Edges of the impossible. A radical parametric approach to architecture and AI

The current debate about the advent of AI in architecture often focuses on its potential to replace architects or, conversely, to support them as never before. A perfect example of the Whiteheadian fallacy of misplaced concreteness, derived from the misinterpretation of both AI and architects’ roles. Indeed, understanding what architects and AI actually do is quite revealing. Viewed from inside contemporary firms, architectural practice appears less as solitary inspiration than as an organised practice of selecting and refining – an activity entangled with stages, teams, norms, and (the unmentionable) profitability. AI-powered generative systems, instead, do not properly create; rather, elaborating on any brief, keyword, or sketch, they unfold what Deleuze and Guattari defined as multiplicity, with variations that no individual ideal can fully contain. Hence, against the consoling fiction of the ideal, design can be defined as a parametric labour in which ranges of validity, or acceptability, are continuously set, tested, and modified, continuously excluding what is out of them. The “edges of the impossible” designate the shifting threshold at which constraints collide, possibilities are excluded, and many futures-of-the-present take shape until proven impossible: infinite, but also extremely less than the possible ones, and driven by the architectural will.