Tele-Ontogenetic Production: Human-Directed Originality in Relational Authorship

Elly’s doctoral research introduces the concept of relational authorship as a framework for understanding production in, and beyond, the discipline of architecture. It examines how machine learning—including large language models (LLMs) and multimodal models—participates in creative production, and the socio-material conditions under which these models might participate in authorship. This talk develops one such condition at the interface between human and model, proposing the concept of tele-ontogenesis to describe a human-directed pursuit of novelty through contingent processes of production. Contrasting tele-ontogenesis with tele-objectivity (Virilio; Rouvroy and Berns), the talk examines how computational systems can be oriented either towards predicting and optimising future states from existing data or towards cultivating contingency to bring new possibilities into existence. Drawing on Henri Bergson’s account of invention, it distinguishes between a future rendered predictable through existing patterns and one that emerges through a process that cannot be fully anticipated in advance. This distinction locates originality not in the formal novelty of a finished artefact, but in the process through which something new comes into being. The talk argues that this process remains human-directed. While authorship may emerge through relational assemblages of human and non-human actors, the telos of novelty is established through human direction. Rather than requiring complete control over the outcome, this direction operates through the configuration of computational conditions, the cultivation or suppression of contingency, and iterative engagement with unexpected results. Tele-ontogenesis therefore offers a framework for understanding originality as a processual and relational pursuit of novelty. It asks how machine learning might participate in invention without reducing creative production to the reproduction, prediction, or optimisation of what already exists, and how human authorship might be understood through the direction of processes whose outcomes remain open.