Mature Poets

“Immature poets imitate; mature poets steal; bad poets deface what they take, and good poets make it into something better, or at least something different.” — T.S. Eliot, The Sacred Wood

Mature Poets is a single-channel, real-time video projection generated entirely by custom software. The generative image-making models central to today’s AI technologies were constructed by scraping and training upon millions of existing artworks, photographs, and films — gathered without attribution or compensation. The resulting outputs are, fundamentally, statistical recombinations of this collective visual archive. Mature Poets is built solely from these visual fragments.

Smith considers whether genuine novelty can emerge from a system trained exclusively on stolen material. To test this, he introduces an algorithm driven by an autonomous reinforcement loop: an artificial agent motivated solely by curiosity.

The projected artwork constantly shifts from scene to scene without cuts, ensuring no image is ever repeated. The agent evaluates every new image against its entire historical visual log, selecting scenes based purely on their structural novelty. Once a visual trajectory has yielded all its variation, the system abandons it and moves on. Drawing on Sol LeWitt’s conceptual framework, where ‘the idea becomes a machine that makes the art’, Smith allows the machine to select its own subject matter. The audience witnesses an artificial curiosity attempting to navigate an environment made entirely of borrowed images.

By doing so, Mature Poets turns the question back towards human creativity. If Roland Barthes described the text as a ‘tissue of quotations’, and every artist builds their vocabulary from what they have previously observed, Smith invites us to consider whether human expression relies on the same fundamental mechanics of rearrangement — and what we are truly defending when we claim otherwise.