David Weinbaum & Viktoras Veitas — Open-Ended Intelligence: The individuation of intelligent agents (2015)
Weinbaum and Veitas argue that dominant AI paradigms model intelligence as goal-directed, bounded, and individuated — a system that solves a defined problem and stops. They propose open-ended intelligence (OEI) as an alternative: a distributed population of heterogeneous agents that achieves “progressively higher levels of coordination” through reciprocal determination, without reference to a predetermined goal or terminal state. In OEI, identity itself is metastable and provisional — agents individuate through their interactions rather than pre-existing them.
The key contrast is between intelligence as computation (reducing a problem to a solution) and intelligence as individuation (transforming through encounter with the environment and other agents). OEI systems are inherently open: their “problems” emerge from their interactions rather than being specified in advance, and their “solutions” are new configurations of agents that themselves generate new problems.
For OM: OEI provides the cognitive science framework for the claim that protocol undergrounds function as distributed collective intelligence — not because they solve pre-defined problems, but because they probe latent problem spaces through stigmergic, anonymous, viral processes of open-ended individuation. The meta-protocol values of the underground (high agency, consent, open aesthetics) are “computational ethics” — attributes of efficient OEI exploration.