The creative possibilities opened up by machine learning and the implications of generating new original content from an algorithm - which has learnt from vast amounts of data using generative neural networks - seem conceptually extraordinary. Early experiments included generating images of electric sheep, clouds in skies, and tricking a censoring algorithm into producing synthetic pornography. I went on to ask how far I could push an algorithm trained to generate humanly recognisable images into abstracted and unpredictable outcomes (Latent Space 2017), setting off language and image generating models to have conversations (Closed Loop 2017), and taking neural birds back into a natural landscape (Cusp 2019).
This rapidly developing field raises many issues, from old philosophical, political and ethical questions to creative dilemmas. What happens to artistic agency when working with artificial intelligence? What does it mean to collaborate creatively with an unpredictable algorithm, learning from datasets unsupervised, and how does this fit into a history of conceptual and systems art?