The argument that as long as they cause less incidents than human drivers they are a win has to go. Because that only works if the statistics of the environment are stationary.
I think they're getting at something like this: If self-driving cars resulted in dramatically more miles-traveled-in-car per person they could be safer per-mile and more efficient per-mile while making some important total outcomes worse.
Relevant to this example: if people travel by car more because they care less about traffic when they're playing video games or on TikTok during it, instead of driving, overall congestion will likely go up which makes emergency services worse.
That’s fair but I think it’s kind of orthogonal to the original point.
As is already the case in cities with robust public transit systems you’d need to make sure you’re applying the right incentives (i.e. taxes and charges) to make sure people are making decisions that benefit everyone. That doesn’t alter the possibility of self driving cars being much safer than human driven ones.
