I liked to points raised by this comment by one Peter Swinton in the article cited:
"So an autonomous car changed lanes and decided to change back when the car in front slowed, hitting a motorcycle that had moved into the space it was vacating. This leads to a number of questions:
Did the car sense that the motorcycle was there? This should be in the data log.
If not, why not.
If it did, why did the protocols choose hitting the motorcycle as the preferred option compared to potentially hitting the braking car in front or potentially being hit from the car behind?
Did the motorcycle honk? I have a loud dual-tone horn and have found it effective in preventing a lane change where the driver didn't see I was there. Even if the rider did honk, can an autonomous car hear, and if so, how would it be programmed to respond?
....
I don't think this is quite as clear cut as the police report implies. And it warrants further investigation into the sensing abilities and protocols of autonomous cars. I would think that knocking down a motorcycle, bicycle or pedestrian should never be a preferred protocol compared to denting another car."
