How AI will deepen the divide between literary masterpieces and cheap reads

Generative artificial intelligence is thought already to be able to produce novels in a style similar to that of established authors

Will AI written novels become the norm? PA

If a thousand monkeys at a thousand typewriters could pen the works of Shakespeare, then the idea of an artificially intelligent software program creating a literary masterpiece could be just as likely.

Turn the clock back a few decades and the idea that machines could write novels would probably have seemed fanciful, a concept raised only in science fiction.

Today, this is reality. The capabilities of artificial intelligence have now developed to the extent that a bestseller could conceivably be written by a series of circuit boards.

Friend or foe?

When accepting a top Japanese literary award for her book Tokyo-to Dojo-to, set in a Tokyo of the future, the author Rie Kudan recently said that she had used generative AI to her benefit to help produce the book, which features the technology as one of its themes.

At the same time, established authors are up in arms that their books are being used to train generative AI programmes, giving the technology the capability to produce novels similar in style to the ones that they have laboured for years to produce.

Dr Kulvinder Panesar, an assistant professor of applied artificial intelligence in the Faculty of Engineering and Digital Technologies at the University of Bradford in the UK, says that the technology can “definitely” produce novels in the style of those that have already been published.

The “large language models” that generate content can, she says, assimilate and use vast amounts of data, such as all of a person’s published novels.

“That’s why these models are absolutely extortionately big and require so much processing power,” she says.

Prompt engineers

A technique called retrieval-augmented generation provides the large language models with extra information so that they can produce output of a particular type.

“You define this novel in terms of what you want it to look like with a series of prompts. [For example,] have an ‘Alastair Campbell style’ or it must have this many characters or be set in this background,” Dr Panesar says.

Such “prompt configurations” specify the type of novel or other output that is required, while subsequent work can be carried out to refine it.

“This prompt engineering is a job in itself. You might have a few more jobs coming out – writers who are becoming prompt engineers,” she says.

The use of AI in creative writing was a key issue behind last year’s Hollywood writers’ strike, which ended in September with a deal between the Writers Guild of America deal and the Alliance of Motion Picture and Television Producers

Hailed as a victory for humans over machines, the agreement allows for AI to be used as a tool, but not to replace human writers.

While wondering whether AI can portray the subtleties of human character in the way a human writer can, James Robert Carson, an associate professor and member of the Creative Writing Research Group at the University of East Anglia in the UK, says that with some types of writing, the target audience may notice little difference whether it is produced by human or machine.

“A lot of writers are producing prescriptive dramas anyway for television shows,” he said. “They’re, ‘Writing in the style of’. In some areas, maybe it’s not that much of a shift.”

Mimicking existing authors

Last year, a whole range of authors, including well-known names such as Margaret Atwood and Dan Brown wrote an open letter to technology companies saying that they should obtain their permission and pay them if they use their published works to train generative AI programmes.

Val McDermid, a popular Scottish crime writer, is among the novelists to have spoken out publicly over such unauthorised use of her books.

While AI may be capable of taking a writer’s back catalogue and producing additional books of a similar kind, Dr Kevan Manwaring, who teaches creative writing at the Arts University Bournemouth in the UK, says that it cannot replace human input.

Dr Manwaring is himself a prolific author, with several novels among the tally of more than 30 books that he has written, co-written or edited.

“I think artificial intelligence is a bit of a misnomer; I think we should call it aggregate intelligence, because that’s what it does,” he says.

“It only regurgitates or reconfigures what’s already there on the internet. It never invents, because it can never draw upon anything beyond the digital sphere – what we choose to upload.

“It’s only ever derivative, never innovative. It’s excellent at pastiche … If you want a Dickens knock-off, a Val McDermid knock-off, I think AI could do that perfectly well.”

It is a point echoed by Adnan Bashir, global lead for external communications at a software company, Hansen Technologies, who says that AI cannot replace “human ingenuity, creativity, reasoning and cultural nuance”.

“Some of the most resonant creative works of the last few decades mine a person’s lived experience, the depths of emotional complexity, as well as the intricacies of human history and world events. AI cannot be a stand-in for any of those elements,” he says.

“A novel, a play, a record, a film, are all artefacts of culture. In various ways, they embody the zeitgeist and the emotions of the society that engender them.”

He said that AI systems are prone to the biases of the programmers behind them and these may play out in the novels, plays, essays, poems or other works of art they produce.

For all its apparent creative limitations, however, with content now cheap to produce and distribute electronically, it seems likely that AI novels will become ever more numerous and easy to gain access to.

Dr Manwaring says that the generation of novels by AI is likely to be “happening on self-publishing platforms all over the place”.

He sees a “splitting off of the market”, with established publishers, who have reputations to protect, expected to continue to release high-quality, human-written books, while self-publishers will churn out cheap AI-generated novels in ever-greater numbers. He is in no doubt which side of the potential divide he would like to be on.

“To those without imagination, actual skills or real talent, AI must seem like the second coming,” he says. “People who have nurtured their little flame of talent, like myself, they want to use that. Like I say, I love writing. It’s my happy place. I don’t want to lose that.”