In 1668, John Wilkins published An Essay Towards a Real Character and a Philosophical Language — an attempt to classify every thing in existence and assign each a unique symbol, so that the world could at last be written without ambiguity. The project failed, and Wilkins knew it. Throughout the Essay's taxonomy he confesses, in hedging phrases that accumulate across hundreds of pages: coral is a 'Strange Original; not being properly Minerals'; fungi 'do want or seem to want some of the more essential parts of Plants'; zoophytes are 'betwixt Plants and Animals'; the whale is a fish, except it breeds its young within it and breathes air. There are creatures he admits belong to the wrong category, hybrids he acknowledges he cannot place, things he can only describe as 'either... or' because no single category holds them. A Catalogue of Exceptions dwells in those failures. Each subject — coral, fungi, jellyfish, zoophytes, slime mould, the ancient fossils Wilkins could not have known and could not have classified — is generated through early machine learning models: StyleGAN, early versions of Stable Diffusion, models chosen for their visible artifacts and anatomical instability. The strangeness is double. The organisms that broke Wilkins' taxonomy in 1668 also resist coherent rendering by these models: their ambiguous forms — neither plant nor animal, neither one organism nor many — produce the same quality of taxonomic anxiety in the machine as they produced in the natural philosopher. The coral dissolves at its edges. The jellyfish refuses to resolve. The slime mould flickers between individual and collective. The things that do not fit the category do not fit the model either. Handwritten text appears throughout — Wilkins' own hedging phrases, scanned and layered over the imagery. 'Not properly.' 'Seem to want.' 'Of various shapes and bignesses.' 'Without any Latin or Greek name.' His confessions of failure become the work's only language. The hand that tried to name what the machine now tries to depict; both failing in the same place, 350 years apart. The piece runs as a 24-hour durational cycle, its subjects shifting through the day in loose correspondence with Wilkins' own taxonomic categories: the imperfect plants and zoophytes of morning, the stranger organisms of afternoon, the ancient and unresolvable at night. At the darkest point of the night cycle, the piece reaches its subjects furthest from any taxonomy — Ediacaran biota, Cambrian fossils named 'problematica', organisms for which the formal category is incertae sedis, meaning only: of uncertain placement.
The decision to use early models — StyleGAN, Stable Diffusion 1.x — rather than more capable contemporary systems was conceptual rather than aesthetic. Later models have been trained to suppress failure: to produce coherent anatomy, consistent edges, resolved forms. The earlier models still show their confusion. When asked to generate a jellyfish described as 'a kind of Gelly, roundish at the top, marked with reddish lines in the form of a Starr' — Wilkins' own 1668 description — they produce something that could not be classified as a jellyfish but could not be classified as anything else either. The failure is not incidental; it is the subject. The parallel that emerged in making the work was not something imposed from outside: it is structural. Every system that attempts to classify everything creates a category for what it cannot classify. Wilkins called them 'Strange Originals' and 'Imperfect.' Linnaeus created a category called Chaos for microscopic organisms he could not see clearly enough to name. Haeckel invented a third kingdom — Protista — for everything that was neither plant nor animal. Contemporary biological taxonomy has a formal term for organisms that cannot be placed: incertae sedis. Contemporary machine learning has mode collapse — the failure state in which a model defaults to the same few outputs because it cannot generate the thing being asked for. The dumping ground is always the confession. The catalogue of exceptions is always the most honest part of any system.
Commissioned by the Schwarzman Centre for the Humanities, University of Oxford, as part of A Perfect Language of Images (2026)