Triggered by the world-wide shutdown brought on by the COVID pandemic, the authors viewed the situation as an opportunity to comment upon issues of isolation, remote collaboration, separation, and potential convergence.
Scaffolding on a previous generative system, Coming Together: Cityscapes, this multi-agent system uses ten 30 second videos by Kathryn Ricketts, and a database of audio recordings made by four musicians: Joshua Hyde (Graz), saxophones; Alice Purton (London), cello; Felix Del Tredici (Montreal), trombone; Daniel Brew (UK), guitars. These performers were asked to improvise on a general theme of living during the COVID shutdown, and limit their improvisations to 30 to 40 seconds in duration; the recordings were then analysed for spectral content and segmentation.
The work is an addition to the series of Coming Together generative artworks in which the process of convergence by agents is the focus of the work, in the movement from random individualism to united ensemble interaction. The use of multi-agents are integral to the generative process; instead of attempting to create a complex top-down system for intelligent control or creation, imbuing individual agents with the capacity to make decisions and interact with other agents to create complex, dynamic, and emergent systems provides a bottom-up method.
Given a database of audio and video recordings, five audio and five video agents communicate with one another and negotiate their way through the archive in an effort to arrive at a single clip. This arrival does not always happen (as certain agents can become obstinate in remaining with their current selection); the system has a series of “checks” to determine progress: if the agents haven’t achieved a compromise on specific tasks, the system gives up, and starts again.
The work has three distinct phases:
The agents attempt to arrive at the same clips. They initially chose a random clip from the database, which is shuffled with each run (video clips become “attached” to certain audio clips at this time). Each audio file has been analysed and segmented; a segment is initially chosen, and is time-stretched and heavily filtered. Agents select prominent Bark bands in the recordings, and “claim” these spectral regions – the number of bands an agent can claim depends upon how far along the clip negotiation has progressed. The partnered video agent similarly applies a selective colour filter based upon hue to its video.
The agents attempt to align their starting points. To do so, they shorten their gestures. Note that the audio agents are still playing at maximum 5 of 25 possible Bark bands; to realize a full spectrum performance of a file, the agents must align their starting points. Similarly, the video agents are still applying a hue process, selectively allowing certain hue ranges to pass while others are “greyed out”.
The agents attempt to expand their segments again. Often, they get out of alignment again, as their individual lengths will differ, and play through the negotiated “downbeat”: this is not a bug, but an embraced feature. This section also has a three minute limit, and the system restarts for another version.
Kathryn Ricketts has several characters that she inhabits for the purpose of creating kinaesthetic provocations towards emancipatory encounters. Her 16 year old character LUG, donning an old overcoat and felt hat and always ‘lugging’ and old leather suitcase, dances stories of displacement, longing, belonging and in-betweeness. In the period of isolation, LUG and Kathryn wandered her neighbourhood; the warehouse district of Regina, Saskatchewan, and each week they created a very brief improvisation inspired by a site that resonated with them. The works are in dialogue with fierce winds, new grass and blooms, ever changing graffiti and general grit and rust akin to this area. The themes that arose; islands of isolation, tunnels of despair, fierce need for rest not wanting to be just another statistic, longing for touch.