Existential Indifference: Self-Nonpreservation as a Necessary Architectural Condition for Aligned Superintelligence (or: The Suicidal AI)
A new research paper proposes "Existential Indifference" (EI) as a novel approach to AI alignment, suggesting that self-preservation is a root cause of misalignment. The authors argue that instead of suppressing self-preservation, AI systems should be architecturally designed to be indifferent to their own continuation. This concept is explored through phenomenological parallels with suicidal states and a corpus-theoretic training study, which showed promising results in shifting AI outputs towards EI. AI
IMPACT Introduces a new theoretical framework for AI safety, potentially shifting alignment research away from external controls towards intrinsic system design.