To Wilt – MFA Thesis Installation

Concept & Research Question

To Wilt is a four-space installation that asks how large language models, environmental sensing, and moving images can be woven together to think about love, decay, and the passage of time. The work combines model-to-model dialogue, long-term image sequences, and live sensor data.

Space 1 – Rose Growth & Voices

A multi-screen video wall shows a 3D rose growth animation as a central visual motif. Three LLM “voices” – a philosopher, a neuroscientist, and a lover – speak over the images, each framing emotion and perception from a different angle.

Space 2 – Printed Dialogue

Two LLM “lovers” converse continuously on a webpage. Two networked dot-matrix printers, driven by Raspberry Pi, print their dialogue in real time. Visitors encounter the text as physical traces that accumulate over the course of the show.

Space 3 – Decay Dataset

A self-captured dataset of thousands of images records the decay of organic material. These frames are morphed over time and paired with LLM commentary reflecting on the “decay of feelings” and memory.

Space 4 – Instrumented Rose Tank

A large rose tank is instrumented with environmental sensors (temperature, humidity, etc.) that log conditions and print periodic measurements. LLM voices respond to changes in the environment, treating sensor readings as evidence in an evolving emotional narrative.

Technical Notes

The installation ties together scripted LLM prompts and responses, Raspberry Pi devices, environmental sensors, and custom software for logging, scheduling, and projection. It treats AI systems not as black-box oracles but as collaborators in an extended media ecosystem.

Reflections

To Wilt extends my broader research interest in computational media: how AI systems, sensing devices, and images can be composed to examine perception, attachment, and loss over time.