Susan Crawford :

Tesla updates its autopilot system through pushed software upgrades, which means big downloads by each car. But there’s also a tremendous amount of data being uploaded by the system as each car learns from its experience: The whole point of “deep learning” computing is that computers like those in the Tesla fleet of cars are finding patterns in enormous amounts of data and making predictions based on those patterns. And then spreading that knowledge around the fleet.
 The data ingested and spewed out by these vehicles will become overwhelming at the same time that our desires for immediate, no-delay inputs and outputs from these vehicles will become requirements. And the only way to move all this data around is to have fiber everywhere, in every neighborhood in the country.