HOV lanes are distinct from driverless care sharing, in my opinion. HOV lanes are separated from the rest of the highway, representing an allocation towards shared driving - whether or not they increase throughput depends on whether or not enough vehicles with extra passengers use the HOV lane as compared to if the HOV lane was just part of the highway.

Driverless car sharing is a scheme where drivers don't necessarily own their cars - that's kind of a gray area that I didn't think was discussed well in the article yet. Depending how they are priced and how the liability works, it's unclear to me if people own their own driverless cars, or if it's a service that you subscribe to and pay monthly or per trip - where vehicles show up as you need them, perhaps with additional passengers. This stuff is super interesting as a planning problem, and one that is likely to be addressed by computer scientists rather than transportation researchers, or perhaps together.

Your post seems to imply that individuals would own driverless cars, in which case the 90% reduction would be silly, and I agree. However, the implication that they are making in the article (I think) has to do more with the case that driverless cars also shift our behavior for how we'd like to live - should we choose denser living and a rental system instead of ownership for cars, it is plausible (to me) that we could vastly reduce the number of cars. Living in Cambridge, MA, where parking spaces are severely limited, it's amazing how I almost never even get into a vehicle, minus a few zipcar trips (which I don't find worth it). However, there are cars parked on all the residential streets largely unused, and I think a city like Cambridge would benefit a lot from driverless cars, where trips are short, although increasing the availability of driverless cars could cause the number of trips to go up. But if it could decrease the number of cars and travel time (by reducing congestion), I think that a lot of parking space could be reclaimed for other uses.

The effect that driverless cars will have on how we live is studied, but I think a lot of the impact depends on how driverless cars would be monetized, assuming everything works out from a technological perspective. I do think that driverless cars will prove to be much safer, statistically, than human operated vehicles, although I'm skeptical as to whether we will have cars platoon with minimal space between cars, at least initially. Regarding the impact on society, if it makes driving cheaper/more appealing, we may just continue with sprawl, although the reduction of congestion could make things "closer", whereas if it becomes more expensive, then we might accelerate the growth of the urban population and a rise in walking, bicycling, and public transit.

Additionally, a significant hurdle is the inclusion of manned vehicles in the system - whether people who still drive their cars exist and how they interact with driverless cars. I don't imagine that legislation would ban users from driving their own cars, but that could decrease the productivity of driverless cars. Ultimately, it would have to be a culture shift where driving would no longer be desirable, but that seems very far off.