Practices of Prediction: Chalk (2025)
LLMs work via many processes of opaque prediction. However, via clean interfaces these processes of prediction become reduced to services which promise objective, universal and effortless knowledge production. One of these processes, is turning words into vectors. Within a LLM, to become predictable our inputs must become computable. This is achieved via processes of deep learning in which the meanings of words and their semantic distances and associations become fixed, certain. It is this creation of a latent space that determines what can and cannot be predicted or generated.
Live Performance Documentation, Academy of Fine Arts, Munich
In ‘Practices of Prediction’, audiences were invited to re-negotiate and probe these fixed associations and meanings prescribed by LLMs. Throughout the live action performance, we embedded and re-embedded our collective future though the shuffling, drawing and reading of tarot cards.
Live Performance Documentation, Academy of Fine Arts, Munich
Video excerpt, Live Performance Documentation, Academy of Fine Arts, Munich