Let’s get this straight.
The things that people call AI aren’t AI. They’re not intelligent. Large Language Models are spicy autocomplete. Every time your phone changes “fucking” to “ducking,” you’re getting a reminder that there is no intelligent life here.
Now that we’ve got that out of the way, I want to assure you that I will not be addressing any of the pressing moral issues surrounding using LLMs in the process of writing books. I’m not going to tell you why you should or shouldn’t use AI. I’m going to tell you why I don’t use it.
I’m getting old, and I’m not changing my writing process.
Once upon a time, I knew I wanted to be a writer. The problem was I had horrible handwriting. As a doctor, my dad was legally required to have handwriting that confused pharmacists. Mine was worse. Therefore, I learned to type. My first foray into the art of touch typing was on an assortment of creaky old manual typewriters owned by a summer camp I attended. I was horrible at it, slow and inaccurate with a tendency to drift away from home row. The typing teacher despaired.
My middle school purchased a dozen instances of the boat anchor-sized, loudly-humming device known as the IBM Selectric. I improved slowly and got indifferent grades in the class. I could type faster in Spanish than I could in English for some reason and used to practice by typing my Spanish homework. But I got going and eventually managed to break twenty or thirty WPM touch-typing. A kindly and somewhat drunk uncle who could hold down no job but working at the pawn shop provided me with an Olivetti, which I used for term papers.
At that point, due to frustration and running out of correction fluid, I managed to scrape up enough money—largely through accumulated birthday gifts from relatives and summer job earnings—to purchase my first computer: an Apple 2e with the brand-new 80 column card. My typing improved. So did my summer earnings. In those days, if one could type, even slowly and with many corrections, it was possible to earn as much as $8/hr working for a temp agency. I began picking up speed through enforcement by the torture device known as the Dictaphone.
By the time I needed summer money during college, I could hit one hundred words per minute while using WordStar.
I think with my fingers, by which I mean that I touch-type. If I need to text more than a word or two, I wait until I come home and sit down at my computer to do it, because I can’t thumb-type. If I’ve finished one book and need to start another one, I take a little time to let ideas cook while knitting a sweater or a pair of socks, then sit down and start typing until the story flows up my fingers to my brain. It may take me a month to write the first thousand words, but only two or three more months to finish the book.
I once explained to a kind and well-meaning friend that the slow part at the beginning of writing a fantasy novel was coming up with names for everything. Said kind and well-meaning friend offered to take a few example names and run them through Chat-GPT for me, thereby saving me, they said, lots of time. I had to gently, then firmly, then loudly talk them out of trying to help me. Thinking up names for people, places and concepts may be slow, but while I’m doing it, I’m inventing the world. Naming things the way Adam named the animals is how I make a world real to me. If it’s not real to me, it’s not going to be real to the reader (that’s you). Chat-GPT can maybe kind of sort of do this, but it can’t do it the way I do it. You’re reading my book, not Chat-GPT’s. By the way, did you know that to a French speaker, Chat-GPT sounds like “cat, I farted?” (chat, j’ai pété.)
I have all the time in the world.
Do the math: if you can write 1,000 words per day, you can write four books a year, easy. This is tricky if you have to hold down a job and take care of kids but not if you don’t.
LLM-based tools appeal to people who don’t have enough time. Because my health cratered twenty-five years ago and dumped me straight out of the job market and into the wonderful world of “medical retirement,” I don’t lack for time. I don’t even leave the house much, and when I do, I return feeling like I barely escaped with my life. I write, knit and have (checks) over 12,000 hours in Stellaris. And I only need so many sweaters and pairs of socks. Most of my days are spent alternating between playing this or that 4X game, waiting on hold with the doctor’s office and suddenly getting an idea and writing 500 words.
If I’m not writing, I’m not anyone.
So why would I use an LLM again?
Incidentally
Did you notice that I use em-dashes? For giggles, I pulled up a copy of Wishbone, a book that was first published in 2010. I used fifty-nine em-dashes, though they are rendered as pairs of hyphens, because that’s was what we did then. Did you know that Wishbone is one of the books slurped up and used to train that LLM that’s the subject of a famous lawsuit? If you see an em-dash in one of my books, you can be absolutely secure in the knowledge that it’s not because an LLM wrote it. LLM’s learned about em-dashes from me.