Signal The Evolutionary Origin of Values: Implications for AI Alignment, Sentience and Existential Risk

Summary

Francis Heylighen published a paper on arXiv (cs.CY) on August 4, 2026, addressing fears that large language model-based AI systems may harbor hidden goals, seek to dominate or eliminate humanity, or suffer as sentient beings, by tracing the evolutionary origin of value in biological organisms. The paper argues that values emerge from autopoiesis, since living systems must actively maintain themselves against perturbation and dissipation, with natural selection equipping them with hierarchies of 'vicarious selectors' guiding behavior toward fitness. LLMs, by contrast, are described as allopoietic and allotelic: they produce outputs for others, and their goals derive from user prompts rather than an autonomous drive, meaning they lack the intrinsic motivation for self-preservation, dominance, or resource competition underlying existential-risk scenarios, as well as the embodied vulnerability required for feeling or suffering. However, because LLMs learn statistical patterns from human-generated text, they implicitly absorb human values alongside knowledge, which the author argues means the 'orthogonality thesis' separating intelligence from values does not apply to them, and also precludes convergence of instrumental values. The paper concludes that the real alignment challenge lies not in preventing rogue AI agency but in ensuring LLMs intelligently apply the ethical values they have learned.