About    News    Events    Shop    Instagram    Contact

Tenement Press is an occasional publisher of esoteric,
accidental, angular, & interdisciplinary literatures.

‘My head is my only house unless it rains’

Don Glen Vliet

‘Were a wind to arise
I could put up a sail
Were there no sailI’d make one of canvas and sticks’

Bertolt Brecht, ‘Motto’
(Bucknow Elegies)

Rehearsal      /     6. Jorge Carrión
Translated from the Spanish by Peter Bush

Scenarios of the Imminent Future

(GPT5, Writing, & Publishing.)

GPT4 was born last week and there are already dozens of books in a variety of languages analysing the new technology, advising on how to make money from it or using it to write children’s stories or novels in dialogue form. This new publishing phenomenon began at the end of 2022 with ChatGPT3. And it looks as if it can only grow and grow.

If the crypto-world is in crisis and the metaverse has lost much of its momentum (Disney has just shut down its metaverse division, only six months after the virtual celebration of its centenary), generative AI is spurting cheerfully towards an ambiguous future.

As GPT4 is already able to write with literary style, sum up lengthy books or describe images, and is being integrated into search engines like Bing or writing and design programmes with content like Word or Power Point, it could be a rewarding exercise in plotting possible futures to imagine what will happen in the world of writing and publishing over the next months or years, when it evolves and GPT5 or GPT6 natural language modules have become part of our daily lives. Open AI has already predicted that professions linked to editing  and mathematics, are at risk. Let’s analyse possible scenarios, adaptations and mutations.

I use the abbreviations GPT as a generic term to refer to the neural networks or those who delegate programmes for generating language through deep learning that are constantly multiplying. I will explore ten key pointers to the changes that are almost upon us.

(Integration of GPT into literary writing.)

¶        If at the turn of the century we have already incorporated into the processes of editing and writing Google searches, automatic correcting, cut and paste and even the option of auto-completion, it more than likely that neural networks will soon become part of our professional and artistic routines.

Currently it is necessary to open the generative writing app in order to have access to them, but it is only a matter of months before they form part of the menu of Word and other tools for editing texts. That reality will open up a wide range of options for more or less creative writing.

Writers will be able to generate paragraphs, plots, examples, scenes or characters with their support and adapt them to their own style and the needs of the plot or the logic of their discourse. It might be a solution to classic writer’s block. And a rather powerful temptation. Some authors will go for it and squeeze out all the potential. Others will prefer to write and be disconnected to the internet. Or by hand.

It will also transform what we understand as brainstorming, the Eureka moment and the gestation of a project. Algorithms can conceive infinite seeds, possible stories, horizons to explore, that can be a stimulating trigger for a more than human imagination.

(Editing, rather than pure creation.)

¶        Those who allow most of their text to be generated automatically will cease to be just writers and will become designers and editors. Technology requires training in order to secure the best results, and these need subsequent operations of correcting, checking, adaptation and re-creation.

Literature has always been a remix, post-production, but that reality will now become more intense. Rather than a creator of original prose, a writer will be at once an architect and a DJ who draws up plans, calculates structures and produces new music from melodies and rhythms that are in part from elsewhere.

Authors whose ambitions are merely commercial, whether writing lyrics or noir novels, will find a great ally in GPT to help them calculate potential successful formulas or generate wholesale pages of dialogues or descriptions and thus be able to publish with the regularity the market demands.

Academic researchers, who must generate output as stipulated in their contracts, will also find a tremendous ally in natural language models. At the beginning of 2023 results are no longer so different when the writer is human. Paraphrase, quotes and technical editing place them on an equal footing.

(The translator as AI assistant.)

¶        Advances in translation are even more striking than in generative writing. Although more demanding literature, in prose and verse, is yet to be challenged, whether in the original version or not, it plainly makes no sense to type out word by word versions in another language of straightforward texts, without layers of meaning, metaphors or polysemy, as is the case of academic papers, technical articles, manuals, commercial bestsellers, self-help books or collections of influencer aphorisms.

With these genres, the translator will probably become checker, editor or quality supervisor. That is, a skilled expert who reads the automatic translations critically, looking for words or expressions that have been wrongly interpreted, incoherence and false data. And who revises the book’s final style. The human gaze is superior when it comes to looking at outcomes as a whole. At least, for the moment. That’s why the work of all those who embrace the whole process of publishing, who cherish it and secure its excellence is not threatened.

(The I+D, and other new publishing departments.)

¶        The fact that GPT and other generative AIs have open coding has meant that they are readily adapted to all kind of applications.

Be my Eyes, that connected non-sighted people to others who see in order to help them with their eyes, now uses GPT4 to recognise objects and help in an automated way; Duolingo uses it to practise foreign languages; Abbrevia.me builds your profile from the hundreds or thousands of messages you have published.

There is such a fast succession in advances in the production of text, image or audio that it seems right for companies to employ one or several people who can keep abreast of innovations and reflect on their practical application. It would probably be a new profession: for experts in generative AI who decide how to implement it in real time and can train staff to use it.

The creation of content for webpages and social networks is already being transformed, and the first impact is evident in the marketing and social medias departments of publishing houses. The success of ChatGPT3 was down to its conversational nature that connected to our ancestral need to communicate with each other and was established in the context of bots serving customers. The dialogues between publishing houses and authors with their readers is also going to change.

But GPT will perhaps be applied most of all to practical issues in the process of book production. The automated writing of readers’ reports will allow originals to be selected that really might have a chance of being published. The first draft of para-texts for a book, like updated author bios or cover blurbs will also be generated by AI. And I wouldn’t be surprised if large conglomerates don’t opt for a new department of algorithmic translation or for it to become part of the editorial department. At least for books whose subtlety and complexity don’t call for the commission of a literary translator. Secondary content, for promotion or to broaden the narrative scope, could also depend on that new structure.

(Personalised literature.)

¶        For some time digital technologies—both in production and in the printing or distribution processes—allow children’s books to be personalised, particularly when it comes to the names of the characters. GPT now adds many more personalised elements, as the existence of My Dream tales shows, an initiative to create and provide stories to the liking of their future readers.

In fact, the whole text can be unique, since in a matter of seconds or hours you can have coherent, interesting narratives created by generative AI that have been polished by a professional with names, surnames, cv, labour rights, and all those things we must never forget.

During the digital 21st century there had been a proliferation of apps for the self-design of books and firms providing editorial services. Although any user can have access to the technology, it goes without saying that many will have recourse to professional management to make their writing projects happen, with differing levels of human or artificial intervention.

Those texts will be part of a broader panorama of writing that is also personalised. If social networks or platforms now select information or publicity to fit the profile of the recipient, it is likely that language models will also be capable of varying their text to fit the recipient. And publicity will be inserted into the texts we receive, determined by our profile as consumers. I can image big brands paying large amounts of money for their products to appear in the stories, poems and content thus generated. From sexist or racist bias to advertising bias.

(Books written by dead authors.)

¶        If drastic makeovers rejuvenate the most renowned actors and bring the dead back to life, we can imagine the existence of neural networks programmed to write like the great authors of the past. Agatha Christie or Ian Fleming will be able to publish new tales of Poirot and James Bond. Several a year, even.

It’s no outlandish hypothesis, if we take into account how this is already a common practice with human writers. With the say-so of Manuel Vázquez Montalbán’s heirs, Carlos Zanón wrote a novel featuring the legendary detective Pepe Carvalho, Problems of Identity; David Lagercrantz has continued the Stieg Larsson saga. And Sherlock Holmes has resurrected hundreds of times after his death and his author’s. In those derivative fictions he often says ‘Elemental, my dear Watson,’ words Sir Arthur Conan Doyle never put in his mouth, and that is relevant if it is all about guaranteeing income. And posterity.

(Self-narrated audiobooks and other synthetic content.)

¶        As e-books are growing slower than podcasts or audiobooks, it is logical for resources to be invested in the industrial production of audio items of culture. Synthetic voices sound less and less robotic and more human. AI helps to process texts and reading and performance. Audiobooks read by actors now coexist with the self-narrated kind.

Moreover, it is easy to clone any voice, so the reading of texts can be automated. That means it is now technically possible for a machine to produce books apparently read by Emma Thompson or Constantino Romero. It is only a matter of time before a market of voice reproduction rights comes into existence. And dubious clauses have already begun to appear in the contracts of professional dubbers.    
But AI isn’t simply advancing in the fields of oral and written language, it is also happening in the fields of music, illustration, video editing or audio-visual production. In other words,  the creation of sophisticated content is getting cheaper by the minute. In a few years it will be possible to accompany the launch of a book not only with its e-book and audio-book and a marketing campaign, but also its graphic re-imagining, its soundtrack, videogame or film versions, all created in record time by AI.

In June, the first +Rain Fest in Barcelona will enable us to take the pulse of that new cinema. The sponsor is Runway, the company whose tools for generating scenes, effects and worlds are spectacular.

(Two parallel literary circuits.)

¶        If books and literature follow the path taken this century by the two games that AI has revolutionised, chess and go, two parallel circuits will soon exist. That of humans who read texts written by humans and those who become accustomed to reading those written by GPT and similar systems.

As ever, the most interesting developments will take place in hybrid, frontier areas. In books whose origin and writing processes aren’t obvious. In cyborg teams. In prizes that will be won by authors who have secretly used neural networks. In pseudonyms that will arouse our suspicions. Will the next Elena Ferrante be an algorithm?

Those hybrid readings are already happening in Spotify lists of replays that mix songs by flesh and blood musicians, with copyright, and lyrics generated by machines. In platforms and digital radios, selection and introductions are also beginning to be delegated to computer programmes.

Literature is the most conservative artistic language, yet at the same time it is in deep dialogue with all those realities. I can imagine poetry anthologies published and read by AI. Or Pessoa’s dream, the possession of vast numbers of heteronyms, developed by a model of natural language almost as gifted as the Portuguese poet. But much more extravagant.

(Ethics committees.)

¶        Microsoft has invested 10,000 million dollars in GPT and sacked 10, 000 employees. That numerical coincidence doesn’t augur well. The members of the ethics committee are among the sacked. Perhaps that’s the most crucial element: ethics must accompany the transition we are experiencing.

It seems obvious that big companies prioritise economic benefits. That’s why it is vital for states and international institutions to establish regulatory bodies.

While we wait for that to happen, the Juan March Foundation stands out as a possible model: it already uses AI for the analysis and managing of information in its DataLab, that comes together with its library or content platform, but it never implements a new direction without evaluating the potential impact on its hundred percent human staff.

(Legal metamorphoses and final justice.)

¶        Over the last  ten years systems like Turnitin, plagiary detectors, have become common currency in academia. They are now being updated to detect the use of AI in students’ work. They form part of a forensic turn that will influence all areas of representation, from deep fakes to journalistic texts. Although the law calls for the source of any digital item to be identified, those located outside the law will continue to proliferate. And we will have to create the tools to check them. If the science fiction magazine Clarkesworld has already detected a flood of stories written with ChatGPT3 and has had to sort that out, it is very likely at this very moment that thousands of stories, poems and novels are being entered for prizes that have been entirely or partially been written by AI. Detection systems must, from now on, form part of the evaluation rubrics of literary competitions. And of academic journals. And of all branches of publishing.

At the same time the approach to work written by neural networks or in collaboration with them must be normalised. The Japanese are leading the way with the Hoshi for science fiction, that is open to human and non-human authors. The Itaú prize for digital stories, given by the foundation of that name in Argentina, is now creating an avenue for AI. It will be the first in the Spanish language.

The concept of copyright will have to be reformulated in the new ecosystem. Language models have been initiated thanks to texts written by humanity, to the language that is our great heritage. The millions in profits corporations will gain from that must be returned to the creators of words and their legitimate owners. In the medium or long-term future, a basic universal income seems the fairest way forward.


Ten pointers according to GPT4,
generated after Carrión’s text was written.

1.                    The massive production of content ... GPT4 will be able to rapidly generate stories, novels and other literary texts in large numbers, and that will allow authors to create and publish more work in less time.

2.                    Help in creating characters and plot-lines ... Writers will be able to use GPT4 as a tool to develop original plotlines and complex characters and supply ideas when writer’s block strikes or by simply taking advantage of its capacity to generate fresh stories.

3.                    Correction of style and grammar ... GPT4 will be able to analyse and correct automatically grammatical and stylistic mistakes in a manuscript, thus improving the quality of the final work before it is sent to publishers, consequently reducing publishing costs.

4.                    Help in editing and correcting ... GPT4 can also act as a virtual assistant in the revision of grammar and style, and will help them improve ways of narrating or writing. Moreover, this programme will have the ability to come up with suggestions related to context and even the structural coherence necessary in academic research.

5.                    The creation of fresh, innovative language ... As GPT4 evolves, new forms of literary and linguistic expression will emerge that we have never seen before. This will permit an expansion in the repertory of words, sentences and idiomatic expressions available to be used by writers and content creators.

6.                    Personalised content ... GPT4 can personalise books and other materials written to fit the individual preferences of each reader, offering, for example, alternative endings, additional scenes, appropriate levels of complexity and character options depending on the needs of the user.

7.                    Interactive narrative ... GPT4 will facilitate the creation of interactive novels and games in which the reader can take decisions, thus improving involvement in the story and driving new literary formats based on the choices and explorations initiated by each reader or player.

8.                    Discovery of young and emerging talent ... AI will be able to identify authors with potential and put them in touch with the appropriate publishers and agents, thus giving new visibility to promising writers and helping them in their professional career.

9.                    The automatic creation of metadata and descriptions ... By automatically generating coherent synopses and precise metadata, GPT4 will facilitate the editing process at the same time as it will improve information offered to interested readers.

10.                    Authorial and ethical concerns ... As the influence of GPT4 increases in the literary world, it will ask questions about the degree of authorship and originality that can be attributed to humans or AI itself. That will open up interesting debates on intellectual property rights.

Jorge Carrión is a writer and the director of the Creative Writing MA at the Pompeu Fabra University in Barcelona. He has written regular columns for La Vanguardia and The Washington Post. Bookshops and Against Amazon have been translated into over fifteen languages. He is the author of five novels, the latest of which are Membrane (2021) and Every Museum is a Sci-fi Novel.

Peter Bush
is a translator. His first literary translation was Juan Goytisolo’s Forbidden Territory (North Point Press, 1989) and Bush has to date translated eleven other titles in Goytisolo’s bibliography, including The Marx Family Saga and Exiled from Almost Everywhere. He has translated many Catalan writers including Josep Pla, Mercè Rodoreda, Joan Sales, Najat El Hachmi and Teresa Solana. His most recent effort is A Film (3000 meters) by Víctor Català, the classic 1919 feminist novel set in Barcelona’s criminal underworld. Bush lives and works in Bristol.

︎︎︎    Back to Rehearsal


Tenement Press