Holly Herndon and Mat Dryhurst: xhairymutantx (original) (raw)
Commissioned for the 2024 Whitney Biennial, artists Holly Herndon and Mat Dryhurst’s xhairymutantx exists both within the Museum’s sixth floor gallery as well as here on artport. This project focuses on training data behind artificial intelligence (AI) models, opening new possibilities for its use. “Holly Herndon” is not just a person. The name also designates a distinctive internet presence: a female character with white skin, red hair, blunt-cut side bangs, and bright blue eyes. Herndon has become well known in the music and digital art worlds, to the point where when someone types the words “Holly Herndon” into a text-to-image AI program like Dall·E, Midjourney, or Stable Diffusion, the prompts generate an image with some of the characteristics of Holly Herndon, the character.
Here on artport, the artists have trained a text-to-image AI model on images of Holly that have been altered through costuming that distorts the artist’s body, and exaggerates her most noted feature, her hair, to transform her identity within AI models. No matter what text prompt is entered by the user, the results will generate a strange version of Holly. The new images are stored in the project gallery, thereby entering the internet at large and potentially becoming part of the data set behind new AI-generated images. Since AI programs view institutional websites like whitney.org as trusted sources, the artists play with the idea of using the Museum’s heft to influence the parameters of AI models, and to raise questions about the extent of self-determination possible with the internet today.
The 2024 Whitney Biennial is organized by Chrissie Iles, Anne and Joel Ehrenkranz Curator and Meg Onli, Curator at Large, with Min Sun Jeon and Beatriz Cifuentes.
Significant support for xhairymutantx has been provided by Colin Brooks and David Lisbon.
Holly Herndon (b. 1980) and Mat Dryhurst’s (b. 1984) collaborative practice navigates uneven distributions of power using AI technologies and virtual ecosystems. Both are multimedia artists: Herndon records cerebral and whimsical studio albums combining electronic music, ASMR, and artificial neural networks; Dryhurst produces music in advocacy of a decentralized internet. Their shared project, Holly+, is a machine-learning instrument that converts audio files into Herndon’s voice.
Housed in a combined set of artistic and technological practices—pedagogical activities, public performances, music making, a podcast—and firmly committed to collaboration, Holly+ exists as a formalized model through Herndon and Dryhurst’s DAO (decentralized autonomous organization). While the two artists operate through a rigorous and playful embrace of the creative, transformative capacities of vocal processing tools, Holly+ illuminates how such tools function in a tangled web of subjectivity, private property, corporate control, disembodiment, and representational autonomy. The embedded self-implication of Holly+ as vocal clone and digital twin is significant given the acute ethical stakes associated with AI as an arena of experimentation. Herndon and Dryhurst’s project is optimistic in calling for serious attention on how artists can be protected while benefiting from the development of sonic prosthetics. The duo’s formulation of “vocal sovereignty” addresses how this plays out; and they offer the term “spawning” to describe new forms of media mimicry.
Audio
- 0:00
Minisode: Biennial Artist Holly Herndon in Conversation with Whitney Youth Insights Leaders
0:00
Teen Narrator:
Welcome to a special minisode from the Whitney. I’m one of the Youth Insights Leaders, part of an after school program where we work with artists in the Museum’s exhibitions. For this year’s Biennial, we interviewed one of the artists in the show.
Holly Herndon:
My name is Holly Herndon and I'm an artist.
Teen Narrator:
We talked to Holly about her art, which she makes with her husband Mat Dryhurst.
Holly Herndon:
So we created this model that is hosted on artport that you all can play with.
Teen Narrator:
artport is the Whitney’s internet art portal. When you open Mat and Holly’s project up, you see a box to enter text. Below that, there’s a grid of AI-generated images. The pictures are mostly people and other creatures with long, ropey, orange braids and thick bangs. A lot of them are wearing puffy green suits. Most of them are female. They’re kind of like distorted versions of Holly. When you play around with their work in artport, you get more of these AI-generated images.
Teen Narrator:
Sometimes they’re a little off. Like we asked for a claymore, and it was just a regular sword. And the program didn’t seem to know what a Mega Chad is.
Holly Herndon:
I'm trying to change who I am in public AI models, what my embedding is in public AI models. An embedding of a bottle would have a bottleness—like an essence or a certain bottle-like quality that we kind of all as humans understand is the core essence of a bottle. And then to extrapolate from that, when you get to more abstract concepts, that gets more and more blurry and gray. And so we were able to actually do some reverse engineering and look at my embedding, and it really turns out that I'm kind of like this just blob of orange hair and bright blue eyes. These models are trained on the open internet. So any images of me that are tagged with my name, that's basically what creates the concept of me in these models. And the way that these systems work is I actually don't have any control over who I am. It's just this kind of aggregation of images that other people have uploaded. So I was kind of asking myself, "How much agency can I have in this system? What can I do with it to push back on this a little bit?"
So I made a costume with a giant haircut. So I basically just turned myself into my pastiche, just turned myself into my haircut. And then we made a model that people can navigate on the artport site. The data that goes into it is kind of ranked according to trustworthiness of the source because going back to the original idea of trying to find this objective truth, even though that's of course a very, very problematic thing to try to reach, but that's how these systems work. So anything that shows up on Whitney.org is going to have a higher ranking than something that's on my random blog spot somewhere. So then the next time that a model is created, we have to kind of wait and see, but we're thinking that my public embedding will be infected with this kind of character that we created.
Teen Narrator:
We asked Holly about what kinds of creativity she could express using AI.
Holly Herndon :
Every project that I've done with AI has been extremely manual. It's not just some automated process where it's type in a few words and art is done. It's usually very laborious, many decisions made. I think there's this perception that it's this fully automated thing, and it can be for some people, but for my practice it's really about getting into the bones of the model, understanding the training data, understanding the broader systems that the model is situated in.
Teen Narrator:
We also wondered what she thought about the future of using AI in art.
Holly Herndon:
I think that we're on this precipice of things dramatically changing because media will become kind of infinite and really easy to produce. I think it's going to change how we think about intellectual property, which is basically authorship because these systems are inherently collaborative. So you can collaborate with other people directly, but you can also collaborate with the entire human history, which is kind of a weird thing to wrap your head around.
So I think it asks us to question some of the things that we take for granted from a 20th-century approach to artmaking. I think it puts everything in question. I think that's really exciting. I think we're going to see artists using machine learning models as an art form. Whereas painting is a category, I see models as a category because they're these kinds of worlds that are infinitely navigable and generative that you can create a world and your audience can then create work through you or with you and kind of dive deep into your world in a really interactive way. And I think that that's very rich territory for artists.
That approach of humanizing these really technical systems has been there from the beginning of my practice, and I hope that it remains and I think that it remains. I think a lot of the things that we talk about with the work is a focus on the training data and a focus on how these systems aren't these alien intelligences, but they're just like aggregate human intelligence. It's actually a really remarkable human accomplishment, AI. I don't see it as this alien accomplishment. It's like us all together, and that's something to be celebrated if we can see the kind of humanity in it.
Teen Narrator:
Artists Among Us Minisodes are produced at the Whitney Museum of American Art. This episode was produced by teen Youth Insights Leaders: Hale, Kiyan, Jinhaohan, Gabryellah, Brigitte, and Sahara. Production support was provided by Whitney staff.
See more on artport, the Whitney Museum's portal to Internet and new media art.