At this point in the conversation, many of us understand that the AI models were trained on data pulled from the internet, which means they were unwittingly codified with all sorts of age-old biases learned from humans, and tend to favor the white cis man as society’s gold standard. So what better way is there to illustrate and subvert the flaws in technology than through the art of drag, with its long history of exposing and ridiculing outdated norms and rigid categories of identity?
This is the idea behind The Zizi Show, a deep fake cabaret by London new media artist Jake Elwes. Although the show exists in an interactive format online – inviting viewers to choose from a list of possible performers and songs – an in-person version ushered in the digital gallery of the brand new V&A Photography Center, which opened its doors. opened to the public last week. The installation features a large projection of three dancing performers while, in the opposite corner, a row of LED screens show six singing heads that periodically change shape into each other with each song change.
These moving images are profound counterfeits. Twenty-one performers representing the many genders, sexualities and subcultures of the London drag scene were each filmed walking around a room so that a still image from all angles could be used as a dataset. Elwes used a technique known as “skeleton tracing” to reduce the numbers to a series of moving dots which the AI then learned, over time, to convert back into a recognizable person.
For example, although only Ruby Wednesday was filmed dancing Five years by David Bowie, all 21 queens, kings and quings are able to do the exact same moves on screen. But do these AI interpreters correspond to reality?
Installation view of Jake Elwes and The Zizi show (2020-2023) at the V&A Photography Center in London. Photo: © Victoria and Albert Museum, London.
“It’s kind of ironic,” Elwes said. “Right now we’re talking about whether AI is going to replace the human performer, but that’s a bit of a joke because AI will never replace a human drag queen or king.”
Certainly, it is impossible not to notice how the performers, despite all their colorful and eclectic accessories and makeup, constantly sparkle and sometimes sparkle on screen. “My favorite moments are when it fails,” Elwes said, recalling a time when a performer, Cara Melle, fell into the divisions and the AI never figured out how to handle that statistical outlier.
In the short history of AI art, Elwes is something of an early pioneer. “Everyone lost their shit about this new technology, but I was kind of aware it was coming,” they said, after starting to work with AI in 2016 at school. machines with an unlikely name, Making and Make-Believe in Berlin. They immediately got into the art of programming, finding open source code to hack into so they could create a new tool and study the ways it broke down.
This conceptual approach has inspired Elwes’ works for several years, as evidenced by a retrospective that has just opened at Gazelli Art House in Mayfair, London (until July 8). Closed loop (2017), for example, explores what happens when AI talks to itself while Machine Learning Porn (2016) show how these tools can be reverse-engineered, in this case to generate rather than filter pornographic images.
Jacques Elwes, AI interprets AI interpretation “against interpretation” (Sontag 1966) (2023). Photo: Gazelli Art House.
Over the past year, Elwes hasn’t cared much about the newest, viral AI generative tools like DALL-E and Midjourney, preferring instead to create their own datasets and models from scratch. An exception was made for AI interprets AI interpretation “against interpretation” (Sontag 1966) (2023), which asks AI to interpret a line of text from Sontag’s seminal essay into an image and then create a label to describe the image, bringing layers of hallucinogenic interpretation to esoteric subjects like mimesis and form. “It’s a bit of a corny art theory joke,” Elwes said.
In general, however, Elwes’ interests have moved away from the philosophical implications of AI towards the political. “I now feel like those things were a real distraction,” they said. “Artists have a responsibility, in fact, to be honest about what they use and to deconstruct it. To show vanity.
In 2019, Elwes started Zizi – Queering the dataset, which took standardized facial recognition systems and fed them thousands of images of drag performers. “We will queer this data so that it shifts all the weights of this neural network from a space of normativity to a space of queerness and otherness. Suddenly all the faces start to break down and you see the mascara turn into lipstick and the blue eyeshadow turn into a pink wig. That’s wonderful.”
For the current iteration of the project, The Zizi show, the issue of consent took center stage. Against amid ongoing controversy Faced with Big Tech’s shameless exploitation of copyrighted content to power their gigantic datasets, Elwes is developing a non-exploitative model for creating art with AI
Jacques Elwes, The Zizi show (2020-2023) at the V&A Photography Center in London. Photo: © Victoria and Albert Museum, London.
After all, in the wrong hands “this technology, especially deep fakes, can be used for really dark purposes.” Elwes paid each artist for their participation, obtained their consent to be included in the dataset, and ensured that they understood exactly how their body would be manipulated and had the freedom to opt out of the project at any time. moment. “The only people who have the right to resuscitate the bodies of our queer performers are people from that community,” Elwes said.
The result is a celebratory, expressive performance with a charming humanity sorely lacking in most AI artwork. The audience’s enjoyment of the work is also pleasingly simple.
“I think the project has extra scale right now,” Elwes said. “It’s amazing to have this at the V&A because Congress is discussing the safety of AI, I have friends who are drag queens who have right-wing protesters in front of their shows right now and there’s obviously a lot of anti-trans sentiment in parliament. This is a difficult time for the queer community.
“What if AI was only built by fags?” Elwes then wondered aloud, with growing excitement. “Imagine that! It could serve the world much better if it were people who thought from an outsider’s perspective.
They pointed out that after all, the technologically optimistic view of AI insists that these tools will help build a better society. “How can we come to this? This is my profound fake drag utopia.
Follow Artnet News on Facebook:
Want to stay one step ahead of the art world? Subscribe to our newsletter to receive breaking news, revealing interviews and incisive reviews that move the conversation forward.
