Pareidolia in the Machine: Unintended Figuration in AI-Generated Video from Abstract Visual Music

This pictorial explores the interplay between algorithmic abstraction and AI-driven figuration in generative visual art. Using an AI video synthesis model, the authors transformed abstract visual music frames—created with the Processing programming language—into dynamic video sequences, aiming to preserve their non-representational aesthetic. However, the AI often reinterpreted geometric and fluid forms as representational entities, such as human figures, birds, and insects, despite the absence of such subjects in the source material. This phenomenon reveals two key dynamics: technical bias, in which AI, trained on representational datasets, prioritizes figuration and mirrors human pareidolia; and creative negotiation, where the artist’s abstract intent is continually reinterpreted, raising questions about authorship and the boundaries of procedural art. This study prompts deeper inquiry into whether natural language is inherently sufficient to evoke abstract aesthetics—a fundamental question that merits further exploration.