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

601–610 of 714 posts

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

#601

Post by MBCook »

abigail95 wrote: post #410
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.

Firstly, human error in at least 7 out of the 8 cars caused this. FSD didn't take control away from the human, it encountered something outside of it's limits and returned full control to the human. Secondly, that's what planes are. A lot of new plane safety comes from very automated, very complicated systems. In some situations these planes fail in a way that leaves the pilot in a more dangerous plane than one with…

> FSD didn't take control away from the human, it encountered something outside of it's limits and returned full control to the human.

We don’t know if that’s what happened.

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

#602

Post by xyzzyz »

brewdad wrote: post #340
xyzzyz wrote: post #307

Earlier quoted context omitted.

It costs Amtrak over $300k a year to maintain a single mile of track on the Northeast corridor. For comparison, it costs around $30k to maintain a single mile of US interstate, i.e. only 10% of the cost to maintain rail track. In what sense rail is cheaper to maintain, exactly?

Citation needed. I can believe that we spend $30k a year filling potholes on a mile of interstate but every 10 years we have to resurface it and every 30-50 years we have to completely rebuild it.

Just Google it, it’s not hard. Are you really surprised that rail also costs money to maintain?

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

#604

Post by joenathanone »

maury91 wrote: post #492

I'm trying to reason "why the Tesla did stop?", in the second video on TheIntercept website, I can notice a left exit before the point where the Tesla stopped. Something I noticed with Google maps ( I don't have a Tesla, and maybe the navigation system of the Tesla is similar ), is that sometimes thinks tunnels can access the road above them ( in a tunnel near my home Google maps always tells me "turn right" in the m…

I have an X and a 3, this is exactly how the car behaves when it has missed a turn because of a mismatch of map/vision data.

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

#605

Post by matthewdgreen »

jackmott wrote: post #522
maury91 wrote: post #492

I'm trying to reason "why the Tesla did stop?", in the second video on TheIntercept website, I can notice a left exit before the point where the Tesla stopped. Something I noticed with Google maps ( I don't have a Tesla, and maybe the navigation system of the Tesla is similar ), is that sometimes thinks tunnels can access the road above them ( in a tunnel near my home Google maps always tells me "turn right" in the m…

I have a Tesla, when using just cruise control it has slammed the brakes: because of shadows that confused it because of concern about cars crossing the intersection ahead where there is only the need to gently slow because of a car crossing an intersection ahead that has already crossed! because of a steep bridge that confused it we can’t use cruise control in the car basically, we decided it is too dangerous at wor…

The worst part of phantom braking is not just the need for a sudden human intervention. It’s that the required human intervention is extremely non-intuitive.

Intuitively when a car suddenly slows down it feels like the right response is to press on the accelerator: however in my experience this puts you in an unstable regime where you are “fighting” with the car. That is, autopilot continues to try to stop but temporarily accepts your override, but only as long as you maintain consistent pressure on the accelerator. Unfortunately if you even briefly let up on the accelerator, the car proceeds to violently slam on the brakes again. This can lead to a feedback loop where the car’s rapid acceleration/deceleration pattern makes it difficult to maintain consistent pressure on the pedal, and so the experience is like being in a rodeo. Worse, there’s no obvious way “out” of this cycle except to take your foot off the accelerator and let the car (briefly) win.

Counterintuitively, the “correct” way to deal with phantom braking is to avoid the accelerator entirely and instead dive right for the brake pedal: this instantly disengages cruise control. But this is not an intuitive response, you have to learn it the hard way.

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

#606

Post by ryanwaggoner »

AmericanChopper wrote: post #569
sumeno wrote: post #556

Earlier quoted context omitted.

Changing lanes and then slamming on the brakes means you are 100% at fault (assuming there is evidence). It's the most common car insurance scam and it is completely indistinguishable from what the Tesla did in this case

Breaking down is also completely indistinguishable from what the Tesla did in this case. The breaks weren’t slammed on either, 7 seconds elapsed between the Tesla indicating for the turn and the collision. It’s very obvious that every participant in the pile up was at least partially at fault. The only confounding factor in this case seems to be that many HN commenters lose a handful of IQ points when any topic relat…

I applaud your self awareness

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

#607

Post by magicalist »

throwaway894345 wrote: post #598
brookst wrote: post #586

Earlier quoted context omitted.

Not the person you’re responding to, but typically this mindset is about punishment more than deterrence. It won’t reduce the incidence of bad outcomes, but some people feel better knowing that judicial vengeance will be meted out.

Fair enough, but do those people hold the same attitude about punishing violent criminals? Seems like there’s a lot of overlap between the “we shouldn’t punish violent offenders” and the “reckless drivers should be punished” people, but maybe I’m misreading the situation.

Yes, there's still a categorical difference between "there should be some consequences for recklessly killing someone" and "jailing people forever is the only right response to a violent crime".

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

#608

Post by frumper »

smaudet wrote: post #579
jiriro wrote: post #392

Earlier quoted context omitted.

5 seconds is nonsense. The average driver’s reaction time is 0.2 to 0.3 seconds. This is before the driver’s foot moves. Then there is approx 1 second before the brakes take effect (foot movement, applying the force, brakes responding) So 1.5 second before the car starts breaking. There is a rule of “be at least 2 seconds behind the car in front” which gives a safe distance to handle any emergency braking. Of course…

It's absolutely not nonsense, I suggest taking some basic high-school physics. Reaction time is the time for the vehicle to come to a stop, not for the driver to first notice an issue. At 60 mph it will take many meters and seconds to deaccelerate, assuming the driver can take exactly the correct action: https://www.wolframalpha.com/input?i=time+to+brake+at+70+met... Estimate here is about 4 seconds, under perfect co…

he said 5 seconds to START braking

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

#609

Post by threeseed »

darknavi wrote: post #288
threeseed wrote: post #60

Earlier quoted context omitted.

This is a bug in Tesla's specific meat. Other self driving implementations e.g. Cruise, Waymo have LiDAR which is an additional and in my opinion necessary sensor for detecting objects. Because in this case it sure looks like a phantom breaking issue.

Is it really a Tesla specific issue? Tesla's definitely phantom brake quite a bit, but it seems like many automated cruise control implementations do this today. https://www.motorsafety.org/us-government-investigates-1-7-m... https://www.macheforum.com/site/threads/phantom-braking-expe... https://www.chevybolt.org/threads/ok-lets-talk-phantom-braki...

It is Tesla specific in the sense that they refuse to use LiDAR.

All those examples just reinforce my point that it's a necessary sensor.

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

#610

Post by sidibe »

rini17 wrote: post #425
abigail95 wrote: post #410

Earlier quoted context omitted.

Firstly, human error in at least 7 out of the 8 cars caused this. FSD didn't take control away from the human, it encountered something outside of it's limits and returned full control to the human. Secondly, that's what planes are. A lot of new plane safety comes from very automated, very complicated systems. In some situations these planes fail in a way that leaves the pilot in a more dangerous plane than one with…

So you categorize the 737MAX accidents as human errors too and thus they should not affect policy?

That's a very different situation. Namely, I don't own Boeing stock.
Post reply on HN