EKPHRASIS

How can poetry be a used to interrogate generative AI? An exploration of C.P. Cavafy.

Marc Da Costa & Matthew Niederhauser (2023)

Single-channel video with stereo sound and accompanying hardcover book.

“Poetry can break open locked chambers of possibility, restore numbed zones to feeling, recharge desire.” –Adrienne Rich

Ekphrasis is a machine learning video work and artistic research project that explores how poetry can illuminate the algorithmic systems that increasingly shape what we feel and imagine. In this iteration, the poetry of C. P. Cavafy becomes a filter and prompt to interrogate the visuality of today’s machine learning tools. These technologies are the magic sorcerers of the moment, spooking and exciting us, yet drawing their material not from a supernatural world beyond, but instead from the zeitgeist of images and text that live online.


C. P. Cavafy’s poetic corpus can provide a glimpse into how artificially intelligent systems attempt to construe and interpret the more subjective edges of language. Edges that can both inspire and confound. Our results are also detailed in a printed compendium which acts as a codex to an installation that will cycle through machine learning responses to his work, including AI-driven readings and other visual renderings of select poems.

This project was created with the support of the Onassis Foundation USA and ONX Studio.