Mentes Extractae is a hybrid post digital artwork combining oil painting, symbolic ciphers, and web linked systems. I view trees as witnesses to human cruelty and failings. Through driftwood, brain fragments, and AI vectors, the work questions the harvesting of human knowledge and creativity without consent.
An LLM does not read words as complete units. It dissects language into tokens, transforming fragments of meaning into vectors that exist within vast multidimensional embedding spaces. This companion digital site extends the artwork beyond the physical object, allowing viewers to interact with hidden ciphers, explore simplified vector spaces, and contribute their own descriptive words into the system. In doing so, viewers temporarily participate in the same processes of translation, abstraction, and computational interpretation that increasingly shape human knowledge, memory, and creativity.
Each token is a concept in disguise · click to decode
Tokenization simulated · Embeddings pre-computed via text-embedding-3-small (1536 dims, first 32 shown)
By placing the work inside her own public GitHub repository — a widely used online platform where programmers store, share, and collaborate on software code — Isabelle intentionally inserts the artwork into the very ecosystem that powers contemporary AI and software development. Like a Trojan horse, the project quietly infiltrates the world of developers and technologists from within, confronting fullstack developers with questions surrounding authorship, extraction, consent, and the growing displacement and appropriation of human creativity, their technical skills by the systems being built to emulate them.