Roni KaufmanInspired by the Dada movement, from Marcel Duchamp’s L.H.O.O.Q. to Hannah Höch’s collages, Portrait Robot is a parody of current “generative AI” programs.
Thread 👇
To generate a new portrait, the algorithm chooses 5 artworks from the Finnish National Gallery's collection. One for the background, one for the base figure, one for each eye and one for the mouth. They're then assembled into a final, intentionally grotesque, collage.
Portrait Robot blurs the lines between digital art and craft by assembling the different shapes as if they had been cut with scissors and glued together. Against the machine's coldness, the “hand” of the algorithm is made visible, embracing human imperfection and playfulness.
While the datasets used to train "AI" models are often opaque and controversial, the source images here are copyright-free, the elements are clearly separated, and this information is both inscribed in the token’s metadata and displayed in the browser console.
The dataset selection was carefully made by me, browsing the collection website by keywords and artists, choosing a few hundred artworks with a predetermined set of criteria. For instance, subjects in portraits had to be facing broadly forward.
Who is represented in the Finnish National Gallery’s collection, and how? In Portrait Robot, monarchs are mixed with mythological figures, scientists, artists and anonymous people. In the background, we see landscapes, interiors and architectures from all around the world.
The result is a sort of facial composite (“portrait-robot” in French) of the Finnish National Gallery’s portraits.
Oh and for context: this project was created for the Combine24, a generative art competition organized by the Finnish National Gallery.
It was selected as part of the 10 shortlist works that will take part in the competition exhibition in Helsinki this autumn and compete for the competition prizes.
You can see all other finalists and vote for the Audience Favourite prize (until October 1st) here: https://combine24.alusta.art/
