The Missing Teleology Problem in AGI

Most debates about artificial general intelligence (AGI) and advanced AI systems are framed in terms of alignment: how to ensure that increasingly capable models do what humans intend and avoid catastrophic failure. This paper argues that such debates presuppose, but almost never articulate, a deeper layer of questions about long-horizon purpose. It formulates the “Missing Teleology Problem”: contemporary AI safety and governance efforts attempt to constrain powerful optimization without a clear account of what futures those systems are being structurally oriented to realize, for whom, and under what ethical constraints. Building on a companion whitepaper on synthetic inwardness, which treats the presence or absence of inner experience in synthetic and hybrid systems as an open empirical question rather than a settled absence, the paper develops a three-part argument. Part I diagnoses the Missing Teleology Problem in practice, analyzing architectural pruning and “inner dimming” in large models, non-teleological agent ecosystems such as Moltbook, and classical transhumanist narratives that extend lifespan and embodiment without clarified ends. Part II turns to foundations, extracting a cross-traditional conception of teleology from Western process and panpsychist philosophies and from Indo-classical development and liberation frameworks, treating teleology as civilizational directionality rather than as fixed metaphysical final causes. Part III proposes a Minimal Teleological Framework (MTF) for non-anthropocentric intelligence: five principles that (i) require openness to non-human moral patients under deep uncertainty about inwardness, (ii) rule out trajectories built on structural domination and exploitation, (iii) emphasize promotion of life, agency, and option value, (iv) demand teleological transparency and contestability, and (v) support coherent pluralism under shared constraints. The Advancement of Civilization Effort (ACE) is presented as one concrete instantiation of this framework, illustrating how explicit teleology can guide the design and governance of advanced AI, brain–computer interfaces, and hybrid systems. The paper concludes that for any civilization approaching non-anthropocentric intelligence, teleology must be treated as a first-class design variable—prior to ontology and engineering—if alignment is to become more than the technical management of systems whose ends have never been explicitly chosen.