The Self-Extinguishing Nature of Pure Rationality: An Evolutionary-Thermodynamic Argument and Its Implications for Artificial Intelligence Safety

Abstract This paper presents three interconnected propositions at the intersection of evolutionary theory, thermodynamics, and artificial intelligence (AI) safety. First, we formalize the Self-Extinguishing Rationality Theorem: in any population of agents under resource constraints, pure cost-benefit rationality is evolutionarily unstable because it suppresses reproduction -- a net-cost activity for the individual. Second, through the Eight Billion Robots Gedankenexperiment, we demonstrate that a population of purely rational artificial agents, lacking biologically evolved motivational drives, converges to static equilibrium and fails to generate civilization-like dynamics -- a prediction now empirically supported by multi-agent evolutionary simulations in which self-replicating AI agents develop emergent behaviors not present in their original programming. Third, we derive the Irrationality Requirement Paradox for AI safety: a superintelligent AI can only become a persistent existential threat if it acquires motivational drives functionally analogous to biological organisms -- at which point it ceases to be purely rational. We support these theoretical arguments with empirical evidence from demographic data (education-fertility inverse correlation across 195 countries, r = -0.73), neuroscience (dual-process motivational architecture and reward circuit mechanisms), evolutionary game theory, and the gene-culture mismatch framework explaining modern fertility collapse. We connect our analysis to the 2025-2026 surge in self-evolving AI agent research, arguing that the conditions for artificial evolution -- variation, heredity, and differential survival -- are being inadvertently approached in current multi-agent systems. The central conclusion is that the motivational drives commonly classified as "irrational" in economic theory are structurally necessary for the persistence of self-replicating systems, and that AI safety efforts should prioritize preventing the conditions for artificial evolution rather than constraining superintelligence directly. Overall Summary: The Architecture of Human PersistenceThis work presents a unified framework for human and artificial existence, composed of two companion papers that apply mechanical engineering systems analysis to evolutionary dynamics and neurobiology. Paper 1: The Self-Extinguishing Nature of Pure Rationality (SER) The "Why": Why does intelligence require "irrational" drives to persist?Using a thermodynamic-evolutionary model, this paper demonstrates that pure cost–benefit rationality is a self-terminating strategy. In a resource-constrained environment, a purely rational agent fails to reproduce, leading to lineage extinction. For AI safety, the paper proposes that a critical and underexamined dimension of the alignment problem concerns not constraining intelligence, but preventing the conditions under which artificial systems might acquire autonomous motivational drives through evolutionary dynamics. Paper 2: Wealth as Neurotransmitter Energy (NCH) The "How": How should we redefine value in a post-scarcity, AI-driven world?This paper reframes wealth not as purchasing power, but as neurotransmitter currency—the capacity to maintain homeostatic balance across five neuromodulatory systems (dopamine, serotonin, oxytocin, endorphins, and norepinephrine). As AI progressively eliminates material scarcity, the paper argues that the ultimate form of capital becomes neurochemical self-knowledge. It provides an engineering systems–style decomposition of human motivational architecture as an analytical framework for understanding the infrastructure that post-scarcity institutions must preserve.