dmstfctn will create a single-channel audiovisual interactive installation provisionally titled “Waluigi’s masks”. The installation will combine real-time 3D rendering and text generation to create a series of masks able to improvise short theatrical sketches in the style of Commedia dell’Arte. Inspired by common Commedia archetypes, the masks will represent different AIs with a tendency to hallucinate, confabulate, repeat rumours, create their own truths or outright cheat – traits that have been observed in Large Language Models trained on vast amounts of text.
Rendered in real-time using a video game engine and voiced through text-to-speech, the masks will improvise sketches within the traits of the archetype they are based on. They will be prompted and teased by an audience using their phone to drop theatrical props in the scene that represent common jokes, sayings and superstitions, as well as by asking open questions.
The masks will speak through a multilingual AI model which the artists will fine-tune during the residency with the help of project partner Fondazione Bruno Kessler.
This project is about teasing AI language models to elicit the confabulation, hallucination or disobedience sometimes observed in Large Language Models. It is loosely inspired by the “Waluigi Effect” meme which comments on the potential of these models to collapse into a shadow version of themselves prone to do the opposite of what asked – possibly due to the prevalence of fictional texts in the training data containing protagonist-antagonist tropes and archetypical characters that change beliefs or sides.
dmstfctn
Jury Statement
Dmstfctn is the artistic duo selected by Sineglossa for the residency at Tecnopole, supported by Cineca and FBK. With the proposal titled Waluigi’s Mask the duo continue their research on the relationship between simulation, machine learning and AI folklore. This project will recover the archetypes of the Commedia dell’Arte to show what is usually hidden in our interactions with generative Ais: the potential of Large Language Models to confabulate, cheat and deceive, producing disinformation. Dmstfctn moves beyond simply scientific debunking and rather engages audiences in an interactive exploration of its outputs, showing how fake news can be easily created and spread.
Federico Bomba, Sineglossa Art Director
![](https://ars.electronica.art/eudigitaldeal/files/2024/02/dmstfctn-at-unsound-bubbles-2023-photo-by-natalia-knycz.jpg)
dmstfctn (UK)
dmstfctn (pronounced “demystification”) is a London-based artist duo working with audiovisual performance, games and video installation. Since 2018, their work investigates complex systems through the study of communication networks, capital, information flows, and, more recently, AI.
Among their latest productions are Waluigi’s Purgatory and GOD MODE (ep.1), interactive audiovisual performances exploring AI folklore and the use of simulation in AI training. dmstfctn have performed and exhibited at Berghain, Serpentine, Design Museum, Onassis, Corsica Studios, HKW, Fotomuseum Winterthur, LUMA Arles, ARTER Istanbul and Aksioma among others, and at festivals such as Unsound, CTM, transmediale, Semibreve, Borderline and Impakt. In 2021, Krisis Publishing published ECHO FX, the artists’ performance about Brexit market manipulation, later also published in Ø: The Book by Flatlines/Hyperdub. In 2019, Mille Pleateaux released Flash Demons, a collection of the artists’ performances focusing on financial market crashes. dmstfctn were recipient of the 2017 transmediale Flusser artistic residency.