The Conjuring of
Failures in
War Machines

This project begins with a miscommunication, a productive failure between human intent and machine cognition. Working with an obsolete GAN-based image generation model originally trained to produce only landscapes, I prompt it to imagine weaponry: fighter jets, drones, and other instruments of war. The model, unable to synthesize these shapes correctly, produces images that are warped and incomplete. Jets with missing wings, cockpit-glass spilling into dunes, distorted human-weapon hybrids, these elements are cut out and recomposed into surreal digital collages. The world they occupy is no longer a representation of the original prompts, of weaponry and violence, but instead, a hallucination of their failure.

 

Through the knowledge gap between a model’s training data and output, this work engages critically with the contemporary integration of AI into military technology. What happens when we miscommunicate with machines? What are the implications when AI models’ perception interface ceases to translate and diverges from reality?