Bruce Gilchrist
The Poetics of Garbage project manipulates waste material via performance, computer vision, object recognition software, and a language model, translating it into a poetic narrative as an expression of how trash can be rationalised by a machine to produce language.
Created as part of a BeFantastic Beyond Fellowship 2022-23 (In), co-produced by FutureEverything (UK) and supported by the British Council.
When moving, the shape of an object appears to change within the computer vision frame, often eliciting misclassifications, which we understand as part of a novel form of interaction and chance operation in art. The text labels produced by YOLOv7 object detection (including misclassifications) were incorporated into language prompts that were consequently ‘autocompleted’ by a language model (GPT-3) into a series of statements—a form of commentary that alludes to the picture in surprising and poetic ways. Iterations of these text outputs were curated, translated into Hindi and Tamil languages and narrated by an Indian voiceover artist as the soundtrack to the video.
The title of the video alludes to the trope of a ritualistic act of pilgrimage, a journey into the unknown in search of new or expanded meaning. We imagined this sense of altered meaning being achieved through the actions of the body and its interplay between waste material and predictive technology, where spent things reclaimed from the realm of the useless are considered an opportunity, an untapped resource waiting to be harnessed.
As part of the FutureFantastic exhibition in Bangalore, we created an interactive, live-camera installation to demonstrate the underlying principle of Poetics of Garbage. The displayed costume in the gallery was subject to the same object recognition algorithm but used a different language model—an RNN text generator—that was trained on the text corpus that was originally generated from the street performance. In effect, the live-camera installation performed an algorithmic remix of the Plastic Prāyaścitta narrative when a gallery visitor changed the position of the camera.
Credits:
A collaborative project made by Aashna Arora (performance & costume), Bruce Gilchrist (concept, video edit & text generation), Chaitali Kulkarni (software programming) and Thaniya Kanaka Mahalakshmi (narration & language translation). Additional language translation: Malvika Jha / Champa Lakshmi.
Music by Traitor.
Cameras: Purvi Joshi / Rahul Naag / Jonathan Peters.
Header image: Aashna Arora performing at Yumana Ghat, New Delhi, India. Production still by Rahul Naag.
