As you might have heard, a couple of months ago Uber found themselves involved in a self-driving car accident resulting in the death of a pedestrian. It was later revealed that Uber had apparently scaled back the number of sensors used in the car, but that might not have been the full story.
According to a report from The Information (paywall; via MIT Review), apparently Uber’s self-driving car sensors did detect an obstacle on the road but chose to ignore it anyway. Now this isn’t as sinister or careless as it might sound. Uber’s self-driving car had been programmed to ignore false positives on the road that would not have affected the way it drives.
For example if you see a plastic bag on the road, would you swerve or brake violently to avoid it? It has also been suggested that this was done out of rider comfort, where Uber tried to make it so that its self-driving cars would have less start-and-stop experiences which are obviously not a fun ride to be on.
It is rather unfortunate but hopefully other companies will be able to learn from their mistakes, and perhaps come up with better sensors or perhaps find a way to deal with false positives in a better way.
Filed in . Read more about Self-Driving Cars, Social Hit and Uber.