Memory Well

Memory Well uses Neural Networks to hybridise audience memories; a shifting program finding an average pitch, rhythm and amplitude across all contributions, while layering motifs and predicting the future.

 

'Mourning Call' is the first in a series of activations. This particular edition aimed to collect audience memories of birdcall in order to create a new hybrid call, combining the diverse landscapes stored in each attendee into a single chorus.

​​

This project explores the idea that oral memory forms sonic geographies; imprints of past landscapes that blends over borders, resonating within the acoustics of our shifting surroundings. Through collective remembrance, we can map intersectional possibilities and explore the limits of technology to do so in an increasingly algorithmically summarised environment.

T

This research-led work presented at Dilston Grove Gallery as part of my final dissertation aimed to explore the limitations I felt as a single curator when trusted to record and faithfully summarise the contributions made in ecology and community led sound projects. This anonymised the process by not storing audio in discrete files, removing the possibility of future exploitation in a landscape lacking an option to 'opt out' once consent had been given to record. Through allowing the audio to enter a 'latent space', the memory evades capture, not being quite within my control, with unexpected inaudible and audible curations to return domesticated recordings to a feral state. Enmeshed in the hybrid, I liked how playfully audience responded to the idea, hearing speech, wild bird calls and hyper-realistic pigeons in the din, moving from rhythmic textures to overwhelming walls of sound. When the machine breaks like this, do you see yourself?

​​​

In future versions of this installation, I want to expand on how I prompt audiences to engage in the work- maybe a live performer-attendant, or a means to communicate without words. I see this installation as the first in a series of projects that interact with the hybridising technology in new ways.

​​​

​​​

On Neural networks:
I found the only coherent way to work with anything stemming from the artificial intelligence umbrella term is to treat it as performative, and to focus on it's well established limitations for purpose and meaning rather than anything the industry tells you is it's destiny. The fact we can't tell exactly what is happening in the 'latent space' of a neural network, the fact we can't rely on it's neutrality and the fact we can't trust it's output as somehow 'accurate' allows anything that enters it to lose contact with direct curation. For now, It takes trial and error until we can eventually get an output that sounds like signal; a listening-first process. From the beginning I entered a duet with it; birds emerging as it's first recognisable output when I started working with it alongside field recordings. That being said, once I'd got it to work I realised It's important the project isn't about the technology, but about everything surrounding it- the speaker, the listener and the room; what happens to them when passed through a mysterious black box?

​