AI BEETLE by Frederik De Wilde
A “Ceci n’est pas une pipe” for AI vision systems.
AI BEETLE is a art/science/technology research project initiated in 2016 in collaboration with Professor Jeff Clune at the Artificial Intelligence Lab of the University of Wyoming.
As an early example of artist-machine co-agency, the project explores the limits of computer vision by using convolutional neural networks and evolutionary algorithms not to improve recognition, but to disrupt it.
Where camouflage once concealed bodies from human sight, AI Beetle proposes a post-digital form of “dazzle” designed to confuse machine-learning classifiers and corrupt metadata systems.
At the centre of the work lies a paradox: AI-generated patterns are used to deceive other AI systems. The sculptural beetles, covered in algorithmically evolved patterns and surfaces, remain visible to human observers while becoming misread, mislabelled, or ignored by computer vision.
An AI Beetle may be identified as a butterfly, an umbrella, or simply noise.
This failure of recognition is not a flaw but the project’s critical force — a reflection on surveillance culture, algorithmic bias, and the fragility of machine knowledge.
In this sense, AI Beetle extends the questions raised by The Treachery of Images and René Magritte’s famous phrase, Ceci n’est pas une pipe. Just as Magritte revealed that an image is never the object it represents, De Wilde exposes the unstable relationship between image, label, and meaning in the age of AI.
What appears to humans as a beetle may be interpreted by a neural network as something entirely different. AI Beetle thus becomes a contemporary “treachery of images,” where the gap between seeing and knowing is no longer only philosophical, but computational.
The project also addresses the politics of metadata and automated targeting.
In a world increasingly governed by machine learning systems, De Wilde asks what happens when those systems can be fooled. Rather than functioning as a neutral technological experiment, AI Beetle becomes a critique of the epistemologies that shape contemporary power. It reveals how artificial intelligence not only sees, but also misreads.
Historically, the work connects to the legacy of World War I dazzle camouflage, a strategy developed to confuse perception rather than hide form. This lineage links AI Beetle to modernist experiments in fragmentation and visual instability, from Cubism to Surrealism, while also resonating with contemporary anti-surveillance practices designed to evade facial recognition and tracking technologies.
The project further raises questions of authorship and co-creation. If images emerge through iterative exchanges between artist and machine, who becomes the author: the human, the algorithm, or the system they produce together? AI Beetle embraces this ambiguity, proposing a hybrid form of creativity in which agency is distributed across human and non-human actors.
Ultimately, AI Beetle is less an object than a proposition: a form of resistance that uses computational systems against themselves.
By weaponising misrecognition, the work exposes the assumptions embedded within AI and challenges the authority of machine vision in contemporary culture.
