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

341–350 of 714 posts

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

#341

Post by modeless »

Hamuko wrote: post #323
modeless wrote: post #286

Earlier quoted context omitted.

Yes, FSD is essentially a complete rewrite of Autopilot which will soon replace it for freeway driving too. Autopilot has been available for many years and FSD is new. This accident could have happened years ago and is not evidence of any issue with the new FSD stuff.

But since there's no behavioural difference, it doesn't matter if it's a complete rewrite or not.

They currently operate on disjoint sets of roads, so they can only be roughly compared, and in rough comparison they do not behave differently. Of course if operated on the exact same road (which, again, is currently impossible) their actions would not be the same in every detail, just as different humans or even the same human driving the same road multiple times wouldn't. But that doesn't mean that there's necessarily any consistent difference that would enable a driver to use the information of which one is active to improve their predictions of the car's behavior.

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

#342

Post by watwut »

modeless wrote: post #325
ncallaway wrote: post #313

Earlier quoted context omitted.

> There is no reason the driver needs to be aware of it. If the person operating the vehicle doesn't need to be aware of the difference, then the general public shouldn't need to either when analyzing an accident. It seems baffling to me that we could imagine a scenario where it's less important for the driver of a vehicle to be aware of the difference, than it is for the public when reacting to a news article. I don…

The difference comes when drawing inferences about the future development of Autopilot/FSD as these articles invariably do. This is completely different than day-to-day usage of the system. If you believe this was a failure of FSD, then maybe it's evidence we should stop the rollout of FSD or whatever. If it's a failure of Autopilot, then it is not evidence about the future behavior of FSD at all. FSD is a complete r…

If FSD automatically turns into autopilot on higways and driver is not supposed to know, then it is fault of SFD.

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

#343

Post by AmericanChopper »

justapassenger wrote: post #243
somethoughts wrote: post #65

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. [Edit] From the video the Tesla is making a lane change and stopping simultaneously which means there could be a case of the Tesla FSD/driver doing a unsafe lane change.[1] Most of the time FSD just wrecks the Tesla itself or injures the driver of the Tesla (i.e. running int…

It’s very unlikely it’s a first time FSD caused accident for others. People using FSD allow it to do reckless things, just to see “if it’ll figure it out”. You can cause accident, without being involved in it, especially if you drive in unpredictable way. Like this one - if driver would stop sudden breaking and moved forward, cars behind him could have still crashed. That’s the thing about testing on the public roads…

Most of the fault in this video seems attributable to the other motorists. I was always taught that it was my responsibility to ensure that I have enough following distance, and am paying close enough attention, to stop if the car in front of you abruptly stops. Is this not the case in California? Because the first car seems at least partially at fault in this respect, and every other contributor to the pile up is significantly at fault.

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

#344

Post by somedude895 »

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…

There are so many tunnels in Switzerland, that using AP is a huge pain. But at least it keeps me attentive and ready to take over at all times.

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

#345

Post by watwut »

Alex3917 wrote: post #84
somethoughts wrote: post #65

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. [Edit] From the video the Tesla is making a lane change and stopping simultaneously which means there could be a case of the Tesla FSD/driver doing a unsafe lane change.[1] Most of the time FSD just wrecks the Tesla itself or injures the driver of the Tesla (i.e. running int…

> 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 you change the lanes to be right in front of the car and then hit brakes, you are at fault. The back driver responsibility applies when they have control over space between you two.

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

#346

Post by frutiger »

masklinn wrote: post #311
soraminazuki wrote: post #225

Earlier quoted context omitted.

Reminds me of how I used to build railways everywhere instead of roads in Sim City. But surely that wouldn't work out in real life?

It works in most of the world’s cities.

Statistically speaking, the vast majority of the world’s cities have absolutely zero local rail transportation.

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

#347

Post by userbinator »

choeger wrote: post #330

Everyone claiming that the autopilot is at fault for stopping suddenly get something very important very wrong: Stopping a car must be possible at all times. Any automation in a car should always be able to stop the car as the safe default. The same holds for human drivers, of course: Consider the possibility that the vehicle in front of the Tesla just lost some of its load. Of course the Tesla then has to stop sudde…

The same holds for human drivers, of course: Consider the possibility that the vehicle in front of the Tesla just lost some of its load.

The difference is that the humans in the cars behind will be able to see that too and anticipate the reason for the braking. In this case, the Tesla just changes into the leftmost lane and then brakes very quickly to a stop for no apparent reason. A human doing the same thing would almost certainly be at fault too.

You can see on the video that it's not just suddenly stopping, it's changing lanes (with what appears to be very little room from the car behind) and then suddenly stoppping.

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

#348

Post by brewdad »

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 have to do the drive from Portland to Seattle about once a month. ACC makes the drive so much more enjoyable. I can pretty much zone out until it is time to pass someone. Then, depending on speed differences, I move over and pass without intervention or I press the throttle a bit to pass more quickly.

When I get to the inevitable slowdowns, my car just slows without me having to think much about it unless it’s a sudden one.

I also have a Golf with standard cruise and I can’t stand using it since most drivers around me can’t maintain a constant speed, so I am always adjusting it or canceling it.

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

#349

Post by watwut »

thaumasiotes wrote: post #6
loopdoend wrote: post #5

Earlier quoted context omitted.

Don’t you just tap the brake to disengage adaptive cruise?

The problem isn't that disengaging the system is difficult. The problem is that deciding whether to disengage is difficult. If you're catching up to a slower car, maybe you should disengage the cruise control. Or maybe you shouldn't. You can't tell. This means that cruise control goes from being something that relieves stress on you to something that causes additional stress.

If you are catching slower car, you dont have to disengage. That being said, I simply disengage whenever I feel like, whenever I feel any doubt. It is easy to turn on again.

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

#350

Post by skhr0680 »

modeless wrote: post #197
actionfromafar wrote: post #174

Earlier quoted context omitted.

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

Automatic switching between FSD and Autopilot is not any kind of UX problem. There is no reason the driver needs to be aware of it. The driver does not need to take any action at all. The only difference is the type of road. But which software was in use is relevant here because people are trying to blame this on FSD. Whenever you're on a freeway, Autopilot is used automatically instead of FSD, so FSD had nothing to…

I’m not familiar with Teslas but to me the UX problem is that Autopilot and FSD sound like they should mean the same thing
Post reply on HN