Finished The Serpent in the Grove

I have finished The Serpent in the Grove this morning. I ran its last 450 words through the Oratrice Mecanique d'Analyse Cardinale, whose verdict is: 100% AI-generated, with high confidence. 

I've read a few of the critiques that people on X have made of the style of the piece. Comparing it with the style used in Shy Girl, with the AI tells that both share.

One of the more interesting comments that I read was from a Madame Fragonard. She takes issue with the "conceptual errors" in the story, like with the "descriptions of voice and laughter." 

I will not look at the screenshotted passage from Shy Girl. I cannot comment meaningfully about it, because I have not read Shy Girl and have no intention of doing so. But one highlighted example from Serpent is "she had a laugh that shook dust from joists". 

That was an image from the story that I actually liked and that made immediate sense to me. This large, boisterous woman who could laugh so hard that it shook dust from joists. The idea that no human author can write this without casting suspicion on themselves is irksome. It's comments like this that make all the AI hysteria turn into a witch-hunt. 

And that's precisely why Granta did say that it is impossible to prove AI plagiarism. It presupposes that no human could have written such a passage. That is a dangerous assumption when we already know that AI "detectors" are systemically biased against non-native speakers of English. I also feel irked because I tend to reflect on the meaning of my own words. I can easily enough come up with a phrase or description that makes sense on reflection but is not idiomatic, so I can see myself in the firing line. 

It's one thing to criticise bad style, like the overuse of certain words or constructions. I read an utterly apoplectic article about a guy who was obsessed with the phrase "It's not x, it's y". But to start picking on metaphors and descriptions because you have to think in order to interpret them? That's placing limits on human creativity. Fragonard's comment that "We have unspoken intuitions around how voices and laughter are supposed to be described" only underscores my concerns. Why would I want to pander to my own prejudices about how anything is "supposed to be described" when I read creative writing? 

And for that matter, why do I want to pander to other people's prejudices when I write creatively? A characteristic of literary fiction is that you have to think in order to understand and interpret it. You don't necessarily get all the symbolism immediately. A person who writes a passage does not necessarily mean for you to understand it when you first read it. But over a multiplicity of readings over time, a multiplicity of meanings can be opened up by the same passage. These meanings can be legitimately different both between different readers and between different readings by the same reader. And I do not want to feel that I have to write a description that everybody can understand the first time. 

Another silly comment comes from Rohit. He thinks it's "embarrassing" to ask AI if something is AI, as if AI detectors are not the catalyst for most of these controversies to begin with. The controversy about Shy Girl began in earnest when Pangram reported that it was 78% AI-generated. And what was one of the first things that people did when Serpent got published on Granta's website? They started posting screenshots of Pangram's analysis that the story was 100% AI-generated. It's only normal for most people to run suspicious text through an AI detector, and Granta is no exception to this. 

Conversely, I strongly suspect that authors will get left alone if their writing passes an AI detector test. I'm thinking of the opening page of Elphie by Gregory Maguire. It certainly felt like AI-generated writing that I had read in the past — which is by no means an insult! So, I ran it through the Oratrice Mecanique d'Analyse Cardinale. Its verdict was that the passage was 100% human-written, with high confidence, so he's safe.

For my part, I loved the short story. The only thing is that knowing that the text is probably AI-generated makes me not want to reach too far for meaning in the metaphors or descriptions. Beyond a certain point, I don't believe that the meaning is likely to be there to be found in the first place. I just hope that the proliferation of AI doesn't make lazy readers of us all in this way.