Meta-Design, Figure and Style between Drawing and Generative AI

Abstract

The use of generative AI in design today complements established modes of ideation through Drawing, understood in its traditional sense as “meta-design”. In text-to-image, image-to-image, and image-to-model pipelines, ideation takes the form of a chain of transductions across media and codes (sketch, plate, text/prompt, image, model, ren­der), where each step redefines pertinences and thresholds of “figurability”. Within this framework, the “question of figure” concerns the stabilization of readings: which con­ventions, competences, and procedures turn a trace into a figure that can be shared, discussed, and verified across different actors.
The paper advances a semiotic theory of Drawing as meta-design, conceived as a form of “writing” capable of governing the alternation between an iconic pertinence (recog­nition of objects, scenes, actions) and a plastic pertinence of images (topological, eidetic, chromatic, and textural differences). This distinction allows “style” and “character” to be treated as outcomes of finite grids and Type/Token regimes. In its computational coun­terpart, drawing relies on explicit image description: predominantly “plastic” descriptors (edge statistics, Gabor, Haralick/GLCM, Radon, moments/histograms) build feature vec­tors on which profiles of recurrence and difference can be learned.
By retracing a genealogy of the architectural still life, we exemplify three modalities of abstract figuration - Sironi/Muzio (Monza, 1930), the Purism of Le Corbusier/Ozenfant, and the meta-avant-gardes (Hejduk) - showing how Drawing can operate today as me­ta-design at the junction between traditional ideation and generative AI.