Google’s dreamscapes – the product of an artificial neural network being asked to amplify and pull patterns out of white noise. Photo credit: Michael Tyka/Google
In 2011, one of the longest-running student-run literary journals in the USA – Archive at Duke University – ran its annual call for poetry submissions for its Fall Issue. The editors, shifting through the reams of poetry, stumbled upon a short poem called “For the Bristlecone Snag”. It was environmentally themed. It struck a slightly aggressive tone. It contained a few of those clunky turns of phrase that can so often be found in student poetry, including the less-than-immortal line: “They attacked it with mechanical horns because they love you, love, in fire and wind”. Regardless of these slight failings, the editors of the journal decided to run with it. An unremarkable decision and an unremarkable nine-line stanza at first glance, except for one thing: the poem was written by a computer algorithm, and nobody could tell.
Of course, it remains too soon to predict when the TS Eliot Prize will be won by a robot. However, what it could mean for the future of poetry – and writing in general – is gradually gathering a great deal of attention, and stimulating significant discussion.
It’s important to point out that Bristlecone Snag is not the only example of machines writing poetry. In 2008, a US high-school student, Sarah Harmon, used Java to create a computer program that wrote poetry. Again, she submitted poetry created by this machine to student journals. And again, the submissions were successful.
There is nothing fancy about these machines. They are not magically complex. They are simple algorithms built by simple tools. They follow predefined rules of grammar and structure to compose poetical-sounding snippets. For example, Harmon’s poetry machine – named OGDEN – came up with the refrain: “He was perfectly strange,/His world was shyly hopeless,/Then he tasted his dreams.”
Image via shutterstock
These more recent examples are nothing new, either. In 1984 one of the first computer bards – Racter – wrote prose largely at random. It produced a book of poetry and surreal dialogues called The Policeman’s Beard is Half Constructed.
But is it surprising that simple coding tools and skills can be used to create poetry that readers find passable? After all, William Carlos Williams wrote that “A poem is a small (or large) machine made of words”. A nice, simple statement of a poetic position. But also one that picks up on the essentially formulaic aspects of writing. If there are reproducible structures and characteristics – as one would find in any industrial machine or piece of new technology – then it stands to reason that computers can do just a good a job at recreating patterns and writing their own poetry as human beings.
Image via Shutterstock
Just as aspiring writers will look to the poems and novels of their favourite authors, and are able to identify similarities of style and structure that they can imitate, it does not seem unreasonable that digital programs are able to identify the same patterns and imitate them. After all, Booker-nominated author Will Self said of creative writing courses that they are a like to working from “a pattern book”. If such formula can be taught, it can just as easily be programmed.
But what next? Can machine-written poetry ever go beyond simple imitation? Can a computer ever be creative in and of itself? Can it ever create lasting poetic expressions that stand the test of time among human readers without having any examples of real, lived experiences to draw on?
And, perhaps a more pertinent question, would we ever want any answers to these above questions to be ‘yes’?
Ever since the Luddite machine-breaking rebellion 200 years ago, advocates of ever-advancing technology have learned to scoff at technofobes. The argument goes that machine efficiency allows resources to go further, so what does it matter if workers are displaced?
Such an attitude has held firm as industries like coal mining, agriculture and banking and finance have seen miners replaced by coal-cutting machines, farm labourers by tractors and combine harvesters, and bank clerks and analysts by computerised ledgers and algorithms. Although of course we all know that this last one has not been without some teething problems.
The digital Shakespeare
Yet as IT systems and ever-more capable artificial intelligence evolve, is it truly desirable to have so many aspects of humanity computerised and automated? Do we want to read poetry and novels written by machines, as writers huddle together in the last vestiges of hipsterism in some dusty London cereal café pining for the old days, trying to remember what pens, pencils and paper were called? And will we come to exist as those humans depicted in Pixar’s WALL.E – utterly reliant on automation for sustenance and entertainment, and unable to think for ourselves?
Quite what poets like Blake – who envisioned an England of “dark, Satanic Mills” at the face of the country changed with the advent of the industrial revolution – would make of computerised poetry remains unknown. Though it’s probably possible to at least take a rough guess about his feelings.
George Orwell’s 1984 envisioned a world in which we have already reached this point in history. Here, the “proles” are entertained by books produced by machines. Perhaps unfortunately – depending on your point of view – such a future may not be far away.
Professor Philip Parker, of Insead Business School, has created software that has generated 200,000 books, with over 100,000 of these titles available on Amazon. He notes: “A computer works very well with rules and the most obvious way is poetry.”
“We did a blind test between a Shakespearean sonnet and one that the computer had written. A majority of people surveyed preferred ours,” Professor Parker added. “That’s not to say it was better, but it was what people preferred.”
Writer as algorithm
The algorithms at the heart of Professor Parker’s software have also inspired a new suite of writing software that threatens to compete with journalists for the already minimal numbers of jobs going within the news and media industries.
Startup company Narrative Science creates articles without a human doing the writing.
With 30 clients for its articles already, written automatically by a machine collating data and writing “rich narrative content” from it, the death of the journalist has been mentioned in more than one speculative column.
Business news site Forbes is using the service for a number of pieces each weekday.
More questions than answers?
What this illustrates is the extent to which digital technology represents a force of change for writers of all ilk and forms. Some writers will no-doubt realise potential opportunities created by the emergence of new technologies. Think, for instance, of Iain Pears’s new novel, Arcadia – a 600 page hardback that works in close conjunction with an app of the same name. Or else Melville House’s line of “illuminated” novels with QR codes that lead to extra digital content. Or alternatively, Picador’s “The Kills” – a 2013 “digital first” thriller that links to online films from the characters’ points of views.
But perhaps an issue with these examples is that they all utilise technology under the assumption that the human being remains in control. Here, poets, novelists, publishing houses and media groups embrace these tools as enablers, but do not consider where the future is heading. How long before all the news stories we read have been written by machines? How long before we are all reading pre-programmed novels created by robots? How long before studentds are studying the poetry of AI-8976R, or the HAL-9000, instead of Blake and Shakespeare? And what would this mean for our culture?
These questions remain purely hypothetical. Yet as technology develops, we must begin to consider how we can answer them.