Zizi in Motion: A Deepfake Drag Utopia
This silent video series consists of life-size deepfakes of numerous London drag artists. Elwes poses the question: how can we, as a queer community, subvert deepfake technology by using it in an ethical and consensual way to celebrate queer bodies? Everyone participating in the project maintains control over their own data and involvement and is part of the same community.
The dataset, which AI made use of, consists of the shapes and movements of 21 drag artists from London. This data was then reused by the famous drag artist Wet Mess. The AI attempted to mimic Wet Mess’s movements using the information in the dataset. The unexpected movements, some of which were not in the dataset, caused the AI to naturally make mistakes. These generated new ideas within AI about what the queer community might look like.
Artificial intelligence is fed with datasets that favour standardisation and is therefore often based on very specific types of people. There are already plenty of cases where AI reinforces biases arising from the examples it is fed with. As a gender-nonconforming identity, drag is a perfect form to playfully demystify gender bias in AI and make people aware of this process.
With thanks to the drag cast: Baby Lame | Bolly-Illusion | Bourgeoisie | Cara Melle | Charlie Wood | Chiyo | Dahc Dermur VIII | Dakota Schiffer | HERR | Lilly SnatchDragon | Lavinia Co-op | Luke Slyka | Mahatma Khandi | Mark Anthony | Me The Drag Queen | Miss Terri Boxx | Oedipussi Rex | Ruby Wednesday | Sister Sister | TeTe Bang | Wet Mess.
