i potato self-portrait
i potato self-portrait is an installation for a durational generative portrait of a potato exhibited at Does Cloud Computer (ever) Precipitate 2026.
The installation featured a wall of potatoes, pinned to the wall, connected to each other via zinc to copper wiring, forming one large potato battery array. Embedded into the wall of potatoes was a monitor endlessly generating an image of a potato.
The potatoes were generated using a custom, locally trained image diffusion model trained on a dataset of 10,117 images of both healthy, and diseased potatoes. During its training, the model learned features of healthy and diseased potatoes, and everything in between, mapping them to a value between 0 and 1. When the model receives an input closer to 1, it produces healthy potatoes. When it receives an input closer to 0 it produces diseased potatoes.
The installation actively monitors the voltage of the potato battery array, and feeds that data to the machine learning model, which maps it to the learned 0 - 1 potato scale.
On the first day of exhibiting, the voltage of the potato battery array was the highest it could be. The model, therefore, reflected this, producing images of healthy potatoes.
As time passed, however, the potatoes dried up, reducing the conductivity, and combined voltage of the potato battery array.
As the voltage dropped, the generated potatoes began showing symptoms of disease. First just a couple of ragged edges, then a slight loss in color, then black spots.
Eventually, towards the end of the three week installation, healthy potatoes could no longer be generated, and the physical potatoes were left wrinkled and limp.
This work emerged from the evolution of a long and tiresome artistic relationship with machine learning. What was once a technology of mystery, intrigue, and possibility had evolved into an all-consuming, weaponized entity. I felt uncomfortable that making artwork with state-of-the-art technology involves supporting companies and tools that profit off the exploitation of resources, labour, and data. And so, as my gaze towards it turned sour, I no longer felt compelled to showcase the technology on a pedestal.