> There was plenty of time to stop or swerve to keep this woman alive

You have never been in a serious car wreck. I've been in three in my life and I am telling you, any human being that wasn't a racecar driver in a state of flow would not have been able to stop the car before hitting that pedestrian. It's terrible, yes, and Uber's autonomous vehicle should have stopped, but it simply wouldn't be possible for a typical human being to take appropriate action in that window of time when the pedestrian came out of the shadow. One way or the other, whatever "action" you think someone could take in that particular moment would have resulted in an injured or killed pedestrian and/or an injured or killed driver.

In this specific situation, Uber's autonomous vehicle was no more dangerous than a vehicle driven by a human being.

People do this all the time with things like this. Car wrecks, combat, disasters, fights, emergencies. Everybody has the inside track on what the failures were and why they wouldn't have done it that way and why they wouldn't have failed in that situation.

Have you seen the video from the guy who drove the same route?[1] Streetlights do not produce a string of light dots surrounded by complete darkness. Either the footage has been darkened, or, more likely, the camera just has incredibly shitty dynamic range.

Furthermore, the driver did not have his eyes on the road, he was looking at a display, or probably his phone. I'm not exactly sure, but it doesn't even look like he had his hands on the wheel. Which isn't surprising, like, at all. You cannot sit somebody into a car without anything to do for hours, and expect them to be able to step in fast enough. If you want those people to stay alert, you have to take the appropriate steps like driving in pairs and extremely frequent rests.

[1]: https://www.youtube.com/watch?v=1XOVxSCG8u0

You know, actually, I did see this earlier today. EEVblog made a new video where he talks about some of the developments. One of those was all the footage out on the internet from other people with other cameras showing that it wasn't nearly as dark as the Uber video. Which raises some questions, as you mention, especially with Uber's reputation being what it is. Something else interesting he said was that the makers of the vehicle's own collision-avoidance system -- which Uber had turned off -- took the footage released by the police and ran it through their collision-avoidance computer vision algorithms, and it detected the pedestrian even in that dark video. So it's beyond any doubt that a computer both could have and should have stopped the vehicle. I'm eager to see the rest of the data (as I'm sure other autonomous vehicle divisions are, too) and how exactly Uber's automation failed so spectacularly.

I don't believe that we have enough information to conclude that. Yes, she wasn't visible in the released video until 1-2 seconds before impact. But there were no obstructions preventing a driver from seeing her two lanes to the left, about five seconds before impact. Landscaping did shadow her from the nearest street light, but it's unclear whether the driver could have seen her. Perhaps in the headlights.

The pedestrian wasn’t in shadow, the terrible low-light performance dash cam only made it seem like she was.

If the safety driver had been paying attention she would have seen her and been able to stop.

If visibility really was that low then they shouldn’t have been driving that fast - drive to the conditions.