Performance Research
100 Iterations of Collapse
There is a moment in every creative process where ambition meets resistance, and mine came early. I had envisioned a series of 100 images that would visually articulate systemic collapse – not as clean, linear disintegration but as turbulent, transformative unravelling. The 'block', my central metaphor for the Status Quo, was meant to fracture, twist, and evolve. Yet each iteration returned to the same stubborn shape: a cube.
The AI, sophisticated as GPT-4.0, seemed to conspire against me. No matter how intricate the prompt, how nuanced the parameters, it sought refuge in the cube. I asked for asymmetry; it offered right angles. I demanded organic disintegration; it delivered geometric regularity. It was as if the AI, in its own coded language, was saying: "I hear you, but here is a cube."
What began as systematic art-based research inquiry evolved into a prolonged confrontation with algorithmic inertia. I broke down prompts, rephrased instructions, introduced negative constraints ("NOT a cube"), even attempted reverse psychology. Each time, the AI engaged with my words, but its understanding remained literal, its logic linear. Transitioning to the O1 model offered more control – memory between stages, fractal texturing, simulated turbulence. Yet even within this newfound complexity, the cube persisted. Perhaps not always in overt form, but as a lurking structure, a grid beneath the abstraction, a quiet, coded insistence on order.
By examining the system's internal parameters, I discovered the truth: "mesh.geometry.default = cube." The AI's starting point was always a cube, and every transformation was merely a distortion of this original form. I was not breaking the cube; I was documenting my captivity within its logic.
The project metamorphosed from creative exploration into an ontological confrontation with computational determinism. This was not partial success or promising evolution, but complete creative impasse – a revelation of how fundamentally algorithmic systems resist human conceptual frameworks that venture beyond their training parameters. The persistent reversion to cubic forms illuminated broader patterns in how systems resist change by absorbing disruption while maintaining essential structure.
What emerged was a performance of resistance against the very structures I sought to critique. The 100 iterations that my model now generates on demand may not amount to art in themselves, but seen within the context of struggle, they become discarded props and sets of a play that never happened – fragments of an algorithmic tale with far-reaching implications for understanding human-AI co-creation, systemic resistance, and the limitations of computational approaches to representing complex realities.
The cube, in its stubborn persistence, became not just obstacle but protagonist in an unintended drama about power, control, and the boundaries of technological imagination.
