> At its most efficient, an Uber model can automatically and
predictively allocate rides based on realtime rider and driver current and predicted locations with realtime traffic/travel data. [emphasis mine]
This seems like a bizarre claim. Precisely when a specific rider will call for an Uber is inherently unpredictable (at least without a lot of surveillance). Meanwhile Uber drivers are not paid to drive from place to place without a rider - so Uber has limited capability to make them do something like head to a high-volume area in anticipation of demand. This is especially true since almost all drivers I talk to now drive for Uber, Lyft, and whoever else at the same time.
The micro-optimizations of dispatching drivers based on what side of the street I'm on or how many left vs. right turns are on the route are nice but only have a meaningful impact on wait time when there are already numerous drivers in the area to choose from (i.e. when the wait time is already short).
Edit: On second glance I realized you may mean that the dispatch is predictive in terms of anticipating traffic conditions and factoring this into the matchmaking between riders and drivers. While this is valuable, the key innovation here is the same kind of predictive routing algorithms that both Google and Apple already provide. Building this yourself may be hard, but is far from novel.