I am surprised at the amount of comments that betray, IMO, a naive and narrow view of the issue.
John’s points are very valid and the complexity of the issue, even if it were about safety alone, is real.
First, the argument that the situation that John, for sake of example, describes would not even be possible because autonomous driving is inherently safer is ludicrous and misses entirely the point. It is equivalent to saying that it is possible to create an unbreakable IT security system.
Second, as autonomous driving reduces fatal accidents and traffic collisions in general, we, as a society, will collectively raise the bar for what is an acceptable “risk”. Today, in the US, we tolerate 30k car deaths a year. It is entirely possible that in the future even 30 would be too high a number (you just need look at commercial aviation to see how different perceptions of what is an “acceptable” failure rate shapes an industry.) In this context, the relevance of fringe, extreme conditions will increase.
Third, human factors will never be taken out of the equation. For once, there will always be someone who wants to keep driving their car, even if it were to become unlawful. There will always be someone behaving in unpredictable, un-logical ways.
To say that design/governance issues are overstated is to not realize that failing to address them now may very well end up delaying the adoption of the technology. Even in the face of its obvious benefits.