Uber halts self-driving car tests after death

[Gunge]

Wait til the climate-change crazies get their way and we're all reduced to using Shanks' pony 'to save the planet' and other utter crapola.

 

[Linalmeemow]

Wait til the climate-change crazies get their way and we're all reduced to using Shanks' pony 'to save the planet' and other utter crapola.

You're like a stuck record!

 

[Gunge]

You're like a stuck record!

I know, but at least I'm right. We can't let them get away with it.

 

[-Bezza-]

The video is now on the BBC so you can consider for yourself the extent to which the car was at fault: http://www.bbc.co.uk/news/world-us-canada-43497364

 

[foxbat]

The video is now on the BBC so you can consider for yourself the extent to which the car was at fault: http://www.bbc.co.uk/news/world-us-canada-43497364

The police should reconstruct the accident to determine if a human in the car that was actually looking at the road could see movement in the grey of night. The camera makes it appear that there's nothing to see before the pedestrian enters the headlight beam but that's because the limited dynamic range of cameras will black out what the human eye sees as levels of grey - and our eyes are hypersensitive to movement. If a human could see movement but the sensors did not then frankly the sensors are not up to the job.

 

[Saisonator]

You would think the cameras would have IR detection?

 

[Cwrw666]

Don't talk to me about sensors. We had someone here helping us with the sheep. Afterwards she drove us back to the house accompanied all the way by constant beeping - the proximity alarm apparently. So presumably any `autonomous car' would just grind to a halt half a mile away, being too afraid to negotiate our narrow lanes and drystone walls.

 

[-Bezza-]

The police should reconstruct the accident to determine if a human in the car that was actually looking at the road could see movement in the grey of night. The camera makes it appear that there's nothing to see before the pedestrian enters the headlight beam but that's because the limited dynamic range of cameras will black out what the human eye sees as levels of grey - and our eyes are hypersensitive to movement. If a human could see movement but the sensors did not then frankly the sensors are not up to the job.

Quite, the jury is out based on the video and one of my immediate thoughts was whether the dynamic range of the camera was obscuring things. I agree the human eye can probably process information quicker than a computer in this sense, but at the same time, people manage all sort of stupid things whilst driving.

I'm neutral over the point about whether the operator should have stopped the car. We should focus on whether the automation of the car was up to scratch as in an non-automated car, a driver can get distracted and cause an accident, so it's largely irrelevant here.

In assessing the technology I think the point is whether, on average, the automated car is better or worse at preventing an accident in any given situation. One accident does not mean the technology has failed if human drivers would have caused 2 accidents had they been driving.

 

[foxbat]

It's a well lit major road at night. Here's some footage from a non-uber camera of what it actually looks like. Fast forward to 30s in: