On Friday, I did something I usually loathe. I went to a Poetry Slam. This one was different, however, as its title indicated: The first ITP Code Poetry Slam, held at the fourth floor of NYU’s Tisch Hall, home of the Interactive Telecommunications Program.
I went, tipped off by Jakob Nolte, in order to have a look at a current of digital literature that does not take its cue from conceptualism in the vein of Goldsmith, Place, et al. I was initially worried that, as a link on the event’s website indicated, code poetry in the most literal (and uninteresting) sense would be present – code as poetry, sometimes executable, sometimes just mimicking the aesthetics of programming languages. But I was somewhat reassured by the fact that Allison Parrish, who wrote the @everyword twitter stream, would be part of the event: Parrish, an artist and programmer, tries to operate not (or not exclusively) from the inside of coding culture (with its obscure in-jokes and references making up most “real” code poetry), but tries to bring the means for generative production of literature to those who are writers in the first place. She taught a course on “Reading and Writing Electronic Text” at ITP, and much of the poetry presented was a result of that class.
I wasn’t aware of ITP before, but the loft-like space with its abundance of open work areas, circuit boards, maker-bots, and the smell of instant ramen reminded me of my one visit to MIT. The difference being that here Kunstwollen was added to technical expertise. Or maybe I had underestimated MIT’s own Kunstwollen. For the slam, which was really more like a traditional poetry reading with a projector showing text, was opened with a talk by Nick Montfort, professor of Comparative Media Studies at MIT, and part of Trope Tank. He showcased a selection of historic of computer generated poetry he had recreated in Python and JavaScript.
The first of them, Christopher Strachey’s “Love Letters” from 1952, was new to me. The others were classics: Theo Lutz’s “Stochastic Texts” (written under the supervision of Max Bense) from 1959, Brion Gysin and Ian Somerville’s “Permutation Poems” from 1960, and Alison Knowles and James Tanney’s 1967 “A House of Dust.” (Thanks to Dick Higgins, we even have the original FORTRAN source code.) Save Gysin/Somerville, all of them work according to the principle of combining a fixed syntax with finite semantics – basically computer generated Mad Libs with a limited number of words to choose from. (Have a look into the source code to see Strachey’s elements – adjectives, nouns, etc. – or have them neatly listed here.) Directly compared to these types of static rotation, Gysin’s work struck me as much more transparent in its process – permutating all words of a sentence, line by line – but also as more poetically effective. While in the end all of these poems could have been produced by hand, only in Gysin’s is the digital both present and represented, instead of hidden away.
This seemed to be the basic choice for the following not-yet-historic code poetry as well: to hide or to show its processes and make-up. It was not altogether clear which yielded better results.
The first segment dealt with Markov Chain poetry. An old favorite of computer generated literature, a Markov Chain is a stochastic model that maps the probability of a state depending on the preceding one (at least this one type of it, here’s a good explanation). In a way, it is what the QuickType function of the new iOS does: “you” is more likely to be written after “love” than most other words, so QuickType would suggest it. Similarly, for Markov Chain poetry, often a body of text is analyzed for the most probable collocations, which then produce the output.
Aankit Patel’s “Centrality” did this with a single verb as an input and Wordnet, a lexical database for English, as corpus. The output contains scraps of definitions, sample sentences, and the odd nonsense word, and is meant to make for a “more intuitive approach to meaning.” (Try it here, read the detailed description here.) The two other entries were by Jason Sigal (“Phrases & Pronunciations“) and Sharang Biswas (“Love Poems“).
Then came two segments: “Remapping Syntax” and “Remapping Structure,” which pretty much means the same thing: Scraping text and rearranging it (sometimes processing it first). Among others there was Sam Lavigne’s “A Method and Device for Comprehending Theoretically the Historical Movement,” in which the Communist Manifesto was automatically re-formulated as a patent application; Salem Al-Mansoori‘s “Recipes of War,” which combined a cook-book with the Wikipedia entry on the Second World War; and Rajit Bhatnagar’s “Pentametron,” which scrapes Twitter for Tweets that, by accident, are formulated in iambic pentameter, and retweets them as rhyming couplets.
In the “Auditory” segment, there was Caitlin Weaver’s “Susan Scratched,” a Python Script that adds random repetitions to a text. Supposedly reminiscent of scratching a record, read out loud the resulting poem sounded more like Beat poetry. Aankit Patel’s “Dada Dial” is an emulation of John Giorno’s Dial-a-Poem, using UbuWeb’s audio database that still requires you to call a number – 917-534-6464 ext. 11 (explanation here).
In the “Interactive” section – a label that twenty five years ago might have been the definition of digital poetry – little stood out save for “Today” by Nick Barr, an iOS app described as “Fridge Magnets meets Asteroids meets hidden Markov models.” And finally, there even were “real” code poems, ASCII art inspired and possibly executable, but rightly relegated to the end, as they mostly presented in-jokes for programmers.
–
With techniques abundant, results varied. As Allison Parish noted during the event, comedic, or absurdist, works are easier to produce than “serious” or at least ambiguous ones; juxtaposition is easy to get electronically, and it also makes for easy jokes. The texts in which this tendency prevailed were also the ones resulting from a programming rather a writerly perspective. And the better texts on this evening were the ones toning down their digital heritage without obfuscating it completely, but sometimes resembling existing styles of the historic avantgardes.
The history lesson at the beginning also showed that the possibilities of code poetry are not at all exhausted, or even fully comprehended; not too much has changed since then, and one of the few major innovations might be what one could call big data lit, the ability to scrape immensely large amounts of texts, parse them and generatively process them. At the same time, wherever poets relied too much on the technical side of their text production, the outcome suffered. The best texts were mostly conceptual, like Levine’s, or mostly performative, like Weaver’s, while not showing off the codes and punchlines of coding culture but still retaining the sense of possibility of the digital.
–
One thought I had over and over during this event: Someone needs to show this to “analog” authors, ideally combining conceptual vision and technical prowess. “Python for Poets” would be good start.