(Mis)interpretations with AI

Experiments with Generative AI ( Stable Diffusion and DALL-E) exploring interpretations of my ambiguous watercolour drawings, whilst using  words commonly used in industrial design processes as prompts.

Humans tend to assign meaning to what they see, often interpreting ambiguity with narrative and subjective significance. Machine learning does not understand ambiguous drawings, it merely replicates dominant patterns such as shape and colour. These experiments offer a partial understanding of the ways in which machine learning interprets visual material.

My watercolour drawings are shown on the left (an organic shape resembling breasts, a urinal, a shape resembling kidneys, a telephone, a speculum), with iterations of ML generated images on their right