Figure Infinity is an AI powered performance project of Jung-In Jung
and Lou McHugh. The work connects human performers in a self-reflexive
network of control and communication with Artificial Intelligence. It
responds to recent narratives surrounding “AI”, which largely obscure
the collective human endeavour that produced the data these systems are
built on, and that continues to be necessary to maintain it’s
plausibility. Instead of disguising the precaritised 'ghost work'
at the centre of AI systems, we foreground the human factor. This
project presents the general public with an opportunity to the interact
with the increasingly main-streamed large-language model ChatGPT in a
creative and critical way, simultaneously showcasing the creative
possibilities and limitations of AI systems.
Themes
Figure Infinity responds to recent narratives surrounding “AI”, which largely obscure the collective human endeavour
that produced the datasets AI systems are built on. It explores themes of platform labour,
algo-management and AI failure, encountered uncannily through absurdity and play.
Sound and visuals are used to highlight cybernetic feedback between human
performers and AI Agents as an “aesthetic of failure”, with feedback and
noise used as compositional materials. Sonically, the environment changes
with the performers position on stage due to the game track controllers
they are tethered to. Live video feedback produces ethereal visuals for
the audience, but is interpreted as confusing noise by the AI agents.
Further to creating the desired aesthetic, these aspects help reveal the
different experiences between AI and audience.
Interaction
In Figure Infinity, a sole dancer is embed in a system of control and communication with
Artificial Intelligence. During the performance,Computer Vision(CV) and Large Language Model(LLM) AI
agents collaborate to algo-manage a human dancer on stage by generating directions in sequence.
Human participants play the part of creative, yet invisible labour working to refine the system, as we micro-task
the audience to tag the actions on stage using their phones. This action further feeds data set
the LLM uses to create the next performance direction. The website they interact with,
http://figureinfinity.io/ presents a nebulous “AI-Powered Performance” firm, which grounds the project in
its core themes.
Technology
The software used to facilitate a network of communication between performers, AI and audiences included
python, javascript, max msp with the YoloV8 computer vision library and Open AIs GPT3.5 making up
the "brains" of our AI agents.
A hacked game track controller was used to create the tethers that bound the performer, while an arduino
printer was used to print out the performance directions.
This helped devirtualise the performance and introduced elements of theatre to help engage the audience.
