‌4

Poetry

Tung-Hui Hu

It is sometimes said that the future will belong to the scavengers: the zama zama miners lowering themselves with a rope and a handheld flashlight into the abandoned pits of industrial gold mines, closed because there is nothing left for a machine to extract; the hobbyists pouring nitric acid over the gold fingers and wires of old computer chips; the gleaners collecting the misshapen vegetables and roots at the edge of a field, or at the dumpster outside the supermarket; the waste pickers salvaging cardboard for sale to recycling plants. It is not hard to imagine this, because such systems of scavenge have been formalised for so long – there is a cartonero union in Latin America, a Zabbaleen system in Cairo since the 1940s, mudlarking on the Thames for chunks of coal and metal in the Victorian period, gleaning before Leviticus codified it – that they are inseparable from the economy that produced the discards in the first place. Thus they are not, as one might initially expect, ‘post’ anything (post-apocalyptic, post-industrial, post-depression); the economy exists because it relies on scavengers to fuel it, whether by feeding workers who are kept impoverished or in the machines that rely on it.

Indeed, this is most obvious in today’s AI engines, which are simply scavenging old text or images from the internet, from posts in the 1990s and 2000s and 2010s, and recycling them into seemingly brilliant new forms of insight. Although today’s large language models take great pains to produce coherence, albeit in a saccharine, Hallmark-card kind of way, I was nevertheless reminded of the moment that poets were aggressively scavenging junk emails – spam, mostly, in the 2000s, producing poems that were deliberately awful in what was known as flarf. It was roughly around the same time that spammers would aggressively scavenge fragments of blog posts, digitised texts scraped from the Gutenberg project, and other detritus from the literary world to evade spam filters meant to detect if a message was written by a human or a robot. Flarf went away, to be replaced by today’s ‘internet novel’, where Twitter feeds and social media are both atmosphere and text of the novel’s typically white woman protagonist. But spam was the future, and it only ceased to be the future when it was replaced by its middle-brow update – AI, the spammers (and authors) themselves banished to the back room of the scam.

It’s the future of poetry that I’m interested in. Inspired by the gold scrapper and the cartonero, I predict that in 10 years, poetry will occupy a similar economic role. First, understand that there will be a wholesale restructuring of how literature will be produced, akin to the way that the programme era restructured literature in the second half of the 20th century. Poets, increasingly, are being hired to train computers to sound more human, given directives to mimic certain emotions in text (like cinema before sound: exaggerated, over-the-top, ‘universal’). This form of microwork will largely replace current funding models, such as publishing and public support, particularly for writers in the Global South. The limited exception – wealthy North American creative writing programmes – will be absorbed into a white-collar version of microwork, namely Schools of Emotive Content, organised around certain product lines: tearjerkers, inspirational, action/thrills, sex/cum/love, how-to/process, other. Poetry technique will be taught alongside other short-form video content. Despite this, some practitioners will use language to define a new space between feeling and the toneless, attempting to evade sentiment analysis by marking out something that comes after post-irony and post-sincerity. Tonelessness, the idea of feeling nothing (or really the ambivalence resulting from oscillating between these states), will become the new form of interdisciplinarity.

Due to an unhealthy fascination (or defensive nostalgia) for ‘human’ creativity, poets will be driven by intense pressure to speak in a ‘unique’ voice. The auteur, as it were, will come back into fashion. Yet in the past, the auteur has often become an auteur by simply iterating on a certain style while changing the subject matter. Warhol, Harlequin, and other factory-based models of artistic production paved the way for today’s self-publishing models in genre fiction, which produce literature on routine schedules, for (as movie studios would say about Marvel, or Star Wars) the purpose of iterating on their intellectual property. This will only scale up and expand, and poets licensing their intellectual property will become an important revenue stream. The irony is, then, that the writing that will be deemed most human will in fact be the most automatable.

But in a decade, the extraction of language by technocapitalism will be exhausted. What will be left are the discards: language deemed useless for operational reasons by a large language model for its statistical improbability and its failure to contribute to the model’s mass appeal. Poetry will be built by scavenging discarded language and sentiments that have been deprecated for being too maudlin or too simple (‘sincerely yours,’ against an expectation of authenticity). Older patterns in language will make a re-appearance – antiquated words that will have peaked and then fallen off the chart in Google’s Ngram viewer – but will be recombined with other Englishes from the globe, such as Singlish and Nigerian English. Even General American will not sound the same in ten years; linguists have already suggested that certain features, such as consonant clusters, might become simplified or dropped altogether, and so vernacular Englishes (think Paul Laurence Dunbar, Linton Kwesi Johnson) will no longer be anomalies but will be a way of giving body to a language that will have become increasingly disembodied. But these Englishes’ main contribution to poetry will not be in vocabulary or diction; it will instead be in prosodic patterning. To take up Singlish’s influence again, poets will begin writing in syllable-timed English rather than stress-timed English.

In that intervening, transitional period, two things will have happened: 1) AI will have taken over the professional uses of language in the corporate-manufactured chatbot or assistant and other ‘white-collar’ forms of communication, and 2) casual language will move entirely into text messages and other forms of instant communication. As a result, the roughly 60-year reign of casual language in ‘mainstream’ American poetry (for example, the legacy of the New York school) will be over. Poetry may well skew in two ways: first, towards operational ends, akin to the increased reign of what Harun Farocki identified early on as ‘operational images.’ This would be not just for emotional content, as indicated above, but also for minoritarian purposes, as in the songs from the mid-19th century that Simone Browne describes as countersurveillant songs of warning against oncoming slave patrols, or as forms of solidarity and resistance by coal workers in 1970s Britain. Reciting a poem may be the new CAPTCHA test for in-groups, or, like the oral ruttiers mariners used to remember navigation directions, a covert form of storage.

A second branch will skew away from these communicative ends and towards the cryptic. Fragmentation or obscurity or other characteristics of the avant-garde will not necessarily move to the mainstream, however. (For example, where it was once a radical act to appropriate material from the archives – e.g. by erasing or blacking out language – the regime of extreme technological appropriation will make this act mundane.) Instead, to understand my use of the word cryptic, think of the current use of homophonic memes to evade censorship in China, using one word that sounds like another word (a name of a forbidden politician, for example), the next meme springing up as quickly as the last one becomes censored. This branch of creative writing will be like the informal worker in a city’s surrounds: tolerated, even essential, when it flies under the radar, when it doesn’t get too big. There will be newly invented poetic forms (and intricately sequenced books) – both are already happening, of course, but the ease by which language can be made programmable will drive poets towards intricately designed, even algorithmic constraints, visible primarily to literary critics who will need to reverse engineer these poems and sequences.

After the exhaustion of literary style and language in 2034, there will be an uneasy coexistence between poetry and the technocapitalist machines that it (indirectly or directly) helped feed in the previous decade. Poetry will survive in part by serving a primarily archaeological function, like the mudlarkers that dug and continue to dig up coins and other fragments of heritage in the Thames mud for sale to collectors. Poetry will be about gleaning what was left of that once prized concept, ‘human literature.’ The year 2024 – the last year that the volume of human-generated content exceeded robot-generated content on the public internet – will serve as a central locus for this nostalgia. Despite the genocide and the increasingly unliveable temperatures of that year, technologists and scholars will look back on 2024 as a final year of creative flourishing and calm – a tipping point – as when (to use another antiquated phrase, referring to incandescent lighting) the lightbulb burns brightest before it flames out.