Is this it, at last? The brilliant book written by AI?
Probably not, though the headline at The Bookseller does raise hopes. It reads ‘Genius novel’ sold in six and seven-figure deals at auction pulled over suspected AI use.
The Mirror reports the story without the need for a subscription. Having been hailed as a work of genius, the book is suddenly unpublishable! The identity of the author is being withheld (though he’s now being identified as male and Nigerian), but The Mirror reveals that Deadline has said the title of the book is Call Me, I’ll Hide the Body — not perhaps a title of genius.
The author is reported as saying
“The story, characters, scenes, humour, structure and voice came from me, and I have time-stamped drafts, written notes and dated Whats-App messages documenting the book’s development. I am also concerned that AI-detector scores are being treated as conclusive evidence.
“I tested passages from books published in 2012 and 2017, and the software classified them as 98 and 99% AI-generated, despite their predating the widespread availability of generative AI.
“Within six hours of the rumours, I lost US, UK, translation, and screen opportunities without being given a proper chance to explain my writing process.”
What I’m finding difficult to understand is how a book which on Wednesday was worth an advance of more than £1.5 million, can on Thursday become a piece of garbage. Are there really publishing companies willing to bid more than $2 million at auction for a book when their employees haven’t even bothered to read the manuscript at all carefully? OK, auctions are exciting, but if you’re going to bid that much you’ve got to believe the book’s good enough (which mainly means salable enough) to sell many thousands of copies. How can professionals be so wrong? Or were they not really wrong, and this is just another of these knee-jerk rejection-of-AI stories? Apparently the book which was so recently brilliant is now adjudged to contain “many of the hallmarks of AI, including negative parallelism, off-kilter metaphors, and a flat prose style.” — How can “flat prose style” ever have been seen as consistent with “work of genius”? The author’s agent claims that “The book was stunningly good, everyone fell in love with it.” which I guess suggests that someone actually did glance at it! Egg-covered faces both here and in England.
The publisher involved is Minotaur, an imprint of Macmillan/St Martin’s Press, part of the Holzbrinck Publishing Group. (It was Hachette who recently got into this kind of knicker-twist with Shy Girl.) We need, as an industry, to get beyond this all-consuming terror of AI. The test needs to be not “Can an AI detection program find AI evidence?” but “Is the AI use damaging the project? Is the book any good?” Hugh Howey weighs in on this story, suggesting that publishing is about to move on to a new paradigm. AI and publishing obviously got off on the wrong foot with the argument over the “theft” of copyright materials to train the LLMs, and this negativity is clearly spilling over into our current editorial attitudes. Time to get over it and move on. Using AI isn’t wicked. Palming off raw, unedited AI slop output and pretending it’s the result of a “normal” writing process may be. The sooner we start behaving like rational agents, capable of making aesthetic judgements, the better. A book isn’t axiomatically good because it was written using a Mont Blanc fountain pen. It isn’t axiomatically bad because AI had some role in its creation. It’s good because it’s good: it’s bad because it’s bad.
Making Book 
Quality notwithstanding, there are significant concerns over claims of copyright on work that’s overly AI-generated without significant documentation of human crafting and refinement. I don’t think it’s surprising that publishers are skittish about offering an advance on a book they may not even be able to make a copyright claim to if litigated!
Okaaay.
I hadn’t thought of that. I do remember the Copyright Office declaring against the copyrightability of AI-generated materials. But their objection is to straight AI-generated stuff, unedited by human hand, isn’t it?
Their report from January this year states
“The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output.”
• “Copyright protects the original expression in a work created by a human author,
even if the work also includes AI-generated material.”
• “Copyright does not extend to purely AI-generated material, or material where there
is insufficient human control over the expressive elements.”
While of course none of us can know for sure just what went on with Call Me, I’ll Hide the Body I can’t believe it’s “purely AI-generated material”. If it’s as good as claimed, that kind of supports that contention.
We face, it seems to me, a clash of two conflicting assessment systems: the law which has to be 100% fact-based, and lit criticism (I suppose) which has to be based on opinion. I can’t believe any lawyer couldn’t drive a Mack truck through any claim that this or that work wasn’t copyright because of AI use. And who’d be bringing the lawsuit anyway — there’s no damaged party is there?
Thanks for making me think though!
Yes to all your points, which is why an author would likely be wise to keep an audit trail of what refining prompts were used to tailor the final output. Case law in the US is slim so far, so publishers seem to be glacially cautious, as is their wont!
Of course. But I read just the other day that audit trails can easily be faked by AI. Can’t win! But of course you’re right —- it’s just that proving one’s innocence has always been a difficult task. Proving guilt is no doubt also hard in this context — but can we really rely on a presumption of innocence?
This whole area will probably have to be subject to legislation — if only legislation was something we did any more!