Oh boy. I have real sympathy for this woman because she simply told the truth, or at least her understanding of it, and now she’s in the free-fire zone for the robot wars.
Her name is Rie Kudan. She’s a Japanese writer who won this year’s Akutagawa Prize — the final Boss level in Japanese literary awards. Her position is pretty nuanced and sounds very human to me, but the media can’t help but reduce it all to numbers.
“Five per cent of this book was written by a robot!”
And her publisher and Kudan herself have been qualifying her original comments at high speed because it’s all anybody can talk about, and some of the talk from the anti-robot alliance is kinda murderous.
It’s all very sad-making, and I say that as someone whose own books were fed into the insatiable maw of the training sets for the Large Language Models without permission or compensation.
I’ve reached a point where I’m not at all bothered by the idea of being replaced by AI. I just don’t think it’s going to happen in my working lifetime. (Another 10-15 years, maybe?) For sure, Amazon is already being catastrophically flooded with AI-generated garbage books. But the thing is - they’re garbage! You can tell as soon as you open them, and you can do that for free by grabbing the sample chapters. The human touch that makes a book worth reading is still very obvious on the page and even more so when it’s not there.
The thing that Kudan said that I suspect really unsettles some people was the admission that AI can help with ‘weakness’.
For instance, I have a weakness with commas. I’m fucking hopeless with them. Cleaning up the commapocalypse I leave behind on every draft is exactly the sort of work I’m happy to leave to a robot. So too, with basic subediting and copyediting tasks. That’s just data management, and that’s what computers are really fucking good at.
Would I still send my books off to be copyedited? Fo’ sure. Again, that human touch on the page is really obvious when it’s missing. But if I can make my copyeditor’s job easier, I will. That way, they can concentrate on whether a particular sentence reads well, whether it’s beautiful if it needs to be, and not just whether I’ve fucked up some syntax. (Spoiler, I almost certainly have. Multiple times).
Kudan’s novel ‘Tokyo’s Tower of Sympathy’ was described by the awards jury as an “almost flawless” work, and it contains a passage where the architect narrator has a conversation with a machine about designing ‘an elegant high-rise prison in the center of Tokyo’. So it’s obvious why Kudan would have been interested to throw the convo to a word-bot, even as an experiment-in-draft.
I’m not sure what weakness in her writing she was talking about when she gave her famous interview about using an AI writing tool, but I was listening to a publishing podcast yesterday and heard an almost perfect example of a writer using one to directly compensate for a neurological weakness.
The writer was Chris Banks, the developer of ProWritingAid, a Grammarly competitor. Banks wrote the original code for PWA to help himself with his own books, but he’s since turned it into a pretty reasonable business. (Full disclosure. I have a subscription, but I never actually use it. I find that MS Word’s editor and Grammarly pick up most of my comma-based atrocities in the first sweep.)
The really interesting thing about Banks, however, is that he suffers from that weird neurological condition where he has no visual imagination. If you tell him to close his eyes and imagine a green apple, he can’t see anything. He could have been holding a green apple just before he closed his eyes. But when he tries to imagine it? Nada.
As you might imagine, this could be a bit of problem for a writer who has to describe a scene on the page. When I write a scene, I can imagine it vividly. I can sit in that thing for hours, imagining every granular detail. I remember interviewing George RR Martin about the moment he imagined the first scene in Game of Thrones. He was sweating in his kitchen in Texas (I think) and he imagined himself into a frozen fantasy land to escape.
Banks can’t do that. Nobody with his condition can do that. But an AI can. A writer like Banks can prompt an AI like Claude or Chatty with their ‘concept’ of a green apple, and the bot can then describe it. Not well. But a helluva lot better than Chris Banks can imagine it. And he can then edit that description to make it betterer and more humany.
And who the hell am I to tell him that he can’t or that he shouldn’t?
I’d been agnostic on AI use by other writers before I heard that story, but I think I’m fast evolving towards a position of you-do-you, I’ll do me, and we’ll let the readers sort it out.