How appealing could a novel written by artificial intelligence be?

That might seem a question for the future. As I write this there aren’t many AI-generated novels around. But there are some, and soon there will be many more. So if you encountered one that appealed to you, would you read it?

Cards on the table: if I heard about an AI-generated novel that was really convincing and was written in a genre I liked, I can easily imagine myself reading it out of sheer curiosity. But this would not be because I thought this is how novels ought to be created; far from it. As a novelist myself, I would simply want to know what I was up against.

Just how good, I would be wondering, was the quality of the AI writing? How sophisticated was the character development? How well was the plot worked through? How correct, how consistent, how idiosyncratic, was the prose? Would I end up with a sense of who the author was, and a feel for his or her vision of the world? Would that even be possible?

My worry, of course, is that the book might pass all these tests and read like something that had emerged directly from a human mind. If it did, where would that leave real human authors like me?

The power of artificial intelligence

I’ve touched on the power of artificial intelligence in my latest novel, The Inducement, which should be published soon after I post this; but I can already see that I will want to return to the subject in more detail in another book.

It’s worth bearing in mind that broadly speaking, artificial intelligence still needs a helping human hand if it’s going to be effective. In theory you could instruct an AI program to write a novel simply by issuing a few lines of guidance; but in practice you would probably fine-tune its suggestions, giving it a steer by providing clues about subject, settings, characters, scope, pace, direction and outcome.

A literary cyborg?

However, in a way this is more worrying than the notion of a book written with virtually no human involvement at all. In practice, we’re looking at a world where the underlying vision of a work will still come from a human author, but the execution will be done by AI, which will churn out the actual words. The outcome will be a weird kind of hybrid – a literary cyborg.

This puts me in mind of various existing best-selling authors, who come up with the basic concept, then work with collaborators who do the hard yards of actually turning their ideas into prose. James Paterson, for instance, is known to take this approach. And historically, there have been strong parallels in the world of art, where masters such as Rubens ran entire workshops of artisans to flesh out the detail of their paintings.

But at least creators like these have used humans with training, skill and experience to flesh out their ideas. What will happen if, instead of using human associates, writers of the future increasingly rely on AI to do the real grafting? Where will the humanity be then?

Recombining and recycling

You might argue that if the prose reads confidently and convincingly, there’s nothing wrong with it. But as many commentators have pointed out, AI will be basing its output on the efforts of countless existing writers and their work. It will be presenting what an author might typically write in this sort of situation, not what a real live author is writing now, in an actual situation. It will be recombining and recycling, not striking out and innovating to produce genuinely new work.

Perhaps we should be glad. If AI really does end up thinking for itself, and comes up with what looks like startlingly creative, ground-breaking new writing, where will that leave mere mortals?

By the way, if you think this sounds like cloud cuckoo land, you haven’t been paying attention. Artificial intelligence can already write frighteningly plausible, convincing prose that looks indistinguishable at first (and second) sight from human-generated prose. The grammar and syntax of the writing will be correct, and it will have natural-sounding rhythm and flow. This kind of thing is now ubiquitous: just look at Chrome’s AI answers to browser queries, or Amazon’s AI summaries of product reviews.

Weakness in content, not composition

If there are weaknesses in current AI-generated text, they tend to lie in the content, not the technique. Analysts will point out that AI writing tends to be general in scope rather than specific; that it equivocates; that it sometimes falls back on clichés. Maybe so, but it will get better at hiding or eliminating these faults, and if you’re not looking hard enough, you might not notice them anyway.

It’s worth reflecting on who will benefit most from AI in terms of novel-writing. It’s going to be people who are poor at spelling; who make mistakes in grammar; who can’t write fluent, rhythmical prose; or who simply can’t type very fast. Clever AI promises to be a leveller. It will allow people who never had a chance of competing at novel-writing before to get into the game, and perhaps come out ahead – potentially leaving writers who are good at these things struggling to keep up.

So far, organisations like Amazon have tried to resist or at least flag up AI-generated content in books, but who knows what will happen in the future?

Not a luddite if you oppose it

One could argue that the advance of AI is already unstoppable, and that the people fighting it are simply luddites. I can’t agree. The textile machines the actual Luddites in Nottinghamshire feared weren’t emulating human creativity, they were merely outpacing humans in terms of replicability and speed of output. By contrast, AI-generated fiction is reducing creativity to a commodity that can be weighed, measured and served up on a plate. Not the same thing at all.

Don’t get me wrong. I accept that AI can be a wonderful force for good. But if we accept it without question in the arts, we’ll be compromising the opportunities for future human creativity, and dumbing down the work we all end up consuming.

While doing research for this post I watched several YouTube videos of people analysing AI-generated writing, and pointing out how to recognise it. There was an extreme irony in the fact that in each case, the video was interrupted several times by advertisements for products that could help people to create AI-generated text: the very thing the videos were warning us about.

The power of the algorithm.