Catastrophic Forgetting
"The tendency of an artificial neural network to forget previously learned information when acquiring new information is called catastrophic interference."
Catastrophic Forgetting is an audiovisual installation that explores a fundamental problem in AI: the loss of previous knowledge when a neural network acquires new data.
Through this lens, the work raises questions about a post-human future: In a world where machines can remember infinitely but humans no longer exist, what good is that information? Will machines feel nostalgia for memories they never truly lived?
This project was designed as a site-specific work for the unveiling of the permanent multi-channel system at X-Church (UK)—one of the largest discrete audio arrays in Lincolnshire.
The System: A massive 25.1 discrete audio setup integrated into a 19th-century Gothic church.
Spatial Strategy: I utilized Object-Based Audio to treat each sonic memory as a physical entity within the architecture of the church. This allowed for hyper-precise localization and a sense of “machine memory” surrounding the audience.
The Tools: The spatial configuration was built using the SoundSquares plugin (designed by Daz Disley), allowing for custom movement protocols across the 25-speaker array.
The Experience: While the live performance was a physical 25.1 immersion, the digital record utilizes Binaural translation to preserve the spatial complexity for headphone listeners.
Original Title: Catastrophic Forgetting
Premiere: Nov 27, 2021 | Cold Diffusion Festival, X-church, Gainsborough, UK.
Audio Performance: Multichannel 25.1 (Object-Based).
Audio Video: Binaural.
Duration: 17:42.
Core Team: Sol Rezza (Concept/Realization), Daz Disley (Engineer/Supervisor), Lena Sass Hughes (Voices), Irina Vainio (Technical Support).