Tesla issues strongest statement yet blaming driver for deadly crash • Hacker News Board

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#761

Post by Piskvorrr » Thu, Apr 12, 2018, 12:37 PM UTC

simias wrote: post #758

corney91 wrote: post #742

Earlier quoted context omitted.

> If they had the ability to measure confidence in the system, they would use it to issue warnings or disable autopilot Yeah, that's basically what I'm asking for, maybe a warning light at a certain threshold would be a better default, personally I'd still find a number more trustworthy though. > The "confidence meter" would almost certainly be 100% right up until it crashes into an obvious obstacle If image recognit…

>I'd be surprised something more complicated like road navigation was 100% certain all the time. If for instance it failed to detect an obstacle how could it tell you that it failed detecting an obstacle? Or that it got the road markings wrong? Given that the consequences for making a mistake might be people dying I certainly hope that's not how it works under the hood. Anything less than "I'm absolutely sure that I'…

Anything less than "I'm absolutely sure that I'm not currently speeding towards an obstacle" - that's an unattainable goal; there's no way to actually reach this, whether a human or a robot is driving. I'm overwhelmingly sure during the most part of my drives, and mostly sure for the rest - but absolutely? Never.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#762

Post by xHopen » Thu, Apr 12, 2018, 12:40 PM UTC

I have to say I was a firm supporter of tesla's autopilot, but you know, this is rediculous, Tesla should be ashamed say sorry, say oour system is crap , is a demo version and it doesn't work, and thats it. The rest is history, I was about to get a new tesla, not anymore.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#763

Post by Piskvorrr » Thu, Apr 12, 2018, 12:41 PM UTC

ryanlol wrote: post #607

maxxxxx wrote: post #554

Earlier quoted context omitted.

" Tesla's system demands more from drivers than manual driving. The driver has to detect automation failures after they occur, but before a crash. This requires faster reaction time than merely avoiding obstacles." I have been thinking that too. At what point do you decide that the autopilot is making a mistake and take over? That's an almost impossible task to perform within the available time.

Taking over from the autopilot is very easy unless you are deliberately risking your own life.

Sure. But from this, it follows that you'll have so many AP disconnects that you might as well keep the AP off, defeating its purpose.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#764

Post by tremon » Thu, Apr 12, 2018, 12:52 PM UTC

amelius wrote: post #735

maxxxxx wrote: post #554

Earlier quoted context omitted.

" Tesla's system demands more from drivers than manual driving. The driver has to detect automation failures after they occur, but before a crash. This requires faster reaction time than merely avoiding obstacles." I have been thinking that too. At what point do you decide that the autopilot is making a mistake and take over? That's an almost impossible task to perform within the available time.

It's a lot like sitting in the passenger seat next to a driver who suddenly turns out to be having a stroke. At what point do you grab the wheel? Clearly, it's a very dangerous situation that you don't want to happen, and certainly not on a regular basis.

But if the driver is having a stroke, you could likely get visual or auditory cues from the driver, you wouldn't have to infer it from the vehicle behaviour.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#765

Post by Robotbeat » Thu, Apr 12, 2018, 12:57 PM UTC

brain_dev wrote: post #723

Robotbeat wrote: post #45

Earlier quoted context omitted.

This isn't self-driving, though. It isn't called self-driving.

How else would you describe an Automatically Piloted Vehicle then? If it's not self-driving, don't call it Autopilot, and don't lull people into a false sense of security. Even with alarms to try and keep the driver engaged, people are inherently going to place more faith in 'Autopilot' than 'Enhanced Cruise Control' or something similarly named. The marketing around this feature is reckless, and Tesla + Musk should…

It’s called autopilot because “autopilot” already has a technical meaning which doesn’t imply self-driving. Heck, some aircraft autopilots are so primitive they only keep the aircraft in level flight without terrain following (although some do that). No versions of autopilot mean the pilot doesn’t have to pay attention. This is all already established through broad use before Tesla introduced the term to cars.

The same linguistic games you’re trying to play for “autopilot” apply just as well to “cruise control” or worse, since that term didn’t have an already-understood technical definition when introduced to cars.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#766

Post by Silhouette » Thu, Apr 12, 2018, 12:59 PM UTC

mijamo wrote: post #756

Silhouette wrote: post #557

Earlier quoted context omitted.

There's no need to make anything up. Mass transit systems are relatively efficient if and only if they are used on routes popular enough to replace enough private vehicles to offset their greater size and operating costs (both physical and financial). That usually means big cities, or major routes in smaller cities at busier times. Achieving 24/7 mass transit, available with reasonable frequency for journeys over bot…

Why so big though? I lived in a 25 000 people town in Sweden and did not need a car more than a few week ends per year. There was 5 bus lines for local transport, and long distance busses and trains with quite high frequency. And that's not taking into account the fact that bicycle is a very viable way to move around in cities I have actually never owned a car, I just rent some once in a while to go out somewhere whe…

You've almost answered your own question, I think. Providing mass transit on popular routes at peak times is relatively easy. It's more difficult when you need to get someone from A to B that is 100 miles away, and then back again the same day. It's more difficult when you are getting someone from A to B at the start of the evening, but their shift finishes at 4am and then they need to get home again.

To provide a viable transport network, operating full-time with competitive journey times, without making a prohibitive financial loss or being environmentally unfriendly, you need a critical mass of people using each service you run. That generally means you need a high enough population density over a large enough urban area that almost all routes become "main routes" and almost all times become "busy times".

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#767

Post by GhostVII » Thu, Apr 12, 2018, 1:03 PM UTC

mrb wrote: post #463

chrisper wrote: post #146

Earlier quoted context omitted.

A safety feature is useless if you can't rely on it all the time.

An imperfect safety feature is still better than no safety feature at all.

But if the safety feature works almost all the time, and occasionally fails catestrophically, I would argue that it is actually more dangerous than having no safety feature, since people will depend on it working and not catch it when it fails.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#768

Post by icc97 » Thu, Apr 12, 2018, 1:05 PM UTC

skgoa wrote: post #759

icc97 wrote: post #749

Earlier quoted context omitted.

> Partially automated cars have been driven by average people on the public roads since 2006, What the GP was referring to, which is what I've see it referred to in Google presentations is level 3 automation. That is what autopilot is supposed to be, where you can take you hands off the wheel and the car can drive itself but needs constant supervision. What you're referring to is passive assistance such as collision…

I'm talking about active level 2 systems, which the Autopilot is an example of. A level 2 system has to be monitored 100% of the time, because it can not be trusted to warn the driver and fall back to manual automatically in every situation. A level 3 system is robust enough to make this guarantee, which means that the driver can take his/her hands off the wheel until prompted to take back control. Autopilot is a lev…

Yep, fair enough, I was wrong Autopilot is level 2.

However I still disagree with your original comment:

> Sorry, but this is the exact wrong conclusion to make from the available data. Partially automated cars have been driven by average people on the public roads since 2006, when the Mercedes S class got active steering and ACC. There is no data that says that these systems, which are available on cars as affordable as e.g. a VW Polo, lead to more crashes. In fact the only data I have seen suggests that they reduce crashes and driver fatigue significantly.

Firstly, I don't see any examples of VW Polo claiming to have a steering assist, even the most recent 2017 edition [1] only has level 1 features [0]. Steering assist only came to the Mecedes S-Class in 2014 [2].

What I (and the parent comment you initially replied to) was tying to claim was that where as Level 1 features [0] (cruise control, automatic braking, and lane keeping) are great, level 2 (and level 3 according to Google/Waymo) is risky because humans cannot be trusted to be alert and ready to take over.

[0]: www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Vehicles_Policy.pdf

[1]: https://en.wikipedia.org/wiki/Volkswagen_Polo#Sixth_generati...

[2]: https://www.motorauthority.com/news/1080502_2014-mercedes-be...

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#769

Post by KarenSatantsby » Thu, Apr 12, 2018, 1:13 PM UTC

maxxxxx wrote: post #554

Animats wrote: post #552

Tesla's system demands more from drivers than manual driving. The driver has to detect automation failures after they occur, but before a crash. This requires faster reaction time than merely avoiding obstacles. Here's a good example - a Tesla on autopilot crashing into a temporary road barrier which required a lane change.[1] This is a view from the dashcam of the vehicle behind the Tesla. At 00:21, things look norm…

" Tesla's system demands more from drivers than manual driving. The driver has to detect automation failures after they occur, but before a crash. This requires faster reaction time than merely avoiding obstacles." I have been thinking that too. At what point do you decide that the autopilot is making a mistake and take over? That's an almost impossible task to perform within the available time.

Driving a 2017 Corolla on busy highway is more exhausting than driving a 2007 Honda civic - the Corolla has all this stuf like it moves the stearing wheel if I drift complains if I dont signal and jerks the wheel - slows down on cruis control and then speeds up.. I would be out of it driving a more bossy car like a tesla. I would rather the car just do the driving rather than act like my dad when I was 15.

Re: Tesla issues strongest statement yet blaming driver for deadly crash

#770

Post by salawat » Thu, Apr 12, 2018, 1:17 PM UTC

corney91 wrote: post #742

shawabawa3 wrote: post #731

Earlier quoted context omitted.

If they had the ability to measure confidence in the system, they would use it to issue warnings or disable autopilot The "confidence meter" would almost certainly be 100% right up until it crashes into an obvious obstacle

> If they had the ability to measure confidence in the system, they would use it to issue warnings or disable autopilot Yeah, that's basically what I'm asking for, maybe a warning light at a certain threshold would be a better default, personally I'd still find a number more trustworthy though. > The "confidence meter" would almost certainly be 100% right up until it crashes into an obvious obstacle If image recognit…

The problem is, given a neural network, the decided course of action IS at 100% confidence.

Unless you design the network to specifically answer "how similar is this?" and pair it with a training set and the output of another executing neural net, a "confidence meter" isn't possible.

This is one of the trickier bits to wrap one's mind around with neural networks. The systems we train for specialized tasks have no concept of 'confidence' or being wrong. There is no meta-awareness that judges how well a task is being performed in real-time outside of the training environment.

Humans don't suffer from this issue (as much) due to the mind bogglingly large and complex neural nets we have to deal with the world. Every time you set yourself to practicing a task, you are putting a previously trained neural network through further training and evolution. You can recognize when you are doing it because the process of doing so takes conscious effort. You are not 'just doing ', you are 'doing , and comparing your performance against an ideation or metric of how you WANT to be performed'. This process is basically your pre-frontal cortex and occipital lobe for instance, tuning your hippocampus, sensory and motor cortex to perfect the perfect tennis swing.

When we train visual neural networks, we're talking levels of intelligence supplied by the occipital lobe , hippocampus, alone. Imagine every time you hear about a neural net that a human was lobotomized until they could perform ONLY that task with any reliability.

Kinda changes the comfort level with letting one of these take the wheel doesn't it?

Neural nets are REALLY neat. Don't get me wrong. I love them. Unfortunately, the real world is also INCREDIBLY hard to safely operate in. Many little things that humans 'just do' are the results of our own brains 'hacking' together older neural nets with a higher level one.