Uber vehicle reportedly saw but ignored woman it struck • Hacker News Board

Re: Uber vehicle reportedly saw but ignored woman it struck

#301

Post by ufmace » Tue, May 08, 2018, 4:35 AM UTC

erobbins wrote: post #215

ars wrote: post #15

Earlier quoted context omitted.

> get 5 9s reliability That's it? Humans have greater than 7 9's reliability when driving. (As measured by fatalities per time of driving, and assuming a fatality takes 5 minutes, and average speed is 30mph.) If the cause of a fatality is 10 seconds of failure, not 5 minutes, humans have greater than 9 9s of reliability. Math: 10 seconds / (1.13 / (30 miles/hour / 100 million miles)) Fatality rate is 1.13 per 100 mil…

> There's this belief that humans are horrible at driving, etc, etc. Until you run that math and realize, no, they're not. There's just a lot of driving going on. I've been hammering this concept for a while :) Unimpaired, non distracted humans are actually really good at driving. We're optimized for processing visual inputs and making fuzzy decisions, have high dexterity and coordination, and fear death. All of that…

Oddly enough, even impaired drivers are probably pretty good compared to some current self-driving cars. I'd give good odds that someone with a 0.10% BAC would have swerved or stopped for that woman that the Uber car hit.

Re: Uber vehicle reportedly saw but ignored woman it struck

#302

Post by c22 » Tue, May 08, 2018, 4:36 AM UTC

eric_h wrote: post #276

jschwartzi wrote: post #177

Earlier quoted context omitted.

Gaze is actually more important when you're trying to figure out whether someone sees you. If I don't see your eyes, I would assume that you don't see me at all.

> If I don't see your eyes I think you really mean if you don't make eye contact. Eye contact is a two-person activity. How does an autonomous car make eye contact with a person?

Something like this could be developed. For instance, the car could have an array of lights on its grill (think Knight Rider) that track in the direction of detected obstacles, if you don't see a light facing towards you then you know you are not detected and to use caution. Alternately a light on the front of the car could turn from yellow to blue to indicate its intent to yield.

Re: Uber vehicle reportedly saw but ignored woman it struck

#303

Post by c22 » Tue, May 08, 2018, 4:49 AM UTC

Sangermaine wrote: post #81

dahart wrote: post #75

Earlier quoted context omitted.

Personally, I’d rather see driving behavior normalize nationally than cater to regional differences in enforcement. The driving laws in each city & state the US are less different than the locally accepted behaviors, but it would be safer all around if drivers and pedestrians all expected the same things, and if we took traffic & safety laws more seriously. More enforcement of existing traffic laws will save many ord…

The "reality should conform to my ideal, not my ideal to the reality" approach is and has always been a guaranteed recipe for failure. "Let's change everyone's minds and behavior everywhere" is much harder than changing software.

A massive enough influx of self-driving cars all driving with the same profile is perhaps one of the few ways reality could be coaxed into this particular ideal.

Re: Uber vehicle reportedly saw but ignored woman it struck

#304

Post by c22 » Tue, May 08, 2018, 4:55 AM UTC

userbinator wrote: post #263

jarfil wrote: post #189

Earlier quoted context omitted.

Or... make Jay walking punishable by death, and autonomous cars won't need to stop, or even look for random pedestrians.

This, absolutely this. It's always bothered me why pedestrians seem to be universally given the right of way, when the irrefutable laws of physics dictate otherwise: a human, although having a lower speed, has a far wider view of its surroundings, and can change direction much quicker than a car whose driver is relatively slow to react and has a much more restricted view. In a collision, the bare human is always goin…

Because the train is on tracks.

Re: Uber vehicle reportedly saw but ignored woman it struck

#305

Post by ars » Tue, May 08, 2018, 5:14 AM UTC

jstanley wrote: post #25

ars wrote: post #15

Earlier quoted context omitted.

> get 5 9s reliability That's it? Humans have greater than 7 9's reliability when driving. (As measured by fatalities per time of driving, and assuming a fatality takes 5 minutes, and average speed is 30mph.) If the cause of a fatality is 10 seconds of failure, not 5 minutes, humans have greater than 9 9s of reliability. Math: 10 seconds / (1.13 / (30 miles/hour / 100 million miles)) Fatality rate is 1.13 per 100 mil…

It's only really 9 9s of reliability if you assume somebody will definitely be killed as soon as you have a 10-second lapse in concentration, which is obviously not true. In motorway driving you could probably shut your eyes for 10 seconds every 20 seconds and be perfectly fine almost all of the time. Especially at night time, or other quiet times of day.

If people actually did that, then someone would crash, and that crash would be factored into the final accident rate.

So that final reliability rates includes any and all things humans do while driving. (Not what they could do, what they actually do.)

Re: Uber vehicle reportedly saw but ignored woman it struck

#306

Post by docker_up » Tue, May 08, 2018, 5:15 AM UTC

davidgould wrote: post #167

hazardmat wrote: post #137

Earlier quoted context omitted.

That's not really true. Thousands of deer get killed on the road every year, because deer aren't smart enough not to be an obstacle when it's dark. Do we therefore change all the speed limits at night? No. It's a driving risk taken, just like any other.

Those deer are killed because thousands of humans aren't smart enough to not overdrive their headlights so they hit obstacles in the dark. Hitting a deer is not a small thing. You will seriously damage your car, and may be severely injured or killed yourself if the deer is tall enough to come through the windshield. In open range country you can also hit a cow. In some areas you can hit a moose. Do we therefore chang…

I've driven in Palo Alto near 280, and from the distance, in broad daylight, I saw a buck run across an entire field at full speed and straight across the road about 3-4 cars in front of me, and get hit by the car. There was no way the driver could see what I saw, and he had no chance to avoid it. The deer was simply acting irrationally. To say that people who hit deer are "driving like idiots" is wrong.

Re: Uber vehicle reportedly saw but ignored woman it struck

#307

Post by ars » Tue, May 08, 2018, 5:16 AM UTC

dllthomas wrote: post #281

hsljekskfh wrote: post #184

Earlier quoted context omitted.

freeway driving causes less accidents than street driving

I think you were downvoted to dead because people saw this as an objection to the parent; more charitably, I assume you're just adding more color - but you might want to telegraph your intentions a bit clearer.

He was not downvoted to dead, his account is banned, all his posts show up dead.

Re: Uber vehicle reportedly saw but ignored woman it struck

#308

Post by ars » Tue, May 08, 2018, 5:24 AM UTC

blattimwind wrote: post #37

ars wrote: post #15

Earlier quoted context omitted.

> get 5 9s reliability That's it? Humans have greater than 7 9's reliability when driving. (As measured by fatalities per time of driving, and assuming a fatality takes 5 minutes, and average speed is 30mph.) If the cause of a fatality is 10 seconds of failure, not 5 minutes, humans have greater than 9 9s of reliability. Math: 10 seconds / (1.13 / (30 miles/hour / 100 million miles)) Fatality rate is 1.13 per 100 mil…

> If the cause of a fatality is 10 seconds of failure, not 5 minutes, humans have greater than 9 9s of reliability. If you're driving by the textbook , then you have two seconds to react to the car in front braking; the assumption being that you take one second to react normally, and get one extra second for safety. On the other hand, in something like a play street law grants you exactly zero seconds of response tim…

> then you have two seconds

Which takes us to 10 9s of reliability. Are there any computers with that level of reliability?

> The question is now, of course, whether only those I thought about this all day. I feel that from the moment you can no longer avoid to accident (no matter what you do), until it occurs, is the correct timeframe to measure. Not sure what that number is though.

Say you texted, got in a dangerous situation, then corrected. Is that really a failure? It's a risk of course, and enough people doing that will increase the final death tally. But for each individual driver it's not a failure, if it's not a failure then you can not count it.

i.e. doing it any other way would be counting it twice: Once for those situations that actually caused an accident, and again for taking the risk.

Or put another way, taking a risk and winning is not a failure. (Otherwise where do you draw the line on what a risk is?)

Re: Uber vehicle reportedly saw but ignored woman it struck

#309

Post by lhorie » Tue, May 08, 2018, 5:25 AM UTC

drcode wrote: post #245

lhorie wrote: post #219

Earlier quoted context omitted.

> it would be very much in Uber's interest to have a super dark video Why? As soon as the video came out, the first thing people (correctly) pointed out was that the dynamic range of the dashcam was crap. Besides, everyone knows that self-driving cars are essentially giant self-telemetry collection machines, so one would have to be either extremely short-sighted or incredibly stupid to think they would be able to get…

To be fair, I doubt the police just "popped the SD card out of dashcam"... more likely Uber had managers (and lawyers) at the scene within a short time window and Uber had time for very careful deliberations before handing anything to the NTSB (though as you say it's still more likely the got they raw dashcam footage instead of an altered video) Here is the more likely order of events: 1. Driver calls supervisor 2. S…

That still sounds pretty farfetched IMHO. You're basically saying that both the supervisor and a senior lawyer were on call at 10:30pm on some random day _and_ were immediately available upon request _and_ knew exactly what to do and who to call next _and_ the police were just twiddling their thumbs the whole time, despite there being a dead body on the scene. And lawyers/engineers can teleport from SF to Tempe somehow since there's no way they would've been able to fly in with that short notice so late at night.

If anything, I'd imagine that the police would be blocking access to the accident area and the car, especially to an alleged Uber employee trying to muck around with evidence without a government-sanctioned body overseeing forensics.

If I were to make a guess, I'd imagine it would've gone like this:

- "omg, I just hit someone"

- driver sees victim on the ground, panics and calls 911

- police and paramedics arrive

- police interviews driver, gets the account of the story, checks paperwork, get driver's information

- driver tries calling supervisor to escalate the issue (potentially unsuccessfully)

- paramedics declare victim dead, driver gets upset, body eventually is taken away while police tries to calm the driver

- tow truck arrives, takes car into police custody

- next morning, local police investigators look at the car in the impound, find the dashcam and look at the video. Uber ATG wakes up to the bad news and scrambles to figure out what went wrong and what to do next. The engineering team is still nowhere near Arizona and probably too busy wading through TBs of sensor/AI data to even remember there was a cheap dashcam installed in the car.

- Police releases a statement (w/ the dashcam video)

- NTSB gets involved in the investigation and gets in contact with Uber to understand how to acquire forensic data out of the SDV's databases.

Re: Uber vehicle reportedly saw but ignored woman it struck

#310

Post by ars » Tue, May 08, 2018, 5:26 AM UTC

sdenton4 wrote: post #104

ars wrote: post #15

Earlier quoted context omitted.

> get 5 9s reliability That's it? Humans have greater than 7 9's reliability when driving. (As measured by fatalities per time of driving, and assuming a fatality takes 5 minutes, and average speed is 30mph.) If the cause of a fatality is 10 seconds of failure, not 5 minutes, humans have greater than 9 9s of reliability. Math: 10 seconds / (1.13 / (30 miles/hour / 100 million miles)) Fatality rate is 1.13 per 100 mil…

Fatalities seems like a pretty low bar to measure failures by... I would prefer to measure reliability by reported accidents per 100k miles. This stat seems to be 183 per million miles in 2009, compared to 1.13 for fatalities per million miles. So quite enough to shave 7 9's down to 5ish...

> So quite enough to shave 7 9's down to 5ish...

From other comments in this thread, the correct number is actually 10 9's, not 7. So using the accident rate instead of fatality takes you to 8 9's.