Candlemark & Gleam’s position on AI is simple: we don’t use it. We won’t knowingly buy or publish any work created with AI, and our contracts require writers to certify that they have not used AI to produce the manuscript in question. We do not use AI-created art for our covers.

The reasons not to use AI have been stated by many people in many places, and our reasons are pretty much the same: the large language models were trained on material scraped from the internet and from pirate libraries, used without permission from or compensation to the authors. They are filled with what are euphemistically called “hallucinations” — in other words, the LLMs simply invent “facts” that seem to buttress their statements. This would seem to be less of a problem in fiction, and particularly in science fiction and fantasy, but the genre depends in large part on plausible extrapolation. “Hallucinations” destroy plausibility. Additionally, these LLMs use an unreasonable amount of energy to produce their results, on a scale that has direct and unwanted consequences on local power grids, electricity rates, and eventually on climate change. Finally, and perhaps most importantly, the resulting works are flattened and derivative, constrained by the limits of their training sets and by the parameters of their prompts. We are looking for stories that offer unique voices and new visions, and those are things that generative AI cannot produce.

Might this policy change? This is a science fiction press, so, yes, I can imagine that there might be a future in which a different kind of computer-based entity developed and expressed a self that was able to write fiction. I suspect that such fiction might best be understood by other computer-based entities, and might be shared in an entirely separate artistic ecosystem, but that’s wandering well off the topic.

And that’s not what we have now. What we have now is a program that purports to assist a human being by doing the hard and/or boring bits, when of course it’s how a writer gets through those hard bits, or makes the seemingly dull utterly engaging, that makes stories worth reading. It’s the unique and individual spin someone puts on what might otherwise seem too familiar, it’s the wild imagination that makes the improbable suddenly believable, it’s the amazement of being shown the world in a way you never imagined possible. It’s the shock of recognition when a story tells secrets you hadn’t admitted to yourself. By its very nature, generative AI can’t produce any of these things. And therefore we won’t use it.