Coherent extrapolated volition

Abstract

This article posits coherent extrapolated volition (CEV) as a solution to the problem of designing Friendly Artificial Intelligence (FAI) systems. CEV is a mechanism for extrapolating the wishes, values, and preferences of humankind as a whole, in order to guide the actions of FAI. The article argues that CEV has several advantages over other approaches to FAI design, including its ability to take into account the wishes of future generations, its flexibility in adapting to changing circumstances, and its potential to avoid undesirable outcomes such as AI takeover or the destruction of humankind. CEV is based on the idea of extrapolating human volition, or what people would want if they knew more, thought faster, were more the people they wished they were, and had more time to think about the issue. Additionally, CEV incorporates auxiliary dynamics that govern the desired rules, degree of influence and renormalization of the system. It is emphasized that CEV is a dynamic that passes through, and extrapolates, the cognitive complexity of human minds, with all their quirks and variations, to produce a unified output that can be used to design FAI systems that are aligned with human values. – AI-generated abstract