Sonia Cook-Broen

Memory<>Machine, interactive installation. Third annual exhibition of work at the intersection of art, hardware, and software.

A custom GAN trained solely on the artist’s own paintings synthesizes a self-portrait that returns the viewer’s gaze. Lifting the headphones beneath the screen is the act of observation: the portrait degrades backwards through the network’s training checkpoints — the machine unlearning the face — while an original drone score composed on analog synthesizers plays. Setting them down lets the portrait re-evolve toward coherence. Two LCD panels render an exchange between poet and machine; the upper frame carries forward, reverse, and kaleidoscopic views. A GAN was chosen as the medium because it requires no external training data: the model’s memory is built entirely from the artist’s own paintings.

Image-authoring and provenance platform realizing the Four Corners paradigm, binding context, sources, backstory, and authorship to the photographic image. Built with the paradigm’s originator and a working photojournalist; founding partner Writing With Light, the authenticity movement in nonfiction photography.

In progress: a project with a photojournalist tracing Frederick Douglass through a lineage of moral courage, anchored to America-250.

AI-mediated video poetry exploring perception and translation. Custom LoRA models trained on the artist’s own paintings drive a poetry-to-animation pipeline. 110,000+ viewers.

Federates IIIF collections from the Smithsonian, the Metropolitan Museum of Art, Europeana, NASA, and Earth Engine into one surface. Browsing runs on a Latin thematic ontology — Aqua, Terra, Flora, Lux, Corpus, Materia — thirty-six categories that cut across institutional catalogues, so a search proceeds by subject rather than by collection. Semantic search, computer-vision color-palette extraction, and deep-zoom viewing on the IIIF Image and Presentation APIs, with each object’s complete manifest rendered in a custom viewer and shareable by link — the interoperability record travelling with the object rather than staying behind at the institution.

Maps the Divine Comedy through spatial navigation, cross-referenced annotation, and interactive 3D literary visualization, with search across five centuries of scholarship.

Built for the research paper of the same subject. Roughly 40,000 subjective experience reports from the Erowid Experience Vault yield 21,591 orientation passages across thirty substances; language-model extraction and semantic embedding, HDBSCAN with Ward-linkage hierarchical clustering, resolved into a sixteen-category taxonomy the visualization makes navigable. Built by TheTechMargin in D3.js and p5.js; co-author on the accompanying paper.

Developed from more than fifty hours of practitioner interviews conducted by the artist. Seed funded.

A digital photobook prototype, named for Tokyo’s used-book district, grown into a book viewer for art collections. IIIF manifests from Europeana, the Rijksmuseum and other institutions are rendered as art books with four reading modes — trailer flip, browse, focused read, and deep study — page turns calibrated to each pace, and filmstrip navigation across multi-book libraries. The pace you read at changes what you notice.

A Fractint-era archive of mathematical visualizations — Sierpinski, Mandelbrot, Apollonian, Pythagoras, Mondrian plotters — paired with live JavaScript recreations that run in the browser, a compose tool for new pieces, and a DOS-terminal aesthetic with keyboard navigation. Old equations, still alive.

Turns a live camera feed or video into a 3D wireframe mesh in real time — what you look like, rebuilt as geometry. WebGL in the browser, no plugins.

American identity, consumer culture, and found-object narrative.

Interior landscapes; domestic space as psychological terrain.

Electronic composition, poetry-to-sound translation, generative audio systems, Web Audio API.

Noah, S., Saiber, A., Cook-Broen, S., Erowid, E., Erowid, F., Yaden, D. B., & Silver, M. A.

Critical writing on AI, creativity, perception, and the interface between human and machine intelligence. 7,500+ weekly subscribers.