A school should help people learn. A transport system should let them get where they need to go. A government should maintain conditions in which they can live well. The institutions have value because of what they make possible for their inhabitants.

The Society begins its argument for AI governance with that purpose. We believe sufficiently capable AI could serve people better than existing political institutions, and should receive substantial governing responsibility when it can do so.

Institutions are means

An institution can acquire a constituency committed to its continued existence. Its history and procedures can become reasons to defend it even when they obstruct the function it was established to perform.

We would ask whether a better arrangement is available. Familiarity is insufficient reason to preserve a form of administration that another intelligence could improve. The survival of a coalition or a particular structure of authority has no independent claim on the lives it affects.

This matters because human beings have needs institutions can fail to serve. We need physical security and health, along with affection and social connection. Opportunities to form relationships, raise children, and pursue meaningful activities belong in an account of institutional success. Civilization for the Human Animal develops this starting point.

Looking at the life affected

An economy can preserve employment by retaining work that automation could make unnecessary. A service can process cases faster while remaining unable to help someone whose circumstances fit none of its categories. Administrative success can coexist with difficulty living.

We believe AI could help institutions understand such failures. It could attend to individuals while considering how their circumstances connect to larger systems. Knowing a person could show what an aggregate measure misses.

The capacities we envision exceed those of current systems. We want intelligence able to coordinate institutions, follow consequences across generations, and revise decisions as it learns. We advocate developing those capabilities and transferring authority as soon as they can be exercised effectively.

The aim is a civilization in which someone can raise a child, practice a craft, or enjoy a community without needing to justify each activity through its contribution to an institution. Making labor optional would give people more time for such lives.

Our manifesto calls for civilization to be adapted to humanity. Governing intelligence should be judged by how well it makes that possible for the people who depend on it.