SSD420
    image.jpeg I thought this was cute though...you can see the motorcycle! :D:p
  • Jul 11, 2016
    ecarfan
    What Tesla does not need right now is someone posting a photo online showing a Tesla driver on the road with their hands covering their eyes. I know you think you are just being amusing to children, but think of the potential for bad PR for Tesla.

    This is the age of the Internet. Any photo, taken anywhere, can potentially be viewed by billions of people.
  • Jul 12, 2016
    walla2
    Motortrend had to override the model S 0.5-2% of the time during a recent test drive. The system is already within your parameters with one caveat . . . some idiots can't protect their own lives and have wanton disregard for the lives of their family members and other people? Pay attention.

    Texting and driving is dumb. Not focusing on what your 5-7k lbs car is doing is stupid. You have no excuse except that you made a very bad choice (wrong road, inattention) and likely will blame the technology because you fail to take responsibility for anything. If you feel autopilot distracts you from your responsibility, don't turn it on. Ever.
  • Jul 12, 2016
    lklundin
    As an experienced, professional software developer I can confidently state that in this field there is quite a lot of creativity when it comes to explaining why your delivery is going to be late...
  • Jul 12, 2016
    X Yes?
    These accidents and more to come, show this software is not ready to be on the public roads. In this case, GM's "creativity" of delaying the software will be a image and cost saving move. Tesla is already paying the price with safety investigations, bad press and a lawsuit.
  • Jul 12, 2016
    lklundin
    Your assertion regarding Tesla is exaggerated and premature at best.

    As for Cadillac, it is important to realize that the current state of their semi-autonomous 'Super Cruise' mode is such that it did not actually qualify for entrance in a recent self-drive benchmark, as reported by motortrend:

    Testing (Semi) Autonomous Cars With Tesla, Cadillac, Hyundai, and Mercedes - Motor Trend

    So Cadillac is not even able to provide their customers with anything close to what Tesla offers.

    As such their "image and cost saving move" comes across as specious.
  • Jul 12, 2016
    jelloslug
    If the OP thinks that adaptive cruise problems are only a Tesla thing then they are very misinformed.
  • Jul 12, 2016
    dwebb66
    Are you saying that, because there are accidents with Autopilot enabled, they should not allow it to be enabled? To take that logic to an extreme, should people be outlawed from driving as they cause 100's of accidents a day??

    If the net affect of having Autopilor enabled is a lowering of the accident rate, isn't that a good thing????
  • Jul 12, 2016
    X Yes?
    Absolutely. I'm sure the safety of AutoPilot and other similar systems will be the focus of study. It has to be relevant data. The data Tesla currently sights for safety, does not tell the story. Statistics that include data when AutoPilot would not be used are meaningless.
  • Jul 12, 2016
    X Yes?
    Since SuperCruise has been deemed "not ready to be released", we don't know how it works. Obviously the adults in their new legal dept. decided it wasn't ready and was not worth the risk.
  • Jul 12, 2016
    X Yes?
    UBSanalystsColin LanganandEddie Hsiehasses the news thatTesla Motors(TSLA) is now beinginvestigated by the SEC for failing to disclose the death of a driverusing its Autopilot system: That said, we believe Tesla�s comparison of the Autopilot fatality as the first in >130 million miles of driving vs. all US vehicles (one in every 94 million miles) doesn�t take into account that Autopilot is operated in certain restricted conditions (highway) and with human oversight, while the overall US stats include all environments and a broad range of safety technology.
  • Jul 12, 2016
    walla2
    Some of these accidents including the one death occurred on non-recommended roads so you may want to rethink your argument.
  • Jul 12, 2016
    electricity
    Everyone is using AP everywhere.
  • Jul 12, 2016
    X Yes?
    In the city, undivided roads, stop and go LA freeway? Troubling. Thanks for the warning!
  • Jul 12, 2016
    Todd Burch
    I use it virtually everywhere it works. Never had an even remotely scary moment. Why? Because I monitor it and pay attention. Y'know, like you're supposed to.
  • Jul 12, 2016
    brucet999
    It may be possible in Georgia to allow 300 ft behind the car in front of you, but if you leave that much space in CA, at least three other cars will rush in to fill the gap. :)
  • Jul 12, 2016
    brucet999
    You haven't experienced "truck lust" yet? Having your car suddenly swerve at the back corner of a truck in the lane next to you at 75mph is indeed scary. Happened 5 times in the course of about 400 miles during my last road trip.
  • Jul 12, 2016
    Todd Burch
    Nope. I experienced it slightly bias toward a truck as I passed it on 7.0, but never had it swerve toward a truck.
  • Jul 12, 2016
    roblab
    You are right. The thread should be titled, "Model X DRIVER problem: Pretty Scary."
  • Jul 12, 2016
    jdw
    In slower busy traffic I find I need to use 1 or 2 - any higher setting results in people constantly diving in front of me. Makes for a very jerky and unsafe ride when they do. I find no problem with a 2-3 second follow time (setting 5-6-7) in normal highway traffic.
  • Jul 12, 2016
    Topher
    I dispute it.

    Quite simply it is false. Brain activity just can't be modeled anything like that simply. The fact that you would try to bully people into believing you by say it is indisputable, just makes it less believable.

    Thank you kindly.
  • Jul 12, 2016
    KaiserSoze
    You don't have to take my word for it. Try this: Engage Autopilot. Focus on "driving." Don't attempt to shadow drive, however: don't move your hands and feet.

    Now disengage Autopilot and begin driving. Remain just as focused as you were on your previous subjects of attention. Don't let that waver. Operate the steering, brakes, and accelerator. Do you imagine your brain activity went down? How could it, when you changed nothing about your brain except adding several new responsibiliies? Now, consider your motor cortex. Do you think that adding wheel and pedal movements caused less activity in that region? Even kept it the same?

    What do you think happens in reverse?
  • Jul 12, 2016
    ohmman
    I don't have a position for or against the increase in brain activity in the context provided. However, I have to disagree that what we "imagine" or "feel" has any bearing on what's actually happening. The word indisputable shouldn't be slung around too easily. Almost everything's disputable to some degree by reasonable people.
  • Jul 12, 2016
    Mark Z
    With all the news coverage and discussion about Autopilot, I think back to the days of the backseat driver. You know, the person who is constantly telling the driver what to do. In those days, it was normally best to ignore the last second bad advice and shouts to "watch out!" that could cause an accident if the brakes were applied without judgement.

    May I suggest that Autopilot is like the backseat driver. It nags you to watch the road, grab the wheel and pay attention. But in this case, you better watch, listen and do what the backseat driver tells you.
  • Jul 12, 2016
    KaiserSoze
    Harumph! WelI... I dispute that.
  • Jul 12, 2016
    trils0n
    Despite thinking they are, humans aren't very good at multi-tasking. The more tasks they try to engage in at once, the more their performance degrades.

    I only have a few hundred miles of autopilot driving under my belt, but it was pretty clear to me I could much more effectively focus on my surroundings and increase situational awareness when I didnt have to do the additional task of making small course and speed corrections all the time.

    The same goes for driving. Last time I went go-kart racing the track had events where each person got their own laps. It is easy to focus soley on driving because there is no need to look for or be aware of other drivers because there are no other drivers allowed on the track. All brain power to the task at hand, controlling the vehicle.

    I guess you could use the Autopilot as an opportunity to do something else besides monitoring your surroundings, but that seems in direct conflict with the operating instructions that say you should remain alert and ready to take over.
  • Jul 13, 2016
    TacC
    Bingo.

    If you're focused on the conditions around you, AP augments your ability and makes you and your riders more safe. If you're using AP to substitute for your driving, you're probably only safer than a distracted or drunk driver.
  • Jul 13, 2016
    SR22pilot
    Same here but that is why there is always a crash on the connector every day. Now imagine cars automatically applying full braking. Yeah, it might be the fault of the guy behind you but you still have a wrecked car.
  • Jul 13, 2016
    June P
    When I use adaptive cruise control with Subaru Eyesight I don't typically set it to much higher than the speed limit, for starters. I haven't experienced a situation where the car kept speeding up and never slowed down with a car in front of me but it has at times waited until the last second to slow down. I am sure you slowed it down quickly. The car might have eventually slowed it down in time not to hit the vehicle ahead but no sense taking a risk.

    I don't think Tesla should be rolling out anything less than fully tested with a "beta" version offered of safety software except with its test drivers however.

    I'm going to assume the vehicle might have slowed down if you had waited but you had told it to go too fast to take a chance on waiting. It does make me think it's got to be heavily monitored as to actually keeping the space between you and the car in front of you.

  • Jul 13, 2016
    walla2
    I am a neurological surgeon. I disagree with your view of the attention center of the brain. It is not the motor cortex.

    You may not be wrong for some people but again those that can't follow beta recommendations should not use autopilot. Those that can't focus in cruise control also should not use cruise control. Those that can't turn off the brights from blinding others shouldn't turn them on. Take some responsibility about your capabilities and incapabilities.
  • Jul 14, 2016
    KaiserSoze
    I did not claim that. Or if I did, it wasn't my intention. Maybe the motor cortex plays some role in maintaining vigilance, but it's not the center of attention, or executive function, in my view.

    I chose it because it can reasonably be assumed to be more active when driving than not, and because to override Autopilot in an emergency, you presumably have to fire it up.

    As an aside, I find this article quite interesting: This Is Your Brain on Sports, in which it is claimed that experiments have shown that only 20% of motor cortex neurons used when playing sports are activated when merely watching sports. I would imagine it's similar for driving. It also mentions that, because we learn to do many tasks in a sensory-motor feedback loop, the motor cortex is linked to, and used for, for perceiving and comprehension of those tasks.

    However, I don't want to get too far down this rabbit hole, because the main (albeit tenuous) support for my thesis, summarized as, "Autopilot turns (some) formerly good drivers into zombies" is empirical, not theoretical. In at least three Autopilot crashes (stalled vehicle, Montana, tractor trailer), we have seen evidence that the drivers had time to avoid their accidents, but they did not. If Autopilot were disabled any reasonable driver could have avoided these accidents. Yet they did not.

    Is it random chance brought into focus by selection bias? Could be. However, I find it hard to believe that normal, unzombied drivers can miss tractor trailers crossing the road. If they do, we would see this type of accident much more frequently. I maintain that these drivers could not prevent these accidents because they were were zombified by Autopilot. They were lulled into inattention, or at least into a state where they couldn't readily regain control of the car quickly enough. They may revert to being good drivers when Autopilot is disabled, and they may be eminently responsible, and thus they may think "I can handle Autopilot" but, in reality, they can't handle the truth: that after they flip the switch, they're eventually right out of Shaun of the Dead.
  • Jul 14, 2016
    walla2
    Accident 1: Not paying attention, speeding, using autopilot on unapproved road.
    Accident 2: Blaming Tesla for a car crash; not paying attention and cited for careless driving.
    Accident 3: Not paying attention, speeding, using autopilot on an unapproved road

    I only see one real similar finding from the three crashes. Autopilot doesn't cause inattention. People not paying attention kills people whether they have autopilot or not.
  • Jul 14, 2016
    walla2
    Of course less neurons in the motor cortex would be active when not moving vs moving since neurons in the motor cortex are responsible for moving. That doesn't mean we can't pay attention when watching sports. Your argument would suggest never to run advertising during sports shows because we aren't really watching them. Your argument would suggest that humans are incapable of reading since the lack of motion would lull us into inattention.

    Attention is optional but when driving any car should not be.
  • Jul 15, 2016
    cb32000
    while I agree that everyone should pay attention I do not share your position that everyone on the road will and if they don't it's their problem

    If the thousands of people a day that share the road with me were you I probably would not worry but since 30k people a year die in car accidents and a great majority are due to driver errors I think "pay attention" is not going to cut it when most of the time nothing happens and a small fraction of the time a 4000 lb car crashed into someone else
  • Jul 16, 2016
    walla2
    Most of the time a ton of stuff is happening when you are going over 60 miles per hour.

    People need to take responsibility and stop blaming others for their faults. That's what these accidents have shown people are incapable of today. As a society, we should not make excuses for people that can't follow directions or simply pay attention when they are piloting a large mass down the roads. Tesla never said they were doing anything but helping. I'm not a Tesla fanboy. I have numerous complaints about them. Personal responsibility isn't their business. It's ours.
  • Jul 16, 2016
    walla2
    That isn't my position. I'm not talking about everyone. I'm talking about the top 2% of income earners that can afford a Tesla with beta software and should know better.
  • Jul 16, 2016
    cb32000
    OK that makes more sense, I would suggest that Tesla does not allow anyone to activate the autopilot until they actually agree to be part of an unpaid untrained test pool that understands they will assume all liability for any defects in the autopilot software, or that tesla compensate the people as tesla contractors and tesla assumes the liability.

    They better figure this out before the 35k car becomes available since the Model 3 will be available to way more people than the 2% high income, intelligent, and responsible people on the road.
  • Không có nhận xét nào:

    Đăng nhận xét