There is much hype surrounding generative AI at the moment, and its increasingly sophisticated ability to create realistic images, including those that mimic artistic styles to depict prompts. The recent release of OpenAI’s GPT4o, and the resultant trend of ‘Ghiblifying’ images using this model, has inevitably thrown fuel onto the already-raging fire of debate surrounding AI and art, and artists’ livelihoods and intellectual property. Ultimately this incredibly powerful and (to many) mysteriously abstract technology is forcing us to revisit questions about what it means to create, and where the frontiers between art and mimicry, human-borne or computer-generated, really lie.

I thought it would be interesting to turn the tables: instead of having a computer generate images of what we want, why not create images of what a computer is ‘doing’? This process can be viewed as the inverse of generative AI, and the hand-off between ‘prompt’ and ‘creation’ is computer to human, not human to computer: