There is no lack of irony in using artificial intelligence to write a novel and calling it Death of an Author, especially when the man behind the bot doesn’t think the title is prescient.
Of course, when he’s not getting novels out of ChatGPTStephen Marche is writing them himself, so he has a vested interest in seeing AI not take over the literary mantle hitherto exclusively claimed by humans.
It’s an experiment, and one with impressive results. Called ‘the first half-readable artificial intelligence novel’ by the New York Times, far from being the newspaper’s most scathing review, it is in fact an easy read, revolving around the story of Canadian writer Peggy Firmin, whose murder the literary critic Gus Dupin intends to solve.
No spoilers, but there’s also a huge AI company.
They say write what you know, but it’s the words, not the story, that have artificial origins.
“I knew that AI created the most banal plots, terrible plots,” says Mr. Marche, who has written extensively about artificial intelligence. ‘I wasn’t going to use that. I had a conscious process of thinking it through and taking a lot of notes and figuring out the narrative structure and from there using the machines to build it.
Full disclosure, Death Of An Author isn’t Mr. Marche’s first attempt at writing using AI. Described as 95% by AI and 5% by Mr Marche, it is the former that receives the most credit, published under the pseudonym Aidan Marchine.
“I’ve been using it to write fiction since about 2017,” he says. “I wrote a short story for Wired called Twinkle Twinkle, for which I had a computer scientist create an algorithmic mode of compositing called SciFiQ. I then wrote another one for MIT Technology Review, and in 2020 wrote a short story for LA Review Of Books that was 17% computer generated.
“Then about a year later, I wrote Autotuned Love Story, which was built out of chatbots for Lithub.”
If it seems Marche is accelerating its own redundancy and that of writers around the world, a closer look at the process provides some reassurance, if only temporarily. It is complex, not yet feasible without a human in command.
“I mainly used ChatGPT with Sudowrite, most of the text was created using those two technologies together,” says Mr. Marche.
“I gave ChatGPT very specific instructions, including grammar, syntax, variability in sentence length, tone, that sort of thing, and then followed up with ‘containing the following information’.
‘That would then create some text, which it would cut, place in Sudowrite, where you can select text and have it shorten it or add detail. It also has a customization button where you can tell it to do whatever you want, like make it more active, make it more conversational, make it sound like Ernest Hemingway, which I did a lot.
‘I used a lot of writing styles that I wanted to imitate, so I wanted him to do that. That was most of the text.
That may also be the bulk of the argument. Until we get to a point of artificial general intelligence, without human intervention, chatbots can only create content based on what has come before them.
Sure, that’s literally Everything on the Internet, making it easy for them to imitate authors, styles, and stories, but these days, without a guiding hand, the raw results aren’t ready for the real world. Not least due to an excessive abuse of fancy and decorative identifiers and adnames.
In other words, sentences that are too descriptive and a bit difficult. Think about that episode of Friends where Joey gets a thesaurus.
“What ChatGPT writes is quite banal,” says Mr. Marche. ‘To get it to whatever state it can be original you essentially need to filter it through a series of stylists and guides, which is Sudowrite.
“Then when I wanted really good individual lines, for metaphors and things like that, I used Cohere.
“In that, I’d create prompts like ‘make a coffee image,’ then I’d train that into a series of metaphors for coffee, and I’d give myself images and cycle through them until I found one I liked.”
If this doesn’t sound like the easiest or most efficient way to write, know that all of these programs will get better. But, says Marche, the experiment was never about efficiency, even if he was able to produce the novel in a matter of months.
“Certainly writing with AI is not a matter of efficiency,” he says. It doesn’t necessarily make it easier.
However, should he and other writers be concerned that it was his agent who pitched it? Chances are, they spotted a niche idea and timed it perfectly, working in tandem with a fully qualified author for the project.
What the AI novels mean for future generations of readers, however, remains to be seen, even if technology isn’t entirely to blame. English and literature subjects are plummeting in popularity at universities across the UK, a harbinger of doom for the next generation of authors and publishers as they see an industry destined to be swept away by technology?
Or will AI allow more people to become authors, finally getting the novel within them onto the page or screen?
“In this creative way, I’m not sure the term author is appropriate for what I did with this material,” says Mr. Marche. ‘For one, I am the creator of a literary artifact, so obviously we would call it an author.
But on the other hand, I didn’t create the words, and that’s another way we think of an author that probably doesn’t make sense when it comes to this text.
‘But the idea that this is going to replace creativity is simply not true.
‘This requires creativity, it requires human intentionality. It’s really an incredibly powerful and wonderful tool, but whether it leads to the death of the author, I think it’s actually more of the author’s transmutation.
‘Everything becomes authorship, and literary creativity will take a new form.’
The death of an author is a Pushkin publication. Available as audiobook and ebook