Recent advancements in AI algorithms have created networks capable of learning visual concepts from natural language text, as well as generating high resolution images. Pairing these algorithms together, it is possible for a language network to guide the output of a generator networks. In other words we can give the AI system a text description, and it will creat an image based on the text. It works very well for literal descriptions, but we were interested in how it would interpret abstract and surrealist language. The AI system is able to take as input a reference image, as well as a text, and it will morph and edit the image to match the text. Using a series of collages made by Broombreg and a series of quotes from The Book of Disquiet by Fernando Pessoa, we enabled the AI to show us the surreal corners of its latent space, the slippages between language and imagery. Made using the Latent Visions notebook by @advadnoun, which uses the CLIP and VQGAN algorithms. A collaboration between @AdamBroomberg and @IsaacSchaal
