Plastic Chairs
Produced with one of the earliest public versions of DALL-E, a neural network for text-to-image generation that arrived with immense expectations about the future of visual production. For this series the system was tasked with a seemingly trivial request: to reproduce the plastic monobloc chair, an object so common it has become almost invisible in daily life. The results, however, subvert the promise of fidelity. The chairs appear uneven and inconsistent, combining traits of photographs and paintings in ways that often collapse into crude and gestural constructions. In this sense the project functions as a test of the system’s capacities, in the same way that the malformed hands of early models became markers of their failure to convincingly depict the human figure. The work functions as an early document of how generative systems negotiated ubiquity, exposing both their technical limitations and accidental expressions. Today such outcomes have become impossible to reproduce, as current models have largely resolved representation, producing an uncanny standard of image generation while losing the expressive instability that marked their earliest stages.
