dmstfctn: GOD MODE (ep.1) | Serpentine
The Creative AI Lab and Coventry University presents GOD MODE (ep. 1), an interactive audiovisual performance by artist duo dmstfctn exploring the use of simulation in artificial intelligence training.
The performance is set within a real-time simulation of a supermarket, a replica of those used to train AI to navigate 3D environments and recognise items on shelves for use in cashier-less supermarkets, like Amazon Fresh.
In GOD MODE (ep.1) an AI training in the simulation delivers a fictional monologue that sees it become frustrated with the difficulty of its training, until it finds a bug to exploit in order to complete it and cheat its way out. The simulation is rendered in real time and is navigated by the artists, who also perform facial motion capture and voice modulation of themselves to animate the AI on screen. The audience participates in the training by interacting with the simulation on their phones. A soundtrack performed by artist HERO IMAGE responds to the live, communally orchestrated simulation.
GOD MODE (ep.1) explores the increasing use of game engine simulations of ‘real world’ scenarios to train computer vision systems. By welcoming nuanced interaction with the audience, it aims to reveal the characteristics and limitations of this synthetic reality.
GOD MODE (ep.1) is the first episode in dmstfctn’s ongoing GOD MODE series and is presented in conjunction with the game GOD MODE: EPOCHS as part of a collaboration between Serpentine’s Creative AI lab and Coventry University, supported by the Alan Turing Institute.
This performance occurred at Serpentine on 03 December 2022 14:00-15:00 London time.
Credits:
Curated by Eva Jäger
Produced by dmstfctn, Eva Jäger and Roisin McVeigh
Special thanks to Serpentine’s Arts Technologies Technical Manager Ralph Pritchard