Could an AI-generated drag show reveal the technology’s biases and limitations? The Zizi Show, a „deepfake drag cabaret” on display at the V&A Museum in London, hopes to investigate just that.
On screen, a drag queen struts down the aisle, decked out in a glittering bodysuit.
Two other AI-generated queens follow her, but as they land on the ground, their legs suddenly lose shape, melt and multiply, blending with the colors and patterns of their dresses.
Faced with an unfamiliar situation – making splits – the artificial intelligence was unable to adapt the movements of the three queens, resulting in a meltdown of the technology.
The Zizi Show is designed to capture these moments of confusion and imperfection, showcasing the flaws in artificial intelligence. The exhibition at the V&A Museum is part of a larger project, the Zizi Project, that artist, coder and producer Jack Elwes started working on five years ago.
„It explores the intersection of traction performance and artificial intelligence”, they tell Euronews Culture.
„The basic idea is that artificial intelligence has a lot of problems. We’re trained on data sets that have a bias toward the norm, so we’re often trained on certain kinds of people. My thought is, what if we train AI on otherness, quirkiness, difference? „
How to connect an AI drag show
The generation of Zizi Show started during the lockdown.
As the Covid-19 pandemic brought everyone’s lives to a halt, Elwes was funded to work on a project that reflected the queer resilience. A key collaborator in the creation of the Drag Me Queen project, Elwes originally helped discover 13 drag artists from the London scene. The group then set up video studios in vacant lots closed by the health crisis.
“We filmed everything [drag performer] About three minutes, walking and moving through space, that became our data set”, explains Elwes.
The resulting content yielded about 7,000 frames, which were then used to create skeletons following the movements of the drag performers.
When creating the AI drag show, Elves used these skeletons to uniquely represent each drag artist’s movements and applied them to each other’s AI-generated image. After receiving more funding, the number of drag performers increased from 13 to 21.
Some artists moved in ways others didn’t when they were first recorded. For example, three queens attempted to do the splits — only one performed the move during the recording. Another was wearing a bra during the recording, but the clothing was unfamiliar to the AI technology. As a result, when the AI tried to adapt the queen’s movements, the breastplate remained stationary, but multiplied.
The Future of AI: All Doom and Gloom?
„Drag is the perfect form of gender nonconformity to interrogate gender bias, something really accessible and fun,” Elwes tells Euronews Culture. They hope the Zizzi show can help people understand how AI works, „unnecessarily” using technology to slow down the „fear mongering”.
„I’d like to see more people using things like Dal·e or Midjourney with an awareness of where the data is coming from and who’s training it.”
Despite the dangers inherent in using the new technology, such as the misuse of AI to deepfake content, Elwes feels „optimistic” about its future.
„I don’t feel super dystopic,” they say, rather than accepting that all is lost as long as critical engagement with AI continues.
Exhibition by Jack Elwes Data • Confusion • Utopia On until July 8 at Gazelli Art House, London.
„Totalny pionier w sieci. Specjalista od piwa niezależny. Ewangelista popkultury. Miłośnik muzyki. Nieprzepraszający przedsiębiorca”.