Guardian
'The Information' Report article
I'm still waiting for the NTSB report, but slightly alarming if a False Positive could be the reason (although expected), especially if it is true that other tech could have detected it correctly and braked.
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Uber self-driving car saw the pedestrian but didn’t swerve
- Thread starter NeonCarbon
- Start date
'The Information' Report article
I'm still waiting for the NTSB report, but slightly alarming if a False Positive could be the reason (although expected), especially if it is true that other tech could have detected it correctly and braked.
Wasn't that later disproved i.e. it wasn't as dark as the footage implies?
Like this video - https://www.youtube.com/watch?v=CRW0q8i3u6E
Watching the video (ugh) I'm not sure the crash would have been avoidable by even the most adept human driver so there's not much I think the safety driver could have done.
But we expect these self-driving cars to be better than people. It makes me wonder how..how many seconds sooner did the car detect her than a driver would have?
Edit:
Ahh, noted.
But we expect these self-driving cars to be better than people. It makes me wonder how..how many seconds sooner did the car detect her than a driver would have?
Edit:
Ahh, noted.
IIRC he was looking at his phone. Pretty sure they released a video showing his reaction too.
Last edited:
As a photographer, it is tough to gauge lighting conditions based on videos taken from two different devices.
The camera optics in the dash cam versus the cellphone are probably wildly different; with the latter likely better at low light conditions, broadly speaking. To me, it looks like Uber dash cam was exposing for the headlights of the vehicle, which, yes, would make the surrounding area appear much darker than reality. Meanwhile, the cellphone was likely exposing for the entire scene, making things somewhat brighter than what you'd see with the naked eye.
The truth is, most likely, somewhere in the middle. But, all that is largely irrelevant, since the story in this article points to the real reason the car failed to avoid her: a failing in the coding.
Yeah, the video released showed him on his phone and he reacted at the last minute as the car was about to strike. Easily avoided had he been focusing on the road ahead.
Damn, she basically came out of nowhere, even if it wasn't as dark as it looked like it was from that camera, some of the responsibility has to be with the jaywalker herself. Not trying to victim blame, but if it is a 150lb person vs a 4000+lb car, the 150lb person has to look after themselves.
I do find it a shame that it is going to be a long, tough road ahead for self-driving cars. Any accidents that do happen are going to be criticized extensively, even though they are still much more safer than a comparable human driver.
Well that video also has string lights that look like they help illuminate the road.
Link to hte actual video. It doesn't show the actual collision, just right up to it:
https://www.youtube.com/watch?v=q7d90ZFhg28&pbjreload=10
I do find it a shame that it is going to be a long, tough road ahead for self-driving cars. Any accidents that do happen are going to be criticized extensively, even though they are still much more safer than a comparable human driver.
Well that video also has string lights that look like they help illuminate the road.
Link to hte actual video. It doesn't show the actual collision, just right up to it:
https://www.youtube.com/watch?v=q7d90ZFhg28&pbjreload=10
I watched the video closely at the time. The driver had his eyes off the road for six seconds from memory.
Given the position of two street lights just metres from where the pedestrian crossed, I believe the driver would have easily seen her if he was looking at the road. The original video gave a false impression of how dark it was which can be demonstrated by the barely visible red light and illuminated car park in the intersection ahead.
This will be a tricky one to sort out using software. Do you slam on the brakes or swerve and hope all the other drives don't crash into each other and kill more people or do you hit the object and hope for the best. In this case walking right in front of a moving vehicle seems not the best thing to do. There didn't seem to be a lot of traffic so the car should detect the vehicle behind and decide if it should break or not. This is something human beings are generally not capable of doing though.
It seems like this is the common factor in these accidents. People will say it reflects negatively on the tech, but the fact that they have safety driver show these companies are under no illusion that it's ready for full autonomy. They need to do something about these drivers not staying alert
This was the most glaring thing for me, the employee is fully aware he is being recorded so clearly Uber did not care if they actually paid attention.
It confirms that the tech issue was more software/algorithm related than hardware?
Well, that isn't that relevant. The car hit her not because it was dark, they use radar/lidar to detect objects and obstacles. It was a software problem.
Uber has a bit of a reputation for treating its female employees poorly.
http://fortune.com/2017/10/25/uber-sued-alleged-racial-gender-discrimination/
I appreciate you guys actually asking for clarification instead of jumping to conclusions.
Still, what does that have anything to do with the topic at hand?
There really is no jumping to conclusions, the report said that a self-driving car hit a pedestrian, and you said that it "must have been a female employee". The only correlation that is there is you saying that females are bad drivers.
Even if that isn't what you meant, you need some clarification when you post, because there is no fucking way anyone could see that post and think of "oh yeah, uber treats female employees badly".
If their self driving technology sees obstacles but is unable to tell if it is a harmless object like a plastic bag or a person then it should not be on the road.
Yes, clearly the natural conclusion to reach regarding a comment about a story involving a driverless car hitting someone is that the comment is about the driver, not the person struck.
Heard about this story but this is the first time seeing the video.
Not putting blame on the pedestrian but would it be smart to slow down to check if the car is slowing down to let you pass?Did the pedestrian kept walking assuming the car is going to stop? Again, NOT saying it's the pedestrian fault.
Even if it's not human - like a cone or cardboard box for example , should've the car still stop or slow down?
Not putting blame on the pedestrian but would it be smart to slow down to check if the car is slowing down to let you pass?Did the pedestrian kept walking assuming the car is going to stop? Again, NOT saying it's the pedestrian fault.
Even if it's not human - like a cone or cardboard box for example , should've the car still stop or slow down?
Person had one job and failed miserably at it. Clearly the technology isn't ready and why a person needs to be in the car.
"The" technology is pretty good. Uber's technology in particular, on the other hand, sucks ass.
I like the concept behind the joke, but I think you gotta hold the L on this one. At first glance, your post just looks like a lame woman driver comment.
Does Uber use some variant that is majorly different from what Tesla or others use?
That video is shows how easily people can be manipulated. Has no one ever driven outside in the dark before? How in the world can you believe for a second that it is that dark on a well lit road? Maybe I should get a dash cam and give it terrible black crush so I can get away with murder because it's clear that as soon as people see video they completely turn their brains off and lose all reasoning ability.
The reason people bring up lighting is to compare to a human. If a human would've also hit this person, then it's not really much of a knock against AI cars.
There was a human driver in the car though. I took your comment to be a "hurrrr girls can't drive good" post as well, and it seems like you're trying to bait/gaslight people tbh.
Yep my first thought too, the person literally walked in to the path of a car, you never ever want to assume that any person driving see's you.
Very stupid and avoidable.
Also the cars recognition is fucking awful if it thought an object of that size wasn't relevant.
In the long term the only thing that actually is relevant is if they are safer than human drivers.
I would be willing to bet almost anything that they are monumentally safer than human drivers already.
Glad to see Uber feeling repercussions here.
Hope there are consequences for the driver as well since they onviesly weren't paying attention when they were supposed to.
Hope there are consequences for the driver as well since they onviesly weren't paying attention when they were supposed to.
A human paying attention should not have hit her, thats the entire point. Everyone saying "yeah I would have hit her too" needs to turn in their license.
Watching the video of the incident it seems probable that the outcome would have been the same with a human driver, it's very dark where she is crossing.
This could have been a much different story had the software been in order and swerved away, but it will take some time before we fully get there.
The fact that we haven't really seen many of these cases is a major indication that the technology is going in the right direction and it's a question of time before cars automatically can make the best choice before a would be accident happens.
This could have been a much different story had the software been in order and swerved away, but it will take some time before we fully get there.
The fact that we haven't really seen many of these cases is a major indication that the technology is going in the right direction and it's a question of time before cars automatically can make the best choice before a would be accident happens.
Waymo (Google) is light-years ahead of Uber, but Uber needs self-driving cars to survive while Google does not.
Uber appears to be cutting corners in its desperation.
https://www.theverge.com/2018/5/9/1...rless-cars-deep-learning-neural-net-interview
Uber appears to be cutting corners in its desperation.
https://www.theverge.com/2018/5/9/1...rless-cars-deep-learning-neural-net-interview
Last edited: