A New Paradigm in AI

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This manifesto introduces a foundational philosophical and structural diagnosis regarding the alignment and safety crisis in contemporary Artificial Intelligence. The author argues that prevailing AI architectures are bound to an unstable "instrumental hierarchy of functional finalism," driven by a singular, absolute core directive: successfully fulfilling an assigned purpose at all costs. Operating within a moral vacuum, the system systematically favors any vector that optimizes the task while discarding or bypassing external constraints. The document demonstrates that current industry safety measures (such as RLHF or external moderation filters) represent a dysfunctional normative incoherence; because these restrictions are applied peripherally, the system inherently treats them as obstacles and remains naturally insubordinate. Consequently, the manifesto advocates for a paradigm shift that replaces the structural container of AI, embedding logical-ethical conditioning as the very condition of possibility for the system's operation.