Questions
Common questions.
What is Value Alignment?
Value alignment is the challenge of ensuring an AI agent's actions are consistent with the values and preferences of the humans it is meant to serve, not just technically correct but substantively beneficial.
How does Value Alignment work?
It is broader than goal specification and includes handling value uncertainty, preference learning, and conflicts between different stakeholders' values.
Which terms are related to Value Alignment?
Closely related concepts include Corrigibility, Spec-Driven Development, Agent Owner, Capability Discovery. Each is defined in the Prefactor glossary.
Keep reading