As I was working on the images in the "Imaginary Objects and Texture" project new machine learning applications became available to me. These started to offer more direct control over parameters using text, although at the same time a worrying trend developed in the world, which is that the applications began to more directly accept commands in terms of reproducing the styles of other artists. My images here are not generated by text prompts, they are mostly created by generating multiple AI variants (often hundreds) using my own photos as an input, then compiling the results back together in Photoshop. I am interested in what the AI tools get "wrong" based on the limited instructions I do provide; the more realistic they become, the less work is involved from the artist, so I think it's doubtful I'd use this method in future without very heavy modification of the software. I have left the project here as a record of a new technology that I was excited to explore - and perhaps a reminder of some how quickly technology can develop into ways that are not always positive for the artist.

You may also like

Back to Top