Filling the internet with AI-created images will harm future AIs

The performance of image-generating AIs drops when trained using images previously created by other AIs, which may create problems for them given the proliferation of these images

By Matthew Sparkes

The popularity of text-to-image artificial intelligences could be their own downfall – if the pictures they produce proliferate too much, they could contaminate the data sets that new models are trained on, harming performance.

AI tools like DALL-E 2, Midjourney and Stable Diffusion can create pictures of whatever people request, and their images are increasingly being shared online.

Such AIs improve their results by training on data sets of billions of images and …