|
The UK speed tolerance has been tightened, you will now be booked if caught at any speed above the limit. That is not the point, though, the limit is an absolute maximum; you are obliged to drive at a safe speed for the road conditions. In the dark on a road that is badly lit and accessible to pedestrians that is perhaps 20mph. The risk of death to a pedestrian rises steeply between 30 and 40mph, so every mph counts. If you were to program autonomous vehicles to be safe, they would be limited to lower speeds than human ones. And would always give way. And in congested traffic, slowing all vehicles makes journey times less, because flow is restricted by junctions, not maximum speed. But try selling that to anyone.
|
|
|
|
|
Be very careful with that, you're quoting years old wisdom.
Many police forces have lowered the "unofficial" tolerance to between 0 and 3% depending on the country, UK included (3%); reasoning is today's speedometers are far more accurate than they were >20 years ago. (In NZ/Australia it's zero and they have days where they enforce that to the brink, often though it mostly it depends on the cop on the day and with respect to prevailing conditions including the speed of the other traffic.)
Anything, even 0.5 mi/km above the speed limit[^] is ticketable even if your speedo is out because it's also "the owners job to keep their vehicle in good, legal, road worthy, and proper working order." Finally they will also state if you are unsure you may drive a little below the posted limit, however too much under, particularly if holding up other traffic is often nearly everywhere also an offence unless you are doing so to avoid a dangerous situation.
Signature ready for installation. Please Reboot now.
|
|
|
|
|
OriginalGriff wrote: I'd suspect other countries do the same thing.
Yep. Brazil is the same. Just slightly different tolerance rules.
To alcohol! The cause of, and solution to, all of life's problems - Homer Simpson
Our heads are round so our thoughts can change direction - Francis Picabia
|
|
|
|
|
HobbyProggy wrote: Common sense and experience would probably have saved her
Yup, but apparently she was lacking some.
On a more serious note it's true that many human drivers (including myself) watch out for other road users that may act inappropriately and prepare for that possibility, e.g. by slowing down or passing at a wider distance than strictly necessary. However, it's not reasonable (and often not possible) to do this for every other road user, and the reports indicate that the woman was acting abruptly, with little to no prior indication, LiDAR or not. This is the point where you have to rely on others to behave with a reasonable amount of self-preservation. Lacking that, not even AI can break the Darwinian law ...
HobbyProggy wrote: And another thing, the car was obviously speeding
Was it? I only saw statements about it not showing any signs of slowing down. Which is kind of odd, given that the LiDAR systems should have been able to recognize an obstacle and should have caused the system to do something to limit the effect of the impact, assuming that it was too late for evasive maneuvers.
GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
|
|
|
|
|
Well, yes.
The latest news i saw was that the woman was crossing the road not abruptly, the car and the safety driver should have recognized her. But we will see what the police will state.
I also i have to correct the speeding thing, initially they said it was driving 40 in a 35 zone, now it shows it was actually a 45 zone, so no speeding.
This article has a good picture of the situation i think -> Pedestrian killed by Uber Self-Driving Car[^]
Rules for the FOSW ![ ^]
if(!string.IsNullOrWhiteSpace(_signature))
{
MessageBox.Show("This is my signature: " + Environment.NewLine + _signature);
}
else
{
MessageBox.Show("404-Signature not found");
}
|
|
|
|
|
Thanks for the link. It's much more detailled than any of the other reports I've seen. Not to mention that there are barely two that offer matching information.
In another forum, someone said that, according to one report, the woman was crossing from west to east, which would mean she had to pass across most of the road before the car hit her. I consider that extremely unlikely as both she should have seen the car well in advance while starting to move across, and the driver should have seen her very clearly as well. Also the statements about the video footage seem to indicate otherwise.
However, if she was indeed coming from the west, I wonder whether she expected the car to see and move around her, not realizing that it was an autonomous car that may not be capable of predicting her (inappropriate) intent correctly. In some places (e. g. Paris, France) this would be a common method of crossing the street (preferably without looking left or right, and your head dug into the wide open newspaper), and the drivers there would indeed go around you, no problem (which means any autonomous system hoping to gain traction there should be able to do the same). But I suspect that in Tempe the traffic standards are not quite like that...
Either way, it seems to me that although AI can beat human intelligence at chess and go, it still fails when faced with true human stupidity.
P.S.: it seems the speed limit changes from 35 to 45 mph a short way to the south of the location shown in that article (speed signs can be seen at the start and end of the bridge in Street View)
GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
|
|
|
|
|
The article you linked was updated to show part of the video footage.
GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
|
|
|
|
|
Saw it, and that clarified for me that the car was malfunctioning and the "safetydriver" failed totally. The darkness is just because the Dashcam doesn't work as good as a human eye, therefore the light seen might be way of what the driver should have seen and this also should not affect the radar components!
Rules for the FOSW ![ ^]
if(!string.IsNullOrWhiteSpace(_signature))
{
MessageBox.Show("This is my signature: " + Environment.NewLine + _signature);
}
else
{
MessageBox.Show("404-Signature not found");
}
|
|
|
|
|
I doubt that the driver could have reacted in time even if she had paid close attention: the woman was crossing midway between two overhead lamps, in the darkest area of the street, not wearing reflective clothing, and no active lights on the bike. Even when considering that the driver's eyes should have adapted somewhat to the darkness, it was near impossible to spot the pedestrian wearing dark clothes in the darkest possible area outside the range of the headlights.
GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
|
|
|
|
|
And that's exactly where the cars systems should have kicked in, by the way, you saw the driver looking to the left ? It felt kinda like she spotted the pedestrian but that is just an assumption. But still, the safetydriver is there to react and pay attention, she failed on that job.
It really is questionable if the accident could have been avoided but i guess since there was no breaking effort done by the car they are mostly responsible for the accident.
EDIT:
An HDR picture of the scene at darkness Where it happened - Album on Imgur[^]
Rules for the FOSW ![ ^]
if(!string.IsNullOrWhiteSpace(_signature))
{
MessageBox.Show("This is my signature: " + Environment.NewLine + _signature);
}
else
{
MessageBox.Show("404-Signature not found");
}
|
|
|
|
|
I agree that the car should have reacted. Even with just the video as input, there was at least a second to hit the brakes. there is no good reason why it didn't.
GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
|
|
|
|
|
And the car has radar installed
Stefan_Lang wrote: Even with just the video as input
Rules for the FOSW ![ ^]
if(!string.IsNullOrWhiteSpace(_signature))
{
MessageBox.Show("This is my signature: " + Environment.NewLine + _signature);
}
else
{
MessageBox.Show("404-Signature not found");
}
|
|
|
|
|
And as I understood it at night, in the middle of the street (crosswalk down the street) and at least on a bend in the road.
|
|
|
|
|
Statistically, robots are already better drivers then humans. That is why the whole world talks about one casualty on the road while every year human drivers extinguish one Vietnam war of Americans.
Understanding politics of fear, I'm afraid a lot of people will perish before we realize that it is better to have less people killed by robots, then demand perfection from machines while at the same time licensing 16 year old kids to drive after 20 hours of training.
|
|
|
|
|
Tomaž Štih wrote: That is why the whole world talks about one casualty on the road while every year human drivers extinguish one Vietnam war of Americans. Of course you are taking into account the disproportionate numbers of human drivers to robot drivers to arrive at such shocking statements, otherwise it could just as well be plain rhetorics.
My old cat was excellent at tracking moving objects and judging distances and speed. It also was arguably far more intelligent than any AI up to now. Perhaps it would have been a good idea to train the cat and give it a driver's license?
I have lived with several Zen masters - all of them were cats.
His last invention was an evil Lasagna. It didn't kill anyone, and it actually tasted pretty good.
|
|
|
|
|
First studies in 2012, understandably, showed lower accident rates for human drivers. But latest (2017) results now shows lower accident rate (per million miles) for robots.
|
|
|
|
|
"Toonces the driving cat" was probably about the same as this AI with one notable difference.
The AI hit a pedestrian.
Spoiler Alert! In case you search YouTube
Toonces injured its own passengers.
|
|
|
|
|
Oh, you have those statistics? That would be an interesting read.
|
|
|
|
|
|
Thank you for that, it was an interesting read.
From that report it seems that the self driving cars have less minor accidents. It is a lot closer with the more significant accidents, but self-driving cars still have less accidents (although by the admission of the report there is too little data to form any conclusions.)
I personally think they over-estimate the number of unreported serious accidents - although I might be wrong there.
One thing they omit is the number of incidents that are averted by the driver interceding. I believe all of the data was gathered with an actual driver. What we are seeing more of now is driverless cars.
BTW in case you have not guessed I am against driverless cars, as I do not think they are ready yet, but I am not against self-driving cars.
|
|
|
|
|
After seeing the accident I now think there is a serious flaw in Uber software. The car didn't even try to apply brakes. Going 40mph directly into human. Besides the fact that there was a time frame of cca. 2 seconds (enough to at least try to apply brakes), and that sensors must've detected the obstacle on the road (Uber has multiple lidars, and radars!) long before that. It probably had at least 6 seconds to react, because it can see in the dark. For self driving car this was an avoidable accident.
|
|
|
|
|
Remember the American Mantra: "Guns People don't kill people, people cars kill people".
Bad command or file name. Bad, bad command! Sit! Stay! Staaaay...
AntiTwitter: @DalekDave is now a follower!
|
|
|
|
|
Excatly! It's so easy to use it...
––––
Bang!
|
|
|
|
|
CodeWraith wrote: even the dumbest human driver has a few million years of evolution behind him
When that human driver happens to be a p***ed up 17 year old blasting out da G-funk at 110 dB whilst smoking a spliff, sending text messages and driving at faster-miles-per-hour past a school hoping to impress the girls, I'd question the value of the evolutionary input!
98.4% of statistics are made up on the spot.
|
|
|
|
|
PeejayAdams wrote: I'd question the value of the evolutionary input!
Nope: "Evolution in action".
Bad command or file name. Bad, bad command! Sit! Stay! Staaaay...
AntiTwitter: @DalekDave is now a follower!
|
|
|
|