Back to publications
Wu-Wei in the Machine: Open-Ended Learning in Goal-Free Generative Agent Societies
Paper Details
Published: 2025/08/31
Container Title: The 7th International Workshop on Intrinsically Motivated Open-ended Learning (IMOL 2025)
Paper Links
HTMLTraditional AI systems optimise explicit objectives, but recent work shows that rich behaviour can emerge from intrinsic drives alone. This paper proposes a framework for goal-free generative agent societies where agents have no extrinsic rewards. Instead, agents are driven by curiosity. Generative agent simulations have suggested that even in the absence of external goals, agents self-organize into social patterns: for example, coordinating events or innovating tool-use. Philosophically, this process echoes Taoist “wu-wei” (action through non-action). Intelligence unfolds not by forceful optimisation but through an effortless, open-ended flow. We discuss how such systems could learn alongside humans in a shared, open-ended environment, akin to gardeners tending a world of unfolding novelty.
