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

Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

fortune.com

Post reply on HN
Print view View original

351–360 of 459 posts

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#351

Post by emodendroket »

BillTheBiker wrote: post #237
emodendroket wrote: post #99

Earlier quoted context omitted.

I don't think that's true at all. There was an article posted about the subject a while ago, but the thrust of it was that researchers think they need to do more to make cars behave like (fully attentive) human drivers rather than being weird anomalies that behave in surprising/irritating (to humans) ways if they're going to actually go on the road, and I think that's completely right.

"(fully attentive) human drivers" probably cause deadly accidents every day, and they certainly behave in surprising/irritating ways. Anyways, a human driver going the speed limit is no hazard, so I fail to see how having self-driving cars follow the limits could be a problem.

Autonomous cars need to share the road with human drivers, so they should not behave in a way that is surprising to humans, to the extent possible.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#352

Post by amelius »

brokenmachine wrote: post #227
mattigames wrote: post #93

Earlier quoted context omitted.

Yeah there are a lot of subtleties that are hard to program; that's why it must be extremely effective on things that _can_ be programmed such as quick breaking, quick steering; prefer lanes that are farther away from pedestrians near-by, beep when entering a non-illuminated zone and hundreds of other details.

Quick braking is the one thing you would have thought they could get right for sure, and they didn't manage even that on this occasion.

Quick braking can also be dangerous. Perhaps even LiDAR has to deal with false positives every now and then.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#353

Post by paganel »

benwilber0 wrote: post #33
friedman23 wrote: post #21

Earlier quoted context omitted.

3mph above the speed limit

That is speeding and also the car was inattentive to potential hazards outside of its vision. Normal human drivers slow down when they see a wild lady on the median looking like she might dart into the road. Normal people slow down when there are kids playing with a ball on the sidewalk. Do these cars do that? Do they even drive the speed limit? edit: Also accidentally killing someone with your car because of minor s…

> Normal people slow down when there are kids playing with a ball on the sidewalk.

The driving-exam manual I had to read before getting my license was mentioning this exact scenario, i.e. if one sees kids playing on the sidewalk close to the road that the driver should slow down. It was even included among the questions given at the written exam itself.

But, then again, the authorities over here in Europe are more attentive when handing out driving licenses to people who are about to handle 2-ton pieces of metal inside populated areas. In this case it seems like the "US driving mentality" (only the driver counts, damn be the pedestrians) has also affected the actual engineers who have programmed these "AI" cars.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#354

Post by huhtenberg »

cthalupa wrote: post #344
r_klancer wrote: post #225

My takeaway from the comments in this thread: self-driving Ubers don't drive defensively, almost surely can't "read" pedestrian behavior, and even with lidar may never have seen Elaine Herzberg. Oh, and they speed. This sounds like straight out of the "hell" scenario for driverless cars: "we cannot accept responsibility for what happens to pedestrians and cyclists foolish enough to stray out of their designated safe…

>Oh, and they speed. 3mph variance is perfectly normal even on modern cruise control. And even on most modern cards, what you see on the speedometer is often off by ~1 mph at low speeds, and even more at higher speeds. (This is the reason you will often see people talk about 'GPS verified' speeds when it comes to talking about the top speed they have achieved in a car). Federal law in the US only requires they be acc…

> 3mph variance is perfectly normal even on modern cruise control.

That'd be a variance between true ground speed and the speed as estimated by the car. Chances are that the 38mph number came from the car itself, so there should be no "variance", leave alone 3mph over the limit. Modern cruise controls are perfectly capable of maintaining set speed, even at a considerable downhill.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#355

Post by zepn »

benwilber0 wrote: post #14

> According to the Chronicle, the preliminary investigation found the Uber car was driving at 38 mph in a 35 mph zone and did not attempt to brake. Herzberg is said to have abruptly walked from a center median into a lane with traffic. The autonomous vehicle was speeding and made no attempt to decelerate given the sudden presence of an obstruction in its path? I would expect some kind of reaction from the car within…

> given the sudden presence of an obstruction in its path

It doesn't even need to have been in its path; it should have detected a pedestrian near the edge of the road, concluded that they might step into the road (it's a possibility), and decelerated. Before she even stepped into the road, the Uber was at fault.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#356

Post by ajb »

invisible wrote: post #279
hnaccy wrote: post #245

Earlier quoted context omitted.

>Seeing as the cars have been on the road in Arizona for a while, and this is the first we are hearing about a pedestrian collision, it seems the algorithm is pretty safe. One fatality in 4 million miles is safe? Humans are 20 times better than that.

I wonder what the margin of error is on that one in 4 million rate since the sample size is one. Also of note: > Arizona has experienced a surge of pedestrian fatalities recently, with more than 10 in a single week of March in Phoenix alone. The state has the highest rate of pedestrian fatalities in the United States.

The sample size is not one. The sample size is 4 million. The 5% and 95% quantiles of the posterior are roughly one in 11 million and one in 800,000.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#357

Post by playworker »

cthalupa wrote: post #344
r_klancer wrote: post #225

My takeaway from the comments in this thread: self-driving Ubers don't drive defensively, almost surely can't "read" pedestrian behavior, and even with lidar may never have seen Elaine Herzberg. Oh, and they speed. This sounds like straight out of the "hell" scenario for driverless cars: "we cannot accept responsibility for what happens to pedestrians and cyclists foolish enough to stray out of their designated safe…

>Oh, and they speed. 3mph variance is perfectly normal even on modern cruise control. And even on most modern cards, what you see on the speedometer is often off by ~1 mph at low speeds, and even more at higher speeds. (This is the reason you will often see people talk about 'GPS verified' speeds when it comes to talking about the top speed they have achieved in a car). Federal law in the US only requires they be acc…

I think there's something off about your arguments, human drivers speed, but they don't do it because of properties of the speedometer or the cruise control, or air resistance, or potholes - they do it because they don't care much about the speed limit.

It's completely possible to stay below the speed limit in a car, a lot of people just don't bother. Cruise control won't apply the brakes (at least in older cars) if you encounter a significant downhill gradient - that's up to the driver, and while speedometers are often out and GPS is more accurate, speedos universally over-read in my experience. If you're doing GPS 38mph your speedo is probably reading 40, which is not really within an excusable error margin of 35mph IMHO.

I would expect a fully autonomous vehicle to be able to accurately obey the speed limit and I suspect the reason it wasn't/doesn't is because that behaviour has been programmed in or developed. You could make arguments that doing what the other cars around you are doing is safer than being an outlier, but that's not the same argument you're making. I think the 38mph figure is significant.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#358

Post by wazoox »

stefan_ wrote: post #10

This is just police standard operating procedure. In any kind of manslaughter where a vehicle is involved, save on paperwork and blame whichever party ended up dead before the driver has even spoken the words "she came from nowhere". Here is the police chief spelling it out for Uber: she came from the shadows I wonder what that LIDAR thinks about the shadows . I mean, this system is in place because everybody drives…

In France, the car is always at blame whatever the cyclist or pedestrian did. This the jurisprudence as set by a case where a drunk cyclist, driving at night without any light in the wrong lane and against traffic was hit by a car coming by. Though the driver obviously wasn't at fault, the decision was to always protect the vulnerable and feeble, i.e. the cyclist.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#359

Post by lmm »

ShorsHammer wrote: post #314
r_klancer wrote: post #225

My takeaway from the comments in this thread: self-driving Ubers don't drive defensively, almost surely can't "read" pedestrian behavior, and even with lidar may never have seen Elaine Herzberg. Oh, and they speed. This sounds like straight out of the "hell" scenario for driverless cars: "we cannot accept responsibility for what happens to pedestrians and cyclists foolish enough to stray out of their designated safe…

I don't see why autonomous vehicles are so big. Virtually all travel is one person, ideally it would be a tiny 4 wheel electric with the sole passenger highly reclined. The hangover from internal combustion is real.

What does autonomy have to do with it? Most regular cars would be better off as single-person vehicles; there are a few small-scale efforts in this direction (the Smart car, the G-Whiz electric car, I think BMW has an enclosed-canopy motorbike) but they largely haven't caught on.

Re: Police Say Uber Is Likely Not at Fault for Self-Driving Car Fatality in Arizona

#360

Post by ygra »

__david__ wrote: post #235
protomyth wrote: post #178

Earlier quoted context omitted.

Wouldn’t that be mitigated by the GPS and the map? It would seem that an automated car using maps as an important input would no its exact position and therefore speed.

GPS can be remarkably bad in urban environments. You can't really trust it for second-to-second speed or positioning. It's only really good for long term/high level planning. For speed, the wheel sensors are going to be way more reliable and consistent.

Actually, with a good receiver you can use GPS for very precise speed measurements by using the doppler shift of the carrier signal. It's just that most receivers don't give you access to that so the only other option is using the current and previous position and the timings of them.
Post reply on HN