Poetry for Robots Puts a Borges Theory to the Test

Open Culture points us toward a site called Poetry for Robots, "a joint effort between Neologic, Webvisions, and The Center for Science and the Imagination at Arizona State University," that seeks to put theory from Jorge Luis Borges to the test, "asking on their website whether it is possible to teach machines the poetic quality of human language.” More:

“What if we used poetry and metaphor as metadata?” asks Poetry for Robots’ front page. “Would a search for ‘eyes’ return images of stars?” To find out, the site has begun crowdsourcing poetry from its users, who they’ve asked to submit pieces of verse (150 characters or fewer) prompted by a series of images posted there: you can write your poetry in response to the open ocean, an urban landscape, a cappuccino, paths diverging in a wood, or 117 other actual images meant to draw out textual imagery.

Then comes the test: can computers learn to make the same poetic associations humans do between word and image, image and word? If the Borgesian vision of metaphors existing in patterns holds true, then they will — computers perform few tasks better than pattern recognition, after all. This could lead not just to, say, artificial intelligence that can compose and even appreciate poetry, but poetic-language search engines — a deeply artistic extension of the seemingly frustrated natural-language search engine efforts pioneered by the likes of Ask Jeeves.

And if none of that works out, we’ll still have witnessed a fascinating thought experiment, just like Borges’ stories themselves. The writer’s original thoughts on the subject will certainly remain compelling, and you can hear them in his 1967-8 Harvard lectures on poetry ... that we first featured here a few years back.

Read all and watch some awesome video clips at Open Culture.