BARRETT WHITE / PERFECT POEMS
The piece combines six clips of "digital poems" produced using an early version of a GAN (General Adversarial Network) AI (VQGAN+CLIP). To produce these pieces, I provide the program with a text prompt, and the network produces images based on that prompt, culled from a digital library containing millions of data points. The results are then converted to video, looped and inverted, combined in a form of rough synchronization, and set to music (an original sample-based composition). The concept behind these pieces is to interrogate the AI's conception of "the perfect poem." The AI was asked to create the following images (in this order in the video): "the perfect sonnet," "the perfect sestina," "the perfect erasure poem," "the perfect concrete poem," "the perfect rhyming couplet," and "the perfect prose poem." The final compositions produce a faint presence of text, but are ultimately unreadable. The kinetics of the images, which pulsate and slightly morph over their thirty second lifespans, seek to engage in a contemporary dialogue with asemic writing, concrete poetry, and digital art. These poems are inspired in part by the formative work of poet-programmer Allison Parrish, as well as the "poetics of non-reading" explored by Tan Lin, who in his piece "Eleven Minute Painting" calls for an ambient literature where "poetry equals wallpaper," "novel equals design object," and "works of literature [don't] have to be read but [can] be looked at, like placemats."
Barrett White edits Tagvverk. Poetry and critical writing has appeared or is forthcoming in Diagram, The Brooklyn Rail, P-Queue, baest, and FLAT, UCLA's design media arts journal.