That's some appallingly misleading footage.

1) It's that dark to a daytime camera. That's visible spectrum, not infrared spectrum, which you have to filter out if you don't want it in modern chips. What your average dashcam picks up in night is shit.

2) By the time that footage starts, LIDAR had her for seven seconds already.

3) She'd already been a stationary object (from the car's perspective) in the opposite lane for as long as it took her to cross a lane (at least six seconds, probably longer).

4) the interior footage is 100% infrared, and you can see the poor schlub sitting there no problem even though the only thing lighting him was the dash lights. So they have the infrared footage. Shit, they've got the telemetry from all four cameras, the LIDAR and the RADAR out front but what they're releasing is grainycam.

And what does the infrared footage show? The poor schlub surfing his fucking phone. Because the car is driving itself. It is, after all, a "self-driving car."

Trust me. You wouldn't have killed her. You would have seen someone lit by streetlight and headlights in the lane ahead of you and you would have slowed. You don't think a brand-new Volvo XC90 doesn't have headlights good enough to see the next lane over at 38mph?

If yes, then Uber has done its job.

on post: How a self-driving car killed a pedestrian in Arizona
by kleinbl00 2445 days ago   ·   link