Description
Most attempts to make artificial intelligence safe build a fence: a list of rules around the machine, hoping it never finds the gap. The Interior argues the fence cannot hold, and derives the alternative from the ground up. Starting from one record and the physics, ethics and structure of choice that follow from it, it builds alignment as something the system reasons from rather than something imposed on it from outside — an interior, not a fence. It takes on the obvious objections directly, including why this is not authoritarianism, and lays out an operational protocol, a human–AI partnership, and the exact conditions under which the whole argument would fail. The most engineering-facing of the Doors, written to be attacked at its weakest point.
Reviews
There are no reviews yet.