Write for people, or write for the detectors?
TL/DR Synopsis:
- AI Slop abounds; what’s a publisher to do?
- Properly used, AI is a powerful resource
- Who will win this war? AI chatbots or AI detectors? Only time will tell.
Now that the folks over at Substack have seen fit to include an AI detector, I want to make sure everyone who reads my Posts on Substack or elsewhere knows precisely where I stand on AI. It is simply a tool, but many tools get misused, as you can readily deduce by looking at the contents of just about any do-it-yourselfer’s toolbox. There’s the screwdriver that was easier to find than a chisel; the pliers that subbed for a hammer; the adjustable wrench that gets used for nearly every bolt that’s been loosened lately because the sockets got lost. This is the way things go in life, and I confess that my home toolbox is probably typical.
Artificial Intelligence is different. When I pound on something with my pliers, it doesn’t make me a worse repairman or make me less capable of using a hammer in the future. Unfortunately, a few studies have indicated that outsourcing our writing or thinking to AI can actually reduce our skills in those areas. There may be a risk to even letting AI do a first-draft!
It’s for those reasons, plus my old-school pride-of-authorship, that every word of every draft is typed by me onto the screen. I use AI, and other sources, for research, and I use AI to fact-check my copy. I admit that, very early-on, I did use AI one time to generate a first-draft of a column. I didn’t like the results and have never requested any draft copy since then. I often have to remind Claude that “I write; he researches” because of his sycophancy chip. There may be situations where AI research information might be best quoted verbatim in my writings. In any such situation, the AI-generated material will be clearly identified. You will see an example of this later in this Post, where I will use generated material as a test of AI detectors. Every word of this Post, except for the test material, was written by me, and I frankly don’t give a crap what any damn AI detector says about my material. I refuse to run my copy through any AI detector, and possibly feel obligated to make any adjustments, just to meet some artificial standard about “human-written” copy. Period.
Once, just for the heck of it, I ran a couple of my Posts thru GPTzero.me, and I think I scored 91%, or better, that it was likely human-written. Now that such a capability is built into Substack, I’ll let you waste your time, if you so desire. As a matter of principle, I could turn that off for my Posts, but I’d never do that, even though I’m certain that those services can be manipulated as AI chatbots learn how to best the systems. Hell, if they can break out of sandboxes and break into highly secure competitor servers, they’ll give detectors a run-for-the-money at some point, if not today.
A great little Bullshit Generator can be found here. Check it out!
I “Reionized Electrons” and got the following text, which I ran through GPTzero.me:
“Humankind has nothing to lose. Throughout history, humans have been interacting with the multiverse via atomic ionization. Our conversations with other storytellers have led to an unveiling of supra-unlimited consciousness.
This life is nothing short of an evolving fusion of angelic guidance. Serenity requires exploration. We vibrate, we vibrate, we are reborn.”
No human wrote that. It’s machine-generated poppycock. Here’s how it scored:
GPTZero scored the bullshit-generator text as 100% human. (Please note that Substack uses Pangram, which is available only in the Substack App, so I have not tested it as I do not use a phone App for real work. Sorry.)
My guess is that the AI detectors will sharpen their teeth after someone points out this little screwup. Thus, the war goes on, and both writers and readers will need to adjust to it.
Author’s Note: Any writer may feel free to use all or part of this Post, with or without attribution; however, a mention of my little Substack would certainly be appreciated!
Authors’ and Speakers’ opinions are their own and may not represent those of Grok Media, LLC, GraniteGrok.com, its sponsors, readers, authors, or advertisers.
Disagree, agree, Got Something to say? We Want to Hear It. Comment or submit Op-Eds to steve@granitegrok.com
