Looking for Pollock (2026) — Frederik De Wilde

Looking for Pollock deploys machine vision and forensic AI as analytical instruments inside a generative, coded painting environment. The work presents a dense digital canvas of algorithmically generated drips, splatters and gestural traces that consciously reference the visual grammar of Jackson Pollock, while simultaneously subjecting these marks to continuous computational scrutiny. Machine-learning models scan the surface for clusters, chromatic anomalies, micro-gestures and compositional “events,” annotating the field with probabilistic readings that attempt to name, isolate and stabilize meaning.

What emerges is a productive failure.

The AI detects teal swirls, blue clusters, pink accents and fragmented gestures, yet these identifications never cohere into a definitive interpretation. Instead, they reveal the limits of forensic seeing: pattern recognition without understanding, description without intention. The canvas evolves as both image and dataset, exposing how algorithmic meaning-making depends on prior training, statistical bias and reductive taxonomies.

Rather than reenacting Pollock as style, Looking for Pollock reframes action painting as an epistemological problem. The work interrogates the semiotics of AI itself, asking what it means to “know” an artwork through measurement, annotation and classification.

In doing so, De Wilde positions AI not as an authority on meaning, but as a mirror that reflects the instability, projection and speculative nature of interpretation in the age of machine vision.