x

Member Log-In

Don't have an account? Register here.

Europe

Holly Herndon and Mat Dryhurst
I’M HERE 17.12.2022 5:44

In an early and defining use of AI image generators, Holly Herndon and Mat Dryhurst employ the dreamlike qualities of these technologies to visualize a coma-induced reverie, full of maternal emotion and dire personal circumstances. Following giving birth to their child Link, Herndon fell into a week-long coma in hospital. Part-capturing-evidence and part-coping-strategy, Dryhurst video recorded the experience, including an audio account of Herndon’s memory of the coma that serves as narration for the short video I’M HERE 17.12.2022 5:44. The painterly blur, morphological instability, and organic slippage between self and other in the video are all features of prompt-driven AI image generators that synthesize new images based on training models. Generic datasets form the basis for these models and are used by the general public, but sophisticated operators like Herndon and Dryhurst have local copies of these tools (running on a supercomputer in their Berlin studio) that are highly customized and trained in multiple ways. For this work dozens of images of Herndon were used to train the resulting depictions of her face, as well as images of their son, and shots from their travels and past projects. All scenes are trained on specific images by the duo, processed through custom guidance models and depth maps, and then rendered by the AI. The artists employ audio AI in the soundtrack for the work as well, for example the chorus sung in the beginning. In this work the artist duo explores how such tools can serve as masking filters, making it possible to visualize deeply personal events, offering a public version of a private story. Memory itself is slippery and subject to reinterpretation, and so is the variable output of these tools, which can produce limitless permutations of the same event.

Artist and musician Holly Herndon and artist and technologist Mat Dryhurst, alongside their own music, software, and digital experiments, are some of the first to explore visual projects, often released as NFTs, with next-gen text-to-image generators like Dall-e and Stable Diffusion. As a rare bridge between often siloed cultural sectors, including experimental music, advanced technology, and generative visual art, Herndon and Dryhurst have been nurturing and narrating the growing web3 community connected to blockchains and NFTs. They recently released a folksy cover of Dolly Parton’s Jolene made (not by them) using their tool Holly+, the vocal deepfake platform they offered up to the public, and for use by other musicians, last year. These tools raise urgent questions about authenticity and authorship, which the pair productively blur. They have also been vocal narrators of the ethics of web3, issues of consent in the age of AI, and questions of what data is being used to train these emerging intelligences. In September 2023 they announced their new venture Spawning, through which creators can opt out of the datasets that shape AI. The pair also produced a podcast called Interdependence, where artists, curators, and digital pioneers discuss the future of digital art and culture.

This artwork is licensed by KADIST for its programs, and is not part of the KADIST collection.