AI: Software learns to create images from written descriptions by destroying data

A fresh approach to generating images based on text descriptions with AI, called a diffusion model, effectively un-destroys new images into existence

By Matthew Sparkes

Early last year, artificial intelligence company OpenAI unveiled software with the surprising ability to create accurate images from text captions – even obscure inventions such as “an armchair in the shape of an avocado”. The company has now released a new AI model that is smaller but capable of producing even better results.

Last year’s program – called DALL-E – was a large, 12-billion-parameter AI model that was trained on a huge set of images with associated captions. In recent years, most progress in AI has come from this sort of approach; training them …