Privatizing Intelligence

There is something peculiar about the idea that a small number of private companies might eventually provide much of the computational capability through which people access knowledge, navigate institutions, produce work, learn, communicate and make decisions, and that we discuss this primarily as an exciting new market.

Of course, nobody is literally privatizing intelligence. Humans do not stop being intelligent because a company builds a model. What can become privatized, however, is access to increasingly consequential cognitive infrastructure. If workplaces, schools, public institutions and entire professions begin reorganizing themselves around these systems, not using them may eventually become about as realistic as running a modern company without internet access. Formally possible, perhaps. Practically rather difficult.

That creates a familiar kind of power. First build something useful. Then make it ubiquitous. Then allow institutions to reorganize themselves around it until dependency becomes normal. At that point, the strongest business model may no longer be selling a product people want. It is owning infrastructure people can no longer reasonably avoid. Privatization becomes considerably more consequential when what has been privatized is the road everybody else is expected to travel on.