First Driverless Car Death

Forums › Transport (cars, bikes and boats) › First Driverless Car Death

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | ... | 12

3438 posts

Uber Geek

Trusted

#1982403 24-Mar-2018 01:12

mclean:

networkn: The difference is, people can be held accountable.

Absolutely people will be held accountable when driverless cars crash. In a situation like this the software will be smart enough to identify several possibilities - maybe steer for a possible safe gap, a child, an animal, or simply carry on and hit the obstruction ahead. The manufacturer's software will make these life-or-death decisions, the logic will be clearly evident, and the manufacturer will be culpable.

I'm not sure that will always be the case (or even if it is now). I don't believe people are programming a set of rules, rather they use machine learning and pattern recognition. As things progress, it could be possible that we won't be able to understand why a car made a particular decision. Similar to trying to get Google to explain how it can recognise a picture of a cat vs a child doing the same thing.

gzt

17140 posts

Uber Geek

Lifetime subscriber

#1982731 25-Mar-2018 08:09

Mill ave at night. The incident occured in the less lit area at 33 sec in apparently.

gzt

17140 posts

Uber Geek

Lifetime subscriber

#1982740 25-Mar-2018 09:14

Accident diagram nyt:

cruxis

482 posts

Ultimate Geek

#1982832 25-Mar-2018 12:23

The car seems to be at fault in this one, It did not even brake or even react to the fact the pedestrian did not even look in the direction of the car.

Which makes me wonder can AI use subtle clues like a defensive human driver can? Like the pedestrian not making eye contact with the approaching car maybe it could started slowing down because they might like blindly continue to walk out?

Is AI is still fooled by cardboard cutouts of a human. Where a human would can easily recognise which is real and which is fake and drive though the fake ones.

FireEngine

1223 posts

Uber Geek

Trusted

#1982906 25-Mar-2018 16:57

cruxis:

The car seems to be at fault in this one, It did not even brake or even react to the fact the pedestrian did not even look in the direction of the car.

Which makes me wonder can AI use subtle clues like a defensive human driver can? Like the pedestrian not making eye contact with the approaching car maybe it could started slowing down because they might like blindly continue to walk out?

Is AI is still fooled by cardboard cutouts of a human. Where a human would can easily recognise which is real and which is fake and drive though the fake ones.

I'm pretty sure it can't - certainly the resolution won't be good enough to see whether eye contact is made (what with though, the LIDAR spinner on the roof of the car???) - I'm pretty sure it saw the outline of her with her bike and bags and didn't ID her as a pedestrian with a bike moving towards its path - it did seem to react to the actual impact and stop though - although that could have been the "safety driver" finally reacting...

Regards FireEngine

gzt

17140 posts

Uber Geek

Lifetime subscriber

#1983029 25-Mar-2018 20:49

mclean:
blakamin: Even Volvos standard "reindeer" avoidance should've prevented that.

Not at that speed - automatic braking in the XC90 turns off at 30 mph.

XC90 2016 manual says 80kmh for pedestrian and cycle avoidance.

Looking at the driver reaction I'd guess there was no driver intervention. Could be Volvo's post collision system may have brought the car to a stop.

Batman

Mad Scientist
29768 posts

Uber Geek

Trusted

Lifetime subscriber

#1983120 26-Mar-2018 07:41

The thing is, you can program for everything you can think of, and then there's always something you haven't thought of.

And then there's malfunction. If something as simple as an iPhone with ios can freeze doing very simple tasks, i suggest the car software is doing pretty complex tasks. Even something as simple as an emergency power cut off on a set of ski chair carousel managed to, instead of cut off power, malfunctioned to make the belt go backwards at high speed.

Cars killing people? Yeah funny how no one saw that coming.

And machines can't be blamed too, that's also a new frontier. Volvo will just move on every time a person dies, oh the car's fault. it's doing more good than harm, it's fine don't worry. Next you'll have AI police and soldiers, and the same arguments.

Quic Broadband

Move to New Zealand's best fibre broadband service (affiliate link). Free setup code: R587125ERQ6VE. Note that to use Quic Broadband you must be comfortable with configuring your own router.

kryptonjohn

2523 posts

Uber Geek

Lifetime subscriber

#1983123 26-Mar-2018 07:58

I think we all understand that no AI software will be perfect. The question for me is: will it be safer, on average, than humans?

tdgeek

29750 posts

Uber Geek

Trusted

Lifetime subscriber

#1983126 26-Mar-2018 08:05

kryptonjohn:

I think we all understand that no AI software will be perfect. The question for me is: will it be safer, on average, than humans?

I believe it will be vastly better than humans. I also dont believe there are that many scenarios it needs to understand.

It need to have enough sensors, thats not hard. It also needs to have a failsafe. Two sensors everywhere, each controlled by a seperate CPU and seperate power supply. If both fail the cars own ECU will cut power in the case of a autonomous vehicle, if it was a vehicle where it required a human in the drivers seat, it will cut power until the steering wheel etc was manually controlled

There are so many accidents caused by human frailty

gzt

17140 posts

Uber Geek

Lifetime subscriber

#1983136 26-Mar-2018 08:31

Batman: And machines can't be blamed too, that's also a new frontier. Volvo will just move on every time a person dies, oh the car's fault. it's doing more good than harm, it's fine don't worry. Next you'll have AI police and soldiers, and the same arguments.

If Uber is relying on Volvo's driver assist technologies to do something they were not designed for ie; self drive, that's probably a bad thing.

tdgeek

29750 posts

Uber Geek

Trusted

Lifetime subscriber

#1983142 26-Mar-2018 08:50

gzt:
Batman: And machines can't be blamed too, that's also a new frontier. Volvo will just move on every time a person dies, oh the car's fault. it's doing more good than harm, it's fine don't worry. Next you'll have AI police and soldiers, and the same arguments.

If Uber is relying on Volvo's driver assist technologies to do something they were not designed for ie; self drive, that's probably a bad thing.

Yes, and it was hardly driver assist, that driver appeared to be doodling on the phone. You cant assist the driver if the driver isnt acting as a party to managing the drive

FireEngine

1223 posts

Uber Geek

Trusted

#1983155 26-Mar-2018 09:02

tdgeek:

It need to have enough sensors, thats not hard. It also needs to have a failsafe. Two sensors everywhere, each controlled by a seperate CPU and seperate power supply. If both fail the cars own ECU will cut power in the case of a autonomous vehicle, if it was a vehicle where it required a human in the drivers seat, it will cut power until the steering wheel etc was manually controlled

There are so many accidents caused by human frailty

Two doesn't do it, you need three and a majority voting system so you can ascertain which of the three is faulty - to do it properly this gets complicated and expensive really fast.

Human frailty affects system design and software coding too. Thats the point. If one individual driver is inattentive then there is usually one accident as a result before corrective action is taken. If a software designer is inattentive on driverless cars then you could have many accidents before it is traced as a root cause and fixed...

Regards FireEngine

Item

1717 posts

Uber Geek

Subscriber

#1983156 26-Mar-2018 09:05

Decent analysis from Ars Technica showing footage which suggests the Uber dashcam was poorly configured or otherwise gave a misleading impression of exactly how well lit the area was. Also puts in some information on the XC90 headlights which are pretty damn good and again suggest that the victim could well have been far more visible for longer than the video is showing.

https://arstechnica.com/cars/2018/03/police-chief-said-uber-victim-came-from-the-shadows-dont-believe-it/

FireEngine

1223 posts

Uber Geek

Trusted

#1983161 26-Mar-2018 09:13

Item:

Decent analysis from Ars Technica showing footage which suggests the Uber dashcam was poorly configured or otherwise gave a misleading impression of exactly how well lit the area was. Also puts in some information on the XC90 headlights which are pretty damn good and again suggest that the victim could well have been far more visible for longer than the video is showing.

https://arstechnica.com/cars/2018/03/police-chief-said-uber-victim-came-from-the-shadows-dont-believe-it/

Yep - and to be honest who would want driverless cars that turn into blind killing machines if something as minor as a headlight bulb blew.

The Police Chief needs to stick to his specialist subject, no way should that car be navigating by visible light only.

The link states she was visible in the headlights for 1.4secs before impact, even if the car detected her only in the headlights that should resulted in 1.39 seconds of emergency braking...we aren't talking human reaction times here...

Regards FireEngine

kryptonjohn

2523 posts

Uber Geek

Lifetime subscriber

#1983177 26-Mar-2018 09:19

Yep - and to be honest who would want driverless cars that turn into blind killing machines if something as minor as a headlight bulb blew.

Oh good grief. "Blind killing machines"? Even my old VW can tell me when a bulb's out. You really think a production approved AI based car would carry on with vital equipment malfunctioning?

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | ... | 12