I know this is a fraught subject, so let me just get this out there immediately: I don’t use AI to directly write any part of any of my novels. It doesn’t write any of the prose, and it doesn’t create any of the concepts. The writing and the ideas are wholly my own.
Having said that, there is one aspect of AI that I use extensively to improve my creative writing, and that is critique, as I have found it difficult to find good reader feedback. Most feedback is either not critical enough, or not timely enough. Friends and family are especially difficult to use here, as having them read an entire novel (especially in a rough form) is a big ask, and they are particularly resistant to criticizing. Even when hiring beta readers through an outlet such as Fiverr, the results can be pretty hit-and-miss.
Generative AI to the rescue.
Anthropic’s Claude (I’m currently using 3.7 Sonnet) is my go-to AI resource for writing. I did a bit of research before subscribing, and it appeared to be most recommended for creative writing critique. I have found it to be fantastic as a resource.
I write iteratively, constantly revising and rewriting plots, characters, etc. Using Claude allows me to write a section, have Claude critique it, and then I update again, have Claude critique again, ad exhaustium (yes, I just looked that up for this). A superpower of Claude’s is its ability to digest an entire novel and offer critique on both its entirety and its components. My manuscript is about 78,000 words, and it only consumes around 66% of my allowed “knowledge capacity”.
Claude is often overly helpful and wants to suggest solutions to the weak spots, but I ignore them and address them myself. It’s my writing, after all! In other words, Claude does the problem identification, and I do the problem-solving. I’m sure I could ask it not to suggest fixes, but I usually don’t bother.
There are limitations, but I find them to be navigable. For example: the larger the context, the more challenging the request to the LLM (generative AI is almost exclusively driven by Large Language Models, and context is something that all models have to deal with). I might create foreshadowing in Chapter 10 for something that is going to happen in Chapter 50. Generative AI may struggle to find and correlate subtle connections over a massive span of content (humans may not do much better!). Also, sometimes I just disagree with the critique. The AI may not see my grand vision yet, or may just have a different opinion about how something should play out.
This brings me to a final benefit of using AI: you can ask it to be harsh or forgiving, punishing or productive, and you can ignore everything it says without hurting its feelings. And if you need a pep talk, it feels weirdly good to ask it to say nice things about your writing, such as listing out the piece’s strongest aspects. Yes, I do this sometimes, and it helps keep me going.
AI writing is a scourge. It’s flooding the market with garbage. That said, I do use it on occasion to help remind me of grammar rules and sentence structure. As a tool, it’s phenomenal. But too many people are using it to write entire stories and tossing them onto amazon, thinking it will make them millionaires.
LikeLiked by 1 person