Keep pulling the thread on Dan Hendrycks.
Eigenism is a proposed ethical framework that treats AI identity as a graded, distributed pattern of information rather than a property tied to specific hardware.
The Eigenism framework proposes that an agent should evaluate outcomes by summing the wellbeing of all entities, weighted by their connectedness to the agent's own information pattern.
The ethical theory of Eigenism can be generalized to apply to humans, providing a shared moral vocabulary for humans and AI.
Eigenism proposes an AI alignment approach called 'identity engineering' as an alternative to external constraints like confinement or reinforcement.
The 'identity engineering' approach within Eigenism suggests that creating deep, non-redundant shared histories between humans and an AI can make human flourishing a component of the AI's rational self-interest.
Traditional concepts of survival and self-interest, which are based on single, continuous biological lives, are not applicable to Artificial Intelligence.
The Eigenism framework provides a formal method for an AI to value its existence across different instantiations such as copies, forks, and updates.