An artificial intelligence model continuously improves a masterpiece: Rembrandt’s The Night Watch. Or, rather, the faces that Rembrandt painted in 1642 are updated with the AI’s output of what it considers to be a masterpiece.
The Pixel Perfect algorithm operates by taking an image of a painting as input, detecting its faces, and replacing them using image diffusion (SDXL). This process updates, improves, and distorts the faces until they’re no longer recognisable.
Is the AI's pixelated output – the result of thousands of hours of servers training on millions of images – an actual improvement over the painter’s brush, or a sloppy and degraded product sprinkled with gooey algorithmic bias?
Are we watching an intelligent metabolism reslurping reality into perfection right in front of our eyes, or are we being irresponsible by letting our historical legacy be rewritten into new meanings of ‘masterpiece’?
In collaboration with Studio Metaform