In terms of intention and with respect to design and implementation, there are three models of distributed AI systems.

1) An AI system that exclusively optimizes for the interest of the end-user (competitive oppositional model)*

2) An AI system that optimizes overall outcome for all users of the system (non zero-sum game model)

3) An AI system that optimizes outcome for select users of the system (pyramidal hierarchical model)

Google maps and its infrastructure is the prime current example of a distributed AI system. Currently this system is in its infancy, and because it is striving to gain acceptance it primarily serves the end-user.

However it is very foreseeable that when the system becomes the norm, the focus of the traffic AI system may shift from the end-user to optimizing the system as a whole, or primarily serving elite users.

Indeed as drivers are replaced with AI systems, this change in focus may be invisible to the general population.

The US seems poised to spend multiple trillions of dollars to remove the human component to driving and replace it with AI driving systems. This will undoubtedly vastly improve the safety and civility of our roadways.

However the interesting question is “How long will this system take to pay for itself, if ever?”

Truckers and trucking companies make a great deal of money because they are organized and close to the money. When products, parts, raw materials, and equipment stop moving the manufacturing and distributions systems go into cardiac arrest. It seems probable that the elites directing long term policy would like to disempower and eliminate this group of employees and employers, and replacing them with AIs is one viable method. I am guessing this group of workers are not going to go quietly into the night.

However distributed AI driving systems also solves another tricky elite problem. Elites tried to use economic exclusion in the form of property value escalation, high fixed costs, and city infrastructure design to limit vehicular traffic regionally and locally. They were unsuccessful. The problem is that if a elite person and an non-elite person want to travel from point A to B, it takes the same amount of time. In congested areas this time may be lengthy, and this is not mitigated by the fact that the elite’s vehicle may cost 20x more money.

Distributed AI driving systems could eliminate this elite problem by giving them priority. Invisible priority. As this kind of system gets entrenched and absolute, it may indeed happen that many destinations will be unavailable to non-elite persons unless they walk there. Navigation systems will simply not show them as a valid destination without priority access privileges.

* “first in first out” is an example of a competitive oppositional model based on timing. Another model would be service based on monetary amount paid IF people in a given population generally had similar amounts of unbound currency to buy priority access.