Ryan Thompson is a Canadian visual artist who uses machine learning and custom-written code, alongside a background in printmaking and photography, to explore ideas he finds hard to grasp through everyday language. He describes his generative approach as one where he “whittles down a concept until he feels it may be readable without language”.
In 2020 he trained a GPT-2 language model on more than 30 years of dream diary entries kept by a Vietnam War veteran, using the output to produce the text and portrait images of GPTSD, a project examining whether trauma can pass from human to machine and back again.The following year he began Dead Precedence, a series of algorithmic portraits of American presidents built by layering each president's image over all those who came before him, with the number of painting iterations tied to the years since that president's death.
Since 2020 he has continued Adversary, an ongoing series using a generative adversarial network trained on publicly available photographs to generate images of imagined opponents, beginning with a body of work centred on boxers.For his 2024 exhibition “Mirror, Mirror.” he built H.U.M.A.N., a sound sculpture that feeds thousands of his own past song lyrics through a Markov chain and then a large language model, voicing the results through a speaker built from an old hot water tank.
