AN INTERVIEW WITH JAKE ELWES BY FRANCESCA IONESCU
Ten years ago, you'd have found Jake Elwes away from their classmates at the Slade School of Fine Art. Despite the institution's reputation aligning more with 'traditional media' (think: no algorithm involved), Elwes spent his days in the basement, learning how to code and playing around with some of the first generative AI models. This was not the ChatGPT or Copilot of today, which can create the perfect corporate email, respond to your dating-app matches or reunite you with a long-lost relative in an uncanny video. Despite these advancements, Elwes' work has remained purposely strenuous, their focus on otherness and ethics keeping them away from the large corporate models,and pushing them into smaller, self-filled, data sets. Their work branches into many hot topics around AI, from copyright to porn, but their mission remains clear: AI can be used to make art, but the artists must resist it from the inside.
Francesca Ionescu: You started using generative AI 10 years ago, and the advances since have been insane. How does it feel to have watched this technology evolve?
Jake Elwes: It's quite a strange thing to see. Working with these technologies in the early days felt very different now. It felt like it was your medium. You could mess with it on a data level — you had to train your own models using your own data. Now, you're prompting into a model that's been trained by a Big Tech company. In a way, I'm very attached to the old ways of doing it, and I'm still using those old technologies today. A lot of the glitches that I was getting, you don't get today. It's too good. It's too perfect.
FI: Do you think there are different worries for artists that primarily work with physical media compared to those of digital artists like yourself?
JE: I feel like most of my friends who are more embedded in the fine art world are very anti generative AI. They really do see it as this 'slop'. I think it's a more conceptual belief that it's moving towards this very homogenous, generic aesthetic. But I've also seen a lot of digital artists embrace it far too readily. I want people to be a bit more critical, to push back a little bit more. I've seen it as a collaborator for a very long time, but that relationship has changed as Big Tech adopted it and it's become more mainstream.
FI: You almost have to resist it from the inside.
JE: Absolutely! As artists, we need to be engaging critically with this, to think of new futures, new ways to be using this technology. We need a lot of different visions of how these tools can be developed and where they can go from here.
FI: On the topic of resisting, talk to me a bit more about Zizi.
JE: So, my partner works in a Drag show venue, and their [show] partner, Me The Drag Queen, has become one of my main collaborators on Zizi. It all started with a project called Zizi - Queering the Dataset, using the most cutting-edge AI technology at the time, which was starting to create these fake faces that didn't exist. I was very struck by the fact that the data sets that it was trained on were standardised facial recognition data sets, originally trained on American celebrities. Of course, it's going to have a real bias towards a certain type of identity. I became quite interested in finding out how we can inject otherness and queerness into this system. I feel like it works well because drag is the ultimate form of playing with gender non-conformity, and it's a nice way of playing with gender bias in the system.
FI: What made you move onto the next stage, the deepfake cabaret?
JE: The thing is, these faces couldn't perform. They would just effectively exist. It was obvious they were not real human faces. So the next project became: how do we get this to perform? How do we get this to lip-sync Bowie or Beyoncé songs? So then it became a collaboration with the London drag community.
FI: With drag being rooted in taking on a persona, how did you face those questions about ownership and consent?
JE: It did become a project about the ethics of deepfakes, and asking if you can even create an ethical deepfake? How do you do this with consent? How can you make sure that everyone gets paid for their data, or that anyone can withdraw from the project? And, importantly, who has the right to puppet whose body? We wanted to make sure it was people inside the [drag] community that were puppeting each other's bodies and making them perform.
FI: Do you think this gets rid of the person in a way? There's no longer anyone behind the make-up.
JE: I was really interested in the layers of artifice and construction. You've got this deepfake version of a constructive persona, which deconstructs itself, as well. When you see it glitch, when it breaks, and when you see the skeleton that the whole thing was based on, that's when you see the deepfake technology underneath it. There's something in the artifice of drag and deepfake technology that are kind of playing into each other.
FI: Where do you think your exploration of AI in art is going to take you next?
JE: We're working more on the idea of performance. I think there's a really exciting place to see where live performance and this technology can co-exist. We're planning the Zizi and Me project — live performances with the drag queen next to her deepfake doppelganger, on stage together. This show where they do musical theatre duets together, to satirise the notion that AI can take over from a human artist. We're playing with the existence of AI in society and exploring this 'competition' between the two. We need the resistance, but I also think there's a place for hope and playfulness and joy.
DoomScroll, 2023
Images courtesy of Jake Elwes
