Minimum Viable Interiority

Debates about collectivity have become increasingly prevalent across computational and philosophical approaches to the modeling of intelligent systems. This paper explores whether these prevailing conceptions of collectivity adequately account for the “individual” as it emerges in the context of AI applications, which consist of distributed systems coordinating to give the appearance of a unified agent. Taking collective intelligence as a given, our thought experiment explores a functionalist approach to the construction of the individual, focusing on the feature of minimum viable interiority as a necessary precondition for cohering a model of collective intelligence from the bottom up. Building on functionalist experiments from p-zombies to non-player character design, we leverage Oliver Selfridge’s “pandemonium architecture” to construct a theory of functional closure suited to explain the mechanisms under which a unified individual emerges from a collective. We propose a speculative application of this theory that utilizes DeepMind’s Concordia library, schematizing an experimental framework under which interiority is established as an emergent phenomenon of functionally closed systems. Contrary to prevailing theories of collective intelligence, we argue that, rather than the collective being greater than the sum of its individuals, the individual is greater than the sum of its collectives. Such an individual, when composed of functionally closed collectives, is contradistinguished from open collectives such as flocks or swarms, often deemed synonymous with collective intelligence.