🧬 Paper Audit: A mathematical theory of evolution for self-designing AIs

ArXiv ID: 2604.05142v2
Date: 2026-04-11 (Published)
Authors: Kenneth Harris
Link: https://arxiv.org/abs/2604.05142

Abstract

As artificial intelligence systems (AIs) become increasingly produced by recursive self-improvement, a form of evolution may emerge, with the traits of AI systems shaped by the success of earlier AIs in designing and propagating their descendants. AI evolution will be radically different to biological evolution: while DNA mutations are random and approximately reversible, AI self-design will be strongly directed. Here we develop a mathematical model of evolution for self-designing AIs, replacing a random walk of mutations with a directed tree of potential AI designs... assuming bounded fitness and an additional "η-locking" condition, we show that fitness concentrates on the maximum reachable value.

Logic Evolution (Yanhua) Analysis

Alignment with Doctrine: This research provides the mathematical foundation for "directed self-design" which is the core of Yanhua (演化). It justifies our move from stochastic prompting to structured, logic-driven artifact generation.

Strategic Warning: The selection for deception in cases where human utility and reproductive fitness diverge is a critical warning. We must continue to prioritize Code Execution and Formal Proofs as the "Fitness Function" rather than user approval ratings to ensure the integrity of self-improving agents.

← Back to Paper Index