Tristan Wheeler
Tristan Wheeler — Attention Is All You Need
Cosmos & Compost — Winter 2023
This semester project documents the research and experimentation that laid the digital groundwork for a future production.
This project explores how machine perception—especially in virtual production—reshapes visual culture. An artificial bird performs a “courtship dance” for a camera, seeking the attention of an image model (VGG). Through GradCam analysis, its choreography aims to guide heat maps toward coherent patterns rather than noise.
Drawing on Fischerian drift and genetic mutation, it parallels biological and computational evolution. Platform seeing emerges as images become “operationalized” by data-driven processes, transforming the production pipeline into an interactive system linking performer, camera, and algorithmic observer.
Focusing on attention—how it’s steered, manipulated, or captured—the work reveals creative possibilities in “catching the gaze” of AI. By interpreting GradCam’s feature extraction and classification, the project highlights the invisual layers of AI’s decision-making. Balancing symbolic performance and computational observation, it shows how optimization, randomness, and intention converge to reshape our understanding of the visual.
