My day job (until Little Miss Muse kicks the dream job into gear) isn’t particularly exciting. Mostly, I edit—and sometimes write—for marketers who create web content for clients across a range of industries. Things like landing pages for divorce attorneys and diving excursions, sales descriptions for toddler toys and used Toyotas, and blog posts ranging from how to unclog your nasty bathroom drain to unkinking your sciatic nerve—either of these can be achieved in fourteen easy steps.
Thrilling, right?
This type of brain work is why Little Miss starts to molt when the day job kicks up. And she whines. And causes problems with glitter and grape bubble gum and bottle rockets. Because one can only get so creative when crafting service pages for what to expect when getting dental crowns or how to find the best dog groomer (these tasks contain a fourteen-point bullet list).
I don’t mind the work. My clients are great, and the project writers work hard at what they do. The goal is to get these businesses to appear top-of-fold in the Google search results when a human like you or I type in the dreaded “I NEED HELP WITH THIS HERE SITUATION NEAR ME.” Whether your “this here” requires a Mexican restaurant to ward off the hangry or a mechanic that can track down that mystery rattle.
A while back, things with the day job took an interesting turn, especially given all the questions and comments I’ve gotten from well-meaning, curious folks regarding what they believe to be the inevitable decline of authors due to AI.
The Powers That Be wanted editors to run each writer’s content through a program to calculate whether the piece was human-written or computer-generated.
Okay. I’ll comply. Because, you know. The rise of AI and ChatGPT and all the others have some unethical copywriters cheating. The clients want human-written content because, as far as we know, it’ll be humans who read the words that might lead to a sale or landing new business. (And I suppose if we’re not alone in the universe, the non-human sentient beings of other-planetary abodes would still like to read quality content to see if Earth is even worth their time, price of gas and all…)
I ran the first piece through the checker program.
80% computer-generated.
This particular writer and I have worked together for years. Years. I read it again. She did not vary her style or quality one bit. I ran it again with the same results. She was “dinged” for the score.
Second writer—another I’ve edited for many times—earned a score of 98% computer generated.
Unfortunately, the platform is automated, so those scores can’t be explained or argued; it automatically flagged the writer as cheating.
All the pieces. Experienced writers, mid-range, and newbies. All the work came up AI-generated.
This couldn’t be. Not this group of people. I spotted no change in their styles or formatting…
So I experimented.
I ran my blog post, Never Again, Betty, through the checker.
I lived every second of the event with that angel food cake, and I wrote every word. Well, Little Miss and I wrote every word, that is. Would testify to this in a court of law (though Little Miss would need a bit of a wardrobe upgrade to appear before the judge, but that’s a tomorrow problem).
That post came back 75% computer-generated. Well. My oven, my cake pan, and my Little Miss Muse beg to differ.
I conferenced with the Powers That Be. She wasn’t completely convinced. I told her to grab a section of something written waaaaaaaaay before AI. Like an early King or Koontz or Cussler. Something fiction. Try it out.
She did. Carrie, copyright 1974. Guess what? Either Stephen King was waaaaaaaay ahead of his time and had a tiny robot next to that manual typewriter, or the algorithm (AI, by the way) is waaaaaaaaaay off.
Content creation came to a bit of a standstill while the Powers That Be sorted out how to weed out the cheaters. They still haven’t found a foolproof solution.
Recently, there’s been an uptick in marketing firms using AI to generate their content. They’ve cut the writers out completely. But…But… they’ll hire editors to make it “sound human.”
Uh… How about pay a human? Blood, bones, brains, and the occasional bad attitude. You know. Alive.
Several of these pieces have come across my desk, and I’ve declined every one based on principle. HIRE THE HUMANS, PRETTY PLEASE.
My favorite one I declined?
The proofing checklist asked that the editor check for hallucinations.
What?
That’s a thing, people. Computers can hallucinate. Particularly when the algorithm tries to fill in the gaps they’ve not had the programming for.
I glanced over the piece and discovered that 90% of the 100% AI-generated content would need to be rewritten. Most of it was nonsense. I kindly declined the job and, as tactfully as I could, suggested the marketer HIRE A HUMAN, PRETTY PLEASE! If he'd like, I could write him a 14-easy-steps process email on hiring flesh and blood.
He declined.
I’d experienced this computer “hallucination” phenomenon when I clicked on an ad saying there was a ChatGPT for novelists. For funsies, I gave it the first 500 words of a chapter from Triage and told it to finish the chapter out. It threw in random soda cans and a praying mantis and created two new characters that I did not request.
In other words, it hallucinated because it didn’t know what in the world it was doing.
So for all of you folks curious about the demise of fiction writers—you can relax. Some things can’t be taught to a computer.
But that program I tried was certainly proud of itself. It believed it had completed the request with gusto and asked if I wanted to do another chapter.
Why yes, yes I do. But not through you, Mr. AI Algorithm.
I’ll leave all the next chapters for my Little Miss Muse to conquer. That little imp doesn’t hallucinate one bit. The fact that I talk to Little Miss indicates I might hallucinate occasionally, but that’s a problem for Couch Lady to solve in quite a few more steps than fourteen.
