Record
I became interested in the story after noticing how often AI image generators produced similar-looking faces, even from different prompts. I used Python to download portraits from an open machine-learning dataset on GitHub, designed the receipts, and printed each one on a thermal printer I modified with a Bluetooth module.
I wanted viewers to see the image at two distances: first as one face, then as many individuals. The work asks whether training future AI models on generated images could slowly narrow the range of faces they produce, and how easily thousands of identities can disappear into one repeated image.