The Misfit Names

Two-channel video, real-time object detection system, large language model
2025

The AI real-time object detection system YOLO (You Only Look Once) watches a soothing aquarium scene together with the audience, featuring coral, sea anemones, and schools of fish. Although the imagery appears familiar and comforting, it represents an environment that YOLO cannot accurately comprehend. The footage was filmed underwater, resembling a tranquil aquarium, yet it holds the true depth and life of the ocean.

In the work, YOLO has only received limited object recognition training. Under unusually loose detection conditions, when it observes underwater creatures for the first time, it might misidentify a fish as a frisbee or interpret coral as broccoli. These “misjudgments,” when passed on to a large language model for sentence generation, create a layered dialectical relationship between the object, limited perceptual capacity, and mistaken judgment—unfolding an alternative form of narrative.

Divers: Te-Mao Lee, Jian-Han Lin, Pei-Chin Hsin

<<