grumpyken52:
Rjan:
[…]That in car video certainly doesn’t give the impression of a well lit urban highway and gives the impression that the poor woman emerged from a particularly dark patch .
I do know of certain areas where the street lighting is more like marker lights than illumination.
The onboard scanning and aviodance systems were certainly not fit for purpose in this intance .
I personally feel that driverless or autonomous vehicle systems are nowhere near ready for general use and should never be used in a situation where they interact with anything but others of the same type .
Agreed. There is no question that either the video has been fiddled or the on-board camera was not fit for purpose.
As the ArsTechnia link shows, the highway is brightly lit with sodium lights, and it is clear by comparison that all street lights were working on the Uber video.
We all know from experience that a sodium-lit road in dry conditions does not cast those sorts of shadows when seen through a human eye - it would be contrary to the very purpose of lighting the road if you couldn’t see things in the road directly ahead (it would be more sensible not to light the road at all and force the car’s main beam to be used).
If you look at the building (or maybe a multi-storey car park) in the background also, the windows/interior clearly glows with the interior lighting and the signage is outstandingly bright (as you’d expect from any corporate setup in an urban area), whereas on the Uber version it’s dull, muddy, and you can barely see it.
Frankly, on the Uber version it appears to be a completely unlit rural road during a new moon, whereas that is clearly not the case.
I personally think it’s an utter deception released to discredit the lady who was killed, because the computer is not dependent on video visuals but on radar principles that can “see” equally well in the dark or through adverse weather, and even if the road had been pitch black, the car should have automatically slowed down within the distance that could be seen (especially if the human occupant was being expected to provide some residual supervision of the driving task).
Even if we accept that it exonerates the driver, because he/she (gender is not clear!) could not have been expected to see the pedestrian clearly, it would then simply show that the computer was driving the car far too fast both for it’s own limitations and that of its human supervisor.