Autopoiesis Without Life? Rethinking AI Through Cybernetics and Systems Theory

Drawing on cybernetics and systems theory, this paper rethinks AI through the lens of autopoiesis. It traces a line from feedback control to second-order cybernetics, showing how machine learning (ML) inherits feedback, estimation, and decision while remaining allopoietic. Autopoietic living systems self-produce their components; ML systems produce outputs for external use and depend on human goals, data, and infrastructure.Yet, when embedded in social communication (Luhmann), generative models perturb autopoietic reproduction and yield quasi-autopoietic effects.The article surveys recursive AI training, info-autopoiesis, embodied/chemical AI, and human-AI hybrids, and argues for design and governance while preserving ontological distinction between life and machine.