AI or not? – Tracy Durnell's Mind Garden

Skip to the content

Menu

Search

Search for:

Close search

Close Menu

I don’t like this game.

I read a book this afternoon that had potential genAI flags. Nothing I found definitive… but the creeping doubt undermined my reading experience.

It doesn’t even matter if I can’t tell the difference for sure — it’s not about quality — I simply don’t have any interest in reading generated text. For me, reading a good book offers a dual pleasure of the story itself, and satisfaction in seeing the execution of artistic skill . If you’ve ever watched Olympic figure skating, think of that moment of anticipation when the announcer alerts us that the skater is about to attempt a challenging jump, and everyone watching across the globe holds their breath in unison, waiting… then we all exhale together, either in relief or disappointment. When I am reading a really good book, I sometimes get that feeling, hoping hoping hoping that the author will stick the landing after a gorgeous set-up. In that moment, viewers are with the skater, readers are with the author, our shared humanity thrilling at the expression of excellence. It is a fucking joy to see someone who has worked hard accomplish the challenging thing they set out to do.

"The point about a good novel produced by a human … isn’t that only a human could ever have produced it. It’s that a human did in fact produce it. There really was another thinking, emoting consciousness at the “other end of the line.” And when you consume the work, you enter into a kind of relationship with them." — Oliver Burkeman

I am also OK with reading flawed and imperfect works — it is interesting to me to think about how story issues could have been resolved. Generated text is a shortcut for people who don’t want to put in the work of learning how to communicate in prose. I’d rather read work by the person trying, even if they’re not that good yet. I’d rather read a person taking a swing at something interesting, even if they’re not skilled enough to pull it off yet — maybe even especially if it’s at the edge of their abilities.

I like giving debut authors and new-to-me self-pub authors a try, but I want to be able to read a book by a new author without doing a bunch of research.

The book in question was Accidentally Fearless by Sapphire Dixon. The premise was fun, and the execution was competent — I’ve certainly read worse self-published books. The main flag for me was that the writing is noticeably internet-y (a la LinkedIn and Substack… and genAI 🫤). But that could be explained by the writing style being popular online, so anyone who reads and writes a lot online could pick it up.

Why it might be genAI<br>Why it might not be genAI

genAI cover

online writing style — lots of triads, punchy, ‘not x. not y.’

some unusual metaphors e.g. reheated pad thai tasting ‘like Tuesday’ (I actually do like this) and a customer service smile fitting like a ‘rented mask’ (don’t like this)

no author presence

no author bio on Amazon or Goodreads

no afterword by the author

other authors by the publisher have been accused of using AI

got this book for free from a giveaway of self-pub books that looked overall low quality based on cover art (though I heard about the giveaway through an author email newsletter)

the book was part of a coordinated giveaway but the author doesn’t have an email list, which is typically the key promotional practice in the sub-genre

front material explicitly prohibits genAI training on the book

the inconsistencies seem more the result of poor editing / inconsistent worldbuilding than generation

snappy pacing (perhaps a bit fast, especially on the relationship, but not out of line with the subgenre)

didn’t notice any unusual words / strange wording choices

followup series titles are listed in the book but have not yet been published three months later… if it were genAI, wouldn’t they have just thrown up the sequels already?

several people have left what appear to be real reviews (not just for ARCs based on timing) enjoying the book and recommending it in Reddit threads

Almost anything I detected could be explained by being a debut author doing this as a side hustle. Can I — should I — give new authors the benefit of the doubt at this point?

404 Media described this pre-print paper that identified some common differences between genAI and human writing:

AI tends to describe emotions as physical sensations rather than identifying the emotions outright,

humans tend to use more complex structures, incorporating flashbacks / moving through time more, and

AI tends to be heavy-handed on theme.

In this book, I didn’t notice emotions being spelled out constantly; as a novella this followed a pretty straightforward structure; and the theme was not hammered overmuch in the story itself, though the book ended with what I interpreted as the author’s dedication (which related to the theme) placed in the wrong spot in the...