Tesla crash in September showed similarities to fatal Mountain View accident • Hacker News Board

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#561

Post by ceejayoz » Fri, Apr 06, 2018, 2:55 AM UTC

mrguyorama wrote: post #499

ceejayoz wrote: post #400

Earlier quoted context omitted.

There's no reason such a system would have to be engineered to be totally trusting. The car, in that situation, would likely be set up to say "nope, my sensors say that's not safe" and hand off to the driver or reject the input. Use it just for additional input to err on the side of safety. If the car ahead says "hey, I'm braking at maximum" and my car's sensors show it's starting to slow down, I can apply more immed…

>The car, in that situation, would likely be set up to say "nope, my sensors say that's not safe" and hand off to the driver or reject the input. You realize you say that in a thread about a car's sensors failing to understand it was in an unsafe condition (again), right?

You realize that the comments in a thread may head off in a related but distinct direction, right?

We're discussing car-to-car mesh networks being used to provide self-driving cars with more decision-making input.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#562

Post by ceejayoz » Fri, Apr 06, 2018, 2:57 AM UTC

Piskvorrr wrote: post #543

ceejayoz wrote: post #400

Earlier quoted context omitted.

There's no reason such a system would have to be engineered to be totally trusting. The car, in that situation, would likely be set up to say "nope, my sensors say that's not safe" and hand off to the driver or reject the input. Use it just for additional input to err on the side of safety. If the car ahead says "hey, I'm braking at maximum" and my car's sensors show it's starting to slow down, I can apply more immed…

That last part is actually scariest: blame the victim, yay. Unless they can prove they had a functional, powered, up to date, compatible I'm-a-car responder, it's their own damn fault for being invisible to the SDV. How about "I'm a human, don't hit me"? Does that also sound like a good idea (RIP Elaine Herzberg)? In other words, "all other road users should accommodate my needs just to make life a bit easier for me"…

I think that's a needlessly uncharitable interpretation.

Having cars communicate information to each other has the potential to be an additional safety measure. It's like adding reflectors to bikes - adding them wasn't victim blaming, it was just an additional thing that could be added to reduce accidents.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#563

Post by bendauphinee » Fri, Apr 06, 2018, 3:30 AM UTC

Ajedi32 wrote: post #393

ajross wrote: post #380

Earlier quoted context omitted.

> 1. Ethical. One thing is that you do a stupid thing and die. Another is that a technology fails in trivial circumstances that in theory are avoidable. A doctor can make also errors, but a medical device is required to be very safe and reliable in what it does. This needs elaboration or citation. You're one of the only folks I've seen come straight out and make this point: that somehow a technology or practice that…

I think the point he's making is that it's worse if a person dies because of something they have no control over (self-driving car malfunction) than if a person dies because of their own stupid choices (driving drunk, driving too fast for conditions, running red lights, etc). This, of course, ignores the fact that stupid choices drivers make tend to affect other people on the road who did nothing wrong, so the introd…

> because of something they have no control over

Expect in this case, the driver has 100% control.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#564

Post by mdorazio » Fri, Apr 06, 2018, 4:42 AM UTC

Dylan16807 wrote: post #560

mdorazio wrote: post #557

Earlier quoted context omitted.

Subtracting your own velocity from what? I've been talking about individual sensor reporting with conflicting information and the complexity of decision making that results. A radar return from a rear bumper at relative zero miles an hour is exactly the same whether the car is actually in motion or not. Reference frame and velocity relative to the ground is up to your software algorithms to determine, not the sensors…

> Subtracting your own velocity from what? You said the camera/radar would be reading another car as "stationary", which is a velocity number of zero. Add/subtract your speed and you get the real speed. A non-defective sensor suite is going to give you either distance or velocity, and you can use distance measurements to calculate velocity really easily. > A radar return from a rear bumper at relative zero miles an h…

Maybe I'm not being clear on my core argument. sensors give conflicting information in different scenarios. One sensor might read an obstacle to be avoided while another shows a clear path. Deciding what to do in these conflict situations is difficult. Saying it's easy is effectively saying that the hundreds of very intelligent engineers working on these systems are idiots because it looks like it should be a piece of cake. You're using similar logic to non-technical people who get angry when a developer can't give an accurate estimate of time needed to implement a new feature - (s)he has the code and requirements right there, it should be easy, right?

I'm quite familiar with how vector math works. Your comments above seem to still be missing that I am talking about situations from the perspective of individual sensors, which have no ability to make judgments about different frames of reference. That only applies at the level of the full driving system. For example, to a radar sensor alone, there is no difference between a car parked 10ft in front of you while you are motionless relative to the ground and a car 10ft in front of you while you are both moving at 60mph relative to the ground. Both are "stationary" as far as the return signature is concerned. There will be no doppler shift in either case. The driving system has to combine this with other sources of information for it to be useful, and that's where problems creep in.

> That's because such things are irrelevant to my argument. I'm only talking about how algorithms should handle objects that have already been found.

We're arguing two perspectives of the same higher-level opinion (algorithms are insufficiently developed to be safe enough for autonomous driving). What I'm trying to say is that there are a number of fuzzy steps to get from sensor readings to actual object detected, and then from actual object detected to "known obstacle" classification, as you put it. I don't think I'm going to be able to argue this case clearly enough in the comments here, so I'm going to add this to my longer article writing list. Thank you for the constructive debate on this.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#565

Post by mike10010100 » Fri, Apr 06, 2018, 4:48 AM UTC

ksk wrote: post #537

stale2002 wrote: post #500

Earlier quoted context omitted.

> I think the point he's making is that it's worse if a person dies because of something they have no control over Perhaps. I reject this argument though. A death is a death. And if you prevent lifesaving technology then you are responsible for a whole lot of deaths, irregardless of how "deserved" those deaths are.

>Perhaps. I reject this argument though. A death is a death. A death caused by shitty engineering (e.g. Tesla in this case) is not the same as a death caused by gross negligence on ones own part. >And if you prevent lifesaving technology then you are responsible for a whole lot of deaths, irregardless of how "deserved" those deaths are. How many deaths has Tesla prevented, which would have happened otherwise?

> A death caused by shitty engineering (e.g. Tesla in this case) is not the same as a death caused by gross negligence on ones own part.

The result is the same, no? Isn't the idea to save lives? Why does one life have more value than another's simply because someone's death was caused by their own negligence?

What about the other people on the road? Do their lives matter less because of the gross negligence of the person not paying attention while they're driving?

This issue is a lot more complicated than you're making it seem.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#566

Post by unityByFreedom » Fri, Apr 06, 2018, 5:03 AM UTC

gfodor wrote: post #493

unityByFreedom wrote: post #472

Earlier quoted context omitted.

> What do you say to people like me, who, if you took the feature away, would be forced to reduce their safety? I would say we should go by the numbers. How many accidents and deaths are there per mile, on similar roads (divided), from vehicles of similar class (new, luxury), in similar weather conditions (fair)? Compare that to miles where autopilot is enabled and we have something. As yet, I don't believe Tesla is…

I agree that the data will be helpful, but it's also important to remember that "Autopilot" itself is a moving target. Tesla has already made changes that are designed to increase the degree to which drivers are forced to pay attention when the system is engaged. In my view, the ultimate goal should be to ensure that Autopilot is used in the way I described, and is disabled for drivers who are unable to use it that w…

Moving target, sure, but Tesla has been making statements claiming using AP is safer than not since AP's inception. You can't have it both ways. Even Tesla couldn't be certain whether certain changes lowered the accident rate or raised it. We should look at total accidents, not miles driven since the latest version.

Perhaps the best comparison is right within Tesla's data. Compare miles driven in the same model with AP activated and with it turned off on the same roads.

> In my view, the ultimate goal should be to ensure that Autopilot is used in the way I described, and is disabled for drivers who are unable to use it that way

Actually there has been some work in determining when drivers are distracted [1]. This seems to be what you suggest, something that can be used to disable AP under certain conditions. Would you support Tesla installing such a system?

[1] https://www.kaggle.com/c/state-farm-distracted-driver-detect...

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#567

Post by wastedhours » Fri, Apr 06, 2018, 7:07 AM UTC

gsnedders wrote: post #488

wastedhours wrote: post #345

Earlier quoted context omitted.

Had to thread a van through a temporary concrete width restriction the other day - when it's that narrow, even the Uber behind me wasn't giving me grief for going that slowly! The country roads one has always dumbfounded me though - why some of those have national speed limits I will never know.

> The country roads one has always dumbfounded me though - why some of those have national speed limits I will never know. Why not? Even roads with lower speed limits you're required to drive at a speed appropriate for the road, the conditions, and your vehicle; the speed limit merely sets an upper-bound, and it's not really relevant whether it's achievable. Just look at the Isle of Man where there is no national spe…

If you set a limit, some people will drive it regardless. Even if you're supposed to drive to the road and conditions, there are enough utter morons out there who'll take a blind narrow corner at 60.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#568

Post by bsder » Fri, Apr 06, 2018, 7:25 AM UTC

ghaff wrote: post #448

bsder wrote: post #126

Earlier quoted context omitted.

> Imagine the autopilot is 100 times safer than human drivers. Full implementation would mean about a fatality a day in the US. It does not seem this would be acceptable by the public even though 30,000 annual human car fatalities could be avoided. Your opinion will be irrelevant--the insurance companies get the final say. When that safety point arrives, your insurance rates will change to make manual driving unaffor…

>your insurance rates will change to make manual driving unaffordable I see people saying that. Why would insurance be more expensive than today--implying that manual driving becomes more dangerous once self-driving cars become common? It's already a common pattern that assistive safety (and anti-theft) systems tend to carry an insurance discount versus not having them.

Because the "manual" driver will almost always be at fault and the automated drivers will have the telemetry to prove it. For example, I recently got hit while backing out of my parking space--I'm legally at fault because of how the collision occurred. The fact that the other driver had tremendously excessive speed (a center mass hit on a Prius that literally lifted it onto the SUV sitting next to it) wasn't proveable by me--a self-driving car will not have that problem.

This means that the insurance companies with more "manual" drivers will be paying out more often and will adjust the insurance rates to compensate.

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#569

Post by Dylan16807 » Fri, Apr 06, 2018, 8:18 AM UTC

mdorazio wrote: post #564

Dylan16807 wrote: post #560

Earlier quoted context omitted.

> Subtracting your own velocity from what? You said the camera/radar would be reading another car as "stationary", which is a velocity number of zero. Add/subtract your speed and you get the real speed. A non-defective sensor suite is going to give you either distance or velocity, and you can use distance measurements to calculate velocity really easily. > A radar return from a rear bumper at relative zero miles an h…

Maybe I'm not being clear on my core argument. sensors give conflicting information in different scenarios . One sensor might read an obstacle to be avoided while another shows a clear path. Deciding what to do in these conflict situations is difficult. Saying it's easy is effectively saying that the hundreds of very intelligent engineers working on these systems are idiots because it looks like it should be a piece…

> Saying it's easy is effectively saying that the hundreds of very intelligent engineers working on these systems are idiots because it looks like it should be a piece of cake.

Saying one very very specific thing is easy is not the same as calling engineers idiots. It's bad to conflate "the system is hard" with "every single piece of the system is hard"

> The driving system has to combine this with other sources of information for it to be useful, and that's where problems creep in.

There are many places where integrating information can let errors creep in.

The car adding its own ground velocity in is just... not one of them.

> What I'm trying to say is that there are a number of fuzzy steps to get from sensor readings to actual object detected, and then from actual object detected to "known obstacle" classification, as you put it.

Agreed. But while some steps are fuzzy, some are easy. The chain from start to finish is fragile and difficult. But some of the individual links are rock-solid.

> I'm going to add this to my longer article writing list. Thank you for the constructive debate on this.

I'm not sure if we really got anywhere productive here, but good luck!

Re: Tesla crash in September showed similarities to fatal Mountain View accident

#570

Post by Piskvorrr » Fri, Apr 06, 2018, 9:24 AM UTC

ceejayoz wrote: post #562

Piskvorrr wrote: post #543

Earlier quoted context omitted.

That last part is actually scariest: blame the victim, yay. Unless they can prove they had a functional, powered, up to date, compatible I'm-a-car responder, it's their own damn fault for being invisible to the SDV. How about "I'm a human, don't hit me"? Does that also sound like a good idea (RIP Elaine Herzberg)? In other words, "all other road users should accommodate my needs just to make life a bit easier for me"…

I think that's a needlessly uncharitable interpretation. Having cars communicate information to each other has the potential to be an additional safety measure. It's like adding reflectors to bikes - adding them wasn't victim blaming, it was just an additional thing that could be added to reduce accidents.

Sure, I understand that you're proposing it as an improvement , and it would even be an improvement - but using this for scapegoating will happen, as long as there are multiple parties to any accident; we have seen this in the last Uber crash ("find anything pointing anywhere but at Uber"), or in any bike crash ("yeah, the truck has smashed into him at 60 MPH and spread him over two city blocks, but he's at fault for not wearing a helmet - it would have saved him!!!").