> To attack your specific example, why do you assume the cars will drive around passenger-less, and not merely continually make productive uses of themselves, such as by delivering other packages or acting as deliverymen for physical goods?

Because they won't most of the time? Let's say that during peak rush hour, N people + M packages need to be transported. That means that in order to satisfy peak demand efficiently, we need N+M cars and road capacity. During off-peak hours, lets say that only 20% of the population needs to travel. That means you've got the other 80% of the cars sitting around doing nothing just like they are today. If the average distance to nearby parking is say 2 miles, then we have 80% of the cars just driving empty 2 extra miles to prevent themselves from taking up road capacity (while today we trade off by creating parking lots).

Despite that, I do think there will be some improvement in terms of car use. Namely:

* the average car size will shrink to probably that of a smart car (2 seats, little cargo room) * parking lots will shrink (but not disappear) due to effective stacked parking * vehicle efficiency will increase (human drivers are inherently bad at maintaining fuel/energy efficiency with acceleration/deceleration)

The danger is that this once again will make car use "cheap" and like the article suggests, extends sprawl yet again. So we'll just have another half a century of poor land use just it might be slightly more efficient this time around.