Dear Reader,

Before you hop over to whichever side of the generative AI fence you tend to feel comfortable on, I'd like to share with you how I manage to be on both sides of that hot-button issue.

And I'm frankly not in the least uncomfortable about it.

First, four of my books were pirated and used to train Claude.ai.

If Anthropic had actually purchased the books, I might just have gone, "Eh. So what." That's because the nature of LLM (large language model) training is very different from copying or plagiarizing. I know there are those who would argue with me about that, but it's more than my brain wants to deal with to wrestle with that particular issue.

The best analogy I've heard is that a single book is like a grain of sand. I wish I could do a better job of explaining the exact nature of this training. But ChatGPT, for instance, is not in itself pirating your books, or making them available to others.

Here are my books that are included in the settlement:

Will I get any money? Eventually, I think. But I felt it was important to go through the process of making the claim no matter the outcome.

Oddly, this did not sour me on being eager to learn how to use AI to my best advantage as a writer.

AI as a writer's tool—the good and the meh.

I'll use ChatGPT as my example here, because it's the one I subscribe to and use most often. I do sometimes dip into Perplexity for Web searches, but ChatGPT "knows" my work. When I ask it subtle things—such as would a woman like X character have been able—believably—to do that in 1814? Or what educational materials would have been available to someone of X class in X place at X time? it has the frame of reference of my previous questions and the chunks of my writing I have shared with it to bring to its answers.

Or—very usefully—Can you give me a list of Cornish surnames servants might have had?

Lists, as you might guess, is one of the ways I use ChatGPT. Lists of things are very handy. As are negative examples to use in workshops. While I could probably find the answers to all my questions in a basic Internet search, the ability to pose contextual questions means that the answers will be more specific and more useful to me.

(Is it a little spooky when it says something like, Sophia's maid might have XYZ name, which would imply... when I haven't specified that I was looking for a name for that character's maid? Yes. Moving on.)

And now, both ChatGPT and Perplexity give you the sources for their answers, so it's possible to check whether they are correct or not.

Surprising research results

What works quite well in ChatGPT when it comes to research is its "eagerness" to keep digging down and finding out more. It also has virtually instantaneous access to information it would likely take me a deal of time to uncover. I gave an example of this in my previous email. It dug up an 1814 text on the diseases of women—which I was able to download. (Definitely out of copyright!)

I'm fast learning that asking the right kinds of questions to get the information you want is a bit of a skill. It matters more when it comes to the other way I use ChatGPT, however. Using AI effectively basically means you have to learn to write prompts that help you direct it to work in the way you want it to.

And yes: AI does hallucinate, so I always double-check information.

Tireless writing buddy

All the writers on this list know that writing can be a lonely occupation. Those doubt demons, those insecurities, are all too ready to leap up at the smallest provocation.

I'm talking about things like, "Does this scene drag?" or "Am I being too wordy?" or "Is my protagonist likable/believable?"

I have many writer friends, and if I had my druthers, I'd have them read my work and answer those very questions for me. That can happen in a critique group session, of course. But humans (thank heaven) bring their own lens to things when asked to opine. Even if you're very specific about the kind of feedback you're looking for, you might not get it. Of course, you also might get truly enlightening feedback you hadn't thought you needed. But that is a luxury most of the time.

Plus, those kind fellow writers might get tired of answering your questions, of helping you soldier on to the next scene, etc. even if you do your best to reciprocate.

Before I continue, let me go on record as saying that there's truly no substitute for human input. Your readers, after all, will be real live people.

ChatGPT, however, never gets tired, and has vast information about best practices and patterns that can be very useful. At busy times, it might slow down, but it always gives you the same amount of attention with each question you ask.

Virtual assistant

I also use ChatGPT to help with:

• Summarizing material
• Marketing copy and ideas
• Brainstorming
• Helping me make slides for workshops and presentations
• Definitions and tech questions

And a few other more mechanical tasks.

What I NEVER do: Get ChatGPT to write part of my books for me. In fact, I almost always modify whatever it spits out in all the instances I've listed above. Often I draft something myself, then ask ChatGPT to give me variations. This is really useful for book description copy on Amazon.

I believe more and more writers are harnessing what generative AI can do to help them in a positive way. Are there potential abuses? Absolutely. Witness all the SPAM emails I get gushing about my published books and saying they'd like to help me bring them to a wider audience (I know they've never so much as cracked open one of my books...).

Proof that I'm not alone is a newsletter from Tiffany Yates Martin (my hero), who uses AI in much the same way.

Moving on.

Last chance to sign up for It's Scary How Fun it Is to Write Romance!

Join me tomorrow, October 16 at 2pm EDT for a fun, interactive workshop where I take you through some of my romance-writing techniques—and analyze the character arcs in one of my favorite rom-coms! It'll be fun. :)

OK, long email today, but I hope it was informative. Happy reading and writing!

Susanne