The uncanny art inspired by evolution and generated by ‘crossbreeding’ images

When viewing the work of the US artist and tool developer Joel Simon, you might find that his images are unlike anything you’ve ever seen before – which makes sense, as they weren’t quite born of nature or a human mind. Inspired by the biological properties of evolution and emergence, Simon uses simple programming rules, which, when applied over and over, give rise to uncanny images that augment the human imagination. His most recent work, explored in this video from Science Friday, applies a machine-learning framework known as a generative adversarial network (GAN) to two images. Guided by human users via Simon’s website Artbreeder, his programs ‘crossbreed’ pictures of everything from animals to artworks. Fascinating digital artefacts in their own right, the resulting, author-less images raise complex questions at the nexus of art, programming and design.