As a university professor, I’ve had quite a few generative AI-related assignment headaches in the classroom, so I feel for the publishers who are trying to navigate submission piles filled with LLM-generated or significantly -assisted stories and novels right now. And as a writer, I found most of my fiction and some of my academic work listed among the properties used to train LLMs, for which I think I’ve made maybe $5 in recompense after the lawsuits? Something not very impressive, anyway. And as a researcher … well, I’m sorry to tell you that when readers don’t know text was generated by AI, they tend to rate it pretty favorably, at least in nonfiction genres like news. Research is still rolling in on AI and fiction. I look forward to learning more about its findings.
None of my fiction has been or will be generated, edited, or polished by LLMs. I have not used LLMs to bounce around plot, character, or setting ideas while writing fiction, and I don’t expect I’ll ever want to use it that way. To date, I haven’t used AI to help me develop, clarify or fact-check concepts or assertions used in my fiction, but I can imagine that being a potentially useful way to use AI, so I won’t take it off the board entirely.
Does that mean I don’t use LLMs at all? Not at all. I read the AI-generated summaries provided by search engines, although I’ve caught those summaries making a nontrivial number of factual errors, so — be careful! I have queried and had “discussions” with LLMs when building roleplaying-game characters (although it’s not always great about differentiating between multiple RPG editions….), writing session logs for roleplaying games, making personal travel plans, developing agendas and priority lists, etc. I’ve found LLMs useful in summarizing legal policies and other jargon-filled documents (“what do I need to know about this cancellation policy?”), explaining new concepts and processes, turning meeting notes into minutes, and other such nonfiction-related information-processing tasks. I used AI to generate thorough, ADA-compliant transcripts for the comic pages found on this site. So I’m not LLM-averse … LLMs can be great tools, as long as you understand their limitations, check sources, and don’t take their output as gospel. …And, you know, are willing to accept that they were trained on stolen works, exploited labor, and are generally wreaking havoc on our economy.
Also, some of the images I’ve used in my collaged book covers and in other elements were generated by AI, according to the various stock-image databases I used while collecting them.
I hope this all serves as a fair statement regarding my use, and avoidance, of generative AI. The bottom line is — you can trust what you read from me. For better or for worse, this fiction is all mine, no artificial ingredients included.