Latent Dispatch is a research project and interactive artwork that stages a conversation between machine precision and human imagination. Given a set of prompts, a novel machine learning system constructs single-line drawings that gravitate towards statistical averages learned from vast amounts of training data. These mechanical, precise plotter drawings are juxtaposed with the diverse and often playful doodles contributed by passing audience members, revealing a tension between the machine’s pursuit of a pure archetype and the human impulse towards endless variation.
The work emerges from a concern about the feedback loops between media and reality: how our perception of the world is increasingly shaped by algorithmic systems, and how those systems, as they train on their own outputs, risk “flattening” our cultural landscape into a shallow median. Latent Dispatch offers an alternative vision of machine learning tools, eschewing conventional image generators in favor of something more chaotic, interactive, and participatory—grounded in mark-making, drawing instruments, and evolutionary strategies. The project exposes the narrowing of machine perception and asks what we might do to resist it. Even a small gesture—a spontaneous doodle, or quick scribble on paper—can inject noise into the system, interrupting the loop and preserving a richer, more organic landscape of human experience.