Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

New video of Tesla crash demonstrates the problem of semi-automated driving

theautopian.com

Post reply on HN
Print view View original

171–180 of 714 posts

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#171

Post by MBCook »

epivosism wrote: post #128
MBCook wrote: post #117

Earlier quoted context omitted.

So if an automaker can make a system that is capable of handling any driving task safely but occasionally ran itself off the road on purpose for no reason, as long as it’s a net positive in lives saved it should be allowed? I don’t agree with that assessment. It needs to make fewer mistakes than humans at all times , not just most of the time. Great with a disaster here and there is terrible.

I see where you're going, but how do you actually justify your choices? A: occasional random suicide (100 deaths/year), no other deaths B: no random suicide, 10k deaths per year from normal driving You would be saying you choose B? It's not really about the exact numbers here - just trying to get on the same page - would you really weigh your uncomfortableness with "random deaths" highly enough that you would accept…

This is a human comfort thing, not a reason thing. As someone else in this discussion said if computers do something that looks stupid to people, they won’t trust it.

What if the investigation here showed the Tesla was 100% correct to do what it did?

What it did is so odd I wouldn’t want to trust it. Maybe the problem is it needs to fail in a more human-like way?

If the death numbers get low enough, I think it will likely speak for itself and people will trust it. But it can’t be a 30% reduction in deaths. It needs to be like 95%.

Enough that it feels safer instead of questionable/dangerous. Even if it was actually safer the entire time.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#172

Post by pensatoio »

greenthrow wrote: post #167
pensatoio wrote: post #158

If autopilot starts braking even slightly and it shouldn't, I immediately take control. For this to happen, the driver clearly was not prepared to take control in any way. It's like driving on cruise control with your legs crossed. The automated cruise control and lane-keep on my previous car, an Audi, would occasionally do crazy things too. The hyper-focus on this being Tesla's fault continues to baffle.

Tesla has much more of an issue with phantom braking than any other manufacturer because they chose to rely only on cameras, and that system simply does not work as well when it comes to this particular issue. Additionally, how can you blame the driver? Did you watch the video? It's not as if the car slowly stopped. It slammed on the brakes. There wasn't time for the driver to correct for this before the car behind -…

Driving in traffic (or in a city in general) with my foot off the pedals sounds borderline insane. I know this is anecdotal, but I've experienced multiple brands slamming on the brakes. You must be prepared to intervene.

It's not a Tesla problem. I think we hyper-focus on Tesla because their system is the best. This is a problem for all automated driving systems (even the basic ones found in almost all brands now.)

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#173

Post by pensatoio »

whatever1 wrote: post #161
pensatoio wrote: post #158

If autopilot starts braking even slightly and it shouldn't, I immediately take control. For this to happen, the driver clearly was not prepared to take control in any way. It's like driving on cruise control with your legs crossed. The automated cruise control and lane-keep on my previous car, an Audi, would occasionally do crazy things too. The hyper-focus on this being Tesla's fault continues to baffle.

Well it’s the only company that charges 20k for “full self driving capability”

Since it's a highway, the car would have been operating on autopilot, not FSD. The two stacks had not, and I think still have not been merged.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#174

Post by actionfromafar »

gfodor wrote: post #57
gundmc wrote: post #44

Earlier quoted context omitted.

I believe the car was in "Full Self Driving" and not "Autopilot"?

It’s unclear because FSD is supposed to switch to autopilot on highways. The driver said it was in FSD mode but they may just not realize that switchover occurs.

And here is the whole problem. It’s not good UX.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#175

Post by scotty79 »

sokoloff wrote: post #107
Alex3917 wrote: post #84

Earlier quoted context omitted.

> I think what's interesting here is it's likely the first instance where Tesla FSD has been involved in an accident which affected other drivers. In a pileup like this it's basically never the fault of the front car, unless maybe if they are purposely causing the accident for insurance fraud or something. Maybe the driver will get cited for failing to maintain the minimum speed, but legally this isn't much different…

If I brake check someone on the highway, causing an accident, and it’s discovered that that’s what happened, I’d be pretty shocked* if the trailing driver was found responsible. That video looks like a combined lane change and brake check on the part of the Tesla. * and disappointed

I think the intention matters. If you did it to spite the guy behind you then you are at fault. If you did it because you thought something might cross the road and just wanted to be safe then it's not your fault even if nothing crossed the road or nothing was really there.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#176

Post by pensatoio »

wyldfire wrote: post #164
pensatoio wrote: post #158

If autopilot starts braking even slightly and it shouldn't, I immediately take control. For this to happen, the driver clearly was not prepared to take control in any way. It's like driving on cruise control with your legs crossed. The automated cruise control and lane-keep on my previous car, an Audi, would occasionally do crazy things too. The hyper-focus on this being Tesla's fault continues to baffle.

> autopilot starts braking even slightly and it shouldn't, I immediately take control I think this is the thing that makes the most sense. But if I were to bend over backwards giving the benefit of the doubt - is it possible the driver feared resuming the previous speed because they were worried the car saw an obstacle that they couldn't? Like the article suggests - if you're not fully engaged, it's as if you're driv…

This is anecdotal, but I'm always looking at the road. If the car starts braking, (and this isn't exclusive to Tesla,) I already know if there is an obstacle or not, and my foot is always hovering over the pedals.

I feel like such a shill responding to these comments, but based on my experience, these same drivers would get into accidents in other brand's cars as well. We only focus on the story because it's Tesla.

I'm not sure what the solution is.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#177

Post by scotty79 »

denimnerd42 wrote: post #23
herodotus wrote: post #2

I have a VW Golf and it has what it calls "adaptive cruise control". What makes it different from my older (simple) cruise control is that it will slow down and speed up again as necessary without my intervention. For example, if I set the speed to 70 mph and it gets close to a car ahead that is only going 65 mph, it will slow down and maintain what the software believes is a safe distance. Similarly, if a car in an…

I do love the ACC on my car. One of the worst parts of road trips is following cars you don't want to or can't pass. The ACC totally solves it. I don't see how you were really able to use cruise to follow another car at all before? Even 1mph too slow is not great and if the car in front isn't using any cruise control at all it could be 5mph too slow and then 5mph too fast.

If you don't want to pass a car you just drop down the speed on cruise control and widen the gap between the car you don't want to pass and you. It really doesn't matter if you are 10m or 100m behind him if you are not going to pass it.

And if you get too far behind him you might bump up the speed a notch.

And if somebody gets in the gap it's even better because they are going to eventually overtake him just like they overtook you and overtaking one car at a time might be safer than two cars at once.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#178

Post by sokoloff »

roguecoder wrote: post #140

Video understanding engineer here. Tunnels and underpasses are the worst. They are a pain in the ass, because shadows mess with all the edge detection and motion models and anything else visual. Humans compensate by thinking "I'm in a tunnel: things are weird." But without a reasoning model that can take context into account, the computer is stuck. In the video from behind, you can see the shadow ahead of the car on…

Interesting observation.

I worked on autonomous vehicles (in vision) at Daimler in 1991. During one of our test sessions, on drying pavement, the vehicle abruptly slammed on the brakes and refused to proceed past a point where the vision system could see a symmetric about a centerline set of horizontal edges on the pavement. It tightly fit our (hand-coded, being 1991) model for a car ahead. We had to revert to manual control and drive back to our staging area and wait for the track (set of runways and taxiways at a disused airbase) to finish drying.

Obviously, the state of the art has significantly improved since then, but some fundamental risk of misinterpretations could easily remain.

Re: New video of Tesla crash demonstrates the problem of semi-automated driving

#180

Post by abduhl »

scotty79 wrote: post #175
sokoloff wrote: post #107

Earlier quoted context omitted.

If I brake check someone on the highway, causing an accident, and it’s discovered that that’s what happened, I’d be pretty shocked* if the trailing driver was found responsible. That video looks like a combined lane change and brake check on the part of the Tesla. * and disappointed

I think the intention matters. If you did it to spite the guy behind you then you are at fault. If you did it because you thought something might cross the road and just wanted to be safe then it's not your fault even if nothing crossed the road or nothing was really there.

Tesla’s FSD/autopilot does not act with intention as we understand it.
Post reply on HN