Shizhe Qian

In "Diffusion," I delve into the intersection of memory and technology, merging machine learning, personal history, and photography to explore the fluidity of recollection. Through a diffusion model, I have trained an AI on my family albums and hometown photos. The model disperses the information within the photograph, introducing noise until the image blends indistinguishably into randomness. It ultimately reconstructs new imagery by denoising based on visual patterns learned from the dataset. In this process, photographs no longer possess fixed meanings; instead, they become malleable data that allows reinterpretation.

Guided by my drawings, the AI creates abstract visuals that blend my current reflections on the past with archival memories. These giants symbolize the influential figures in my childhood, who once shaped who I was. By transforming them into tapestries, those digital fragments of memory behind the giants are deconstructed into materials and woven into tangible expressions. Yet, paradoxically, these tapestries—traditionally used to preserve narratives—are crafted by textile machines following predetermined programs. This natural and organic appearance is, nonetheless, achieved through technology which both enhances and distorts our understanding of the past.

The diffusion process, with its erosion and reconstruction of images, mirrors my journey of reimagining the past. "Diffusion" reflects on our methods of capturing and remembering the past and also invites you to reconsider how we can reshape our historically constructed narratives.