Oh my, AI (part 3)

Sorry I didn’t get back yesterday with the promised post. Several appointments ran long and the day was gone before I knew it. But it let me think about how I wanted to do this next post because, as I worked more with ChatGPT and its output, the more I realized I’d been overly ambitious about what I wanted to do here on the blog. So, before posting the results of the prompts and how I would change/edit them, here are some warnings about what not to do. The short version is, don’t use it as a replacement for the actual work of being a writer.

I know I probably sound like a broken record, but programs like ChatGPT, Jasper, etc., are tools for writers to use and should not be viewed as a way to replace the actual work of writing. Unfortunately, there are a number of folks out there who haven’t figured that out and they are going to make life hard for the rest of us for a while as platforms like Amazon, Apple, and Google figure out what to do about AI-generated content.

If you go to Youtube and search for how to write a book using one of the AI platforms, you’ll find a ton of videos. Most of them will walk you through the steps of how to “write” and then publish your new book. There is little to no discussion about how to polish the output generated by the AI. There is certainly little to no discussion about the danger of uploading the raw or barely polished AI output and therein lies the danger.

While the various publishing platforms are still attempting to address the issue of AI-generated content, as authors we need to be proactive. That’s especially true right now with the screenwriters strike. One of the issues raised in the talks before they walked out was the use of AI-generated scripts. That’s brought the issue to the front page and I guarantee you Amazon and the other platforms are watching closely what happens.

Then there’s the fact Google announced more than a year ago that AI-generated content was against its guidelines and was considered spam.

Honestly, it is only time before we start seeing platforms like KDP or Google Books kicking books–and possibly even author accounts–over AI content. So don’t risk falling into that trap just because such content isn’t against the rules yet.

Don’t get me wrong. I’m not saying to avoid using apps like ChatGPT.

What I am saying is to use it like a tool and don’t get sloppy or lazy. Use the app if you want but then take the additional steps to make the output yours.

How do you do that?

It’s really simple. It starts with reading what has been output. Edit it. Expand it. Make it your own and in your own voice. That’s step one and in many ways will be the most difficult because it’s easy to zone out doing something like this and start skimming over the text. But don’t. This is the most important step to creating AI-inspired (note the change from AI-generated) text.

For example, here is a prompt I used this morning to illustrate what I’m talking about:

genre: urban fantasy. Write an opening paragraph to a story about a female police detective coming upon a crime scene that looks like an ordinary robbery but she can feel the magical energies and knows it is supernatural in nature. The problem is no one on the police force knows she is a witch and most people still refuse to believe in magic.

And here is the output:

Detective Ava Rodriguez approached the crime scene with a sense of unease, her instincts telling her that this was no ordinary robbery. The convenience store was in disarray, with shelves overturned and merchandise strewn across the floor, but there was something more than just the physical chaos. She could feel the pulsing of magical energies emanating from the scene, an energy that only someone with her particular gifts would detect. It was a curse, a dark magic woven with intent, and she knew that it was her responsibility to solve this case, even if no one on the police force knew of her abilities. Ava took a deep breath and stepped forward, ready to face the challenge that lay ahead.

Now, that’s not bad as a first try, but it doesn’t fit the general rule right now in the genre that it needs to be in first person. Let’s see if I can change that using ChatGPT.

I input the following prompt to do so:

rewrite the above paragraph using first person point of view

Here’s the result:

Approaching the crime scene, I felt a wave of unease wash over me. Despite the apparent chaos of the convenience store, I knew this was no ordinary robbery. As a police detective with a secret, I could sense the presence of magical energies pulsing through the air. It was a curse, a dark magic that had been intentionally woven into the scene. I knew that it was my responsibility to solve this case, even if no one on the police force knew of my abilities. With a deep breath, I stepped forward, determined to face whatever challenges lay ahead.

Now, how would I do it? (Remember, I am doing this on the fly and without enough coffee to really think yet. So don’t expect perfect.)

Flashing lights from a pair of squad cars reflected off the exterior of the store, a beacon to all the looky-loos and night desk reporters in town. Now they called out questions to the officers standing post outside the store, demanding to know what happened, if anyone was hurt, had they caught the perp? I didn’t know the answers to the first two questions, but I did to the last. No, they hadn’t caught the perp. That’s why I was there. Now to find out the answers to the rest of it.

“Detective.” The older of the two uniforms nodded as he checked my ID.

“You first on scene?”

“Yeah. Got the call half an hour ago. We requested a detective after seeing the inside.”

I pursed my lips, not liking the implication. Unfortunately, I couldn’t ask him what he meant, not with reporters and civilians close enough to overhear. Instead, I pulled on a pair of nitrile gloves and said I’d speak with him and his partner shortly. Then I pushed open the glass door and stepped inside.

The moment the door swung closed behind me, I hiss out a breath. The sense of unease I’d felt outside exploded as I took in the scene. It looked like a interior took a direct hit from a tornado. Shelves were overturned, their contents strewn around like a mad game of pick up sticks. That’s not what had my hand reaching for the gun at my hip. It wasn’t even the way a dozen cans and jars were embedded in the ceiling. There was something else, something much more dangerous and something none of the others knew about.

Damn it.

My skin prickled as the remnants of magical energy danced over it. I fought the urge to run from the building. Hiding in bed, the covers over my head sounded pretty good right now. Anything but staying here did. How in the world was I supposed to explain to my lieutenant that the perp hadn’t been on a drug-fueled rampage? I’d much prefer that rather than having to explain to him the damage had been caused by dark magic and, unless I missed my guess—and I never did when it came to magic—there’d been an ineffective curse involved as well.

Okay, it looks a lot different from the two suggestions from ChatGPT. But that’s not enough to make sure I don’t run afoul of any current or reasonably anticipated rules about AI-generated content on Amazon KDP, etc. So it is time to take the next step (assuming I’m happy with the current state of the text). That step is to run the text through a tool that checks for AI-generated text. One such tool is Originality.ai. Another option is turnitin.com. This is a site often used by students and instructors to check for not only originality but plagiarism.

After going through that level of check, you may need to find a paraphrasing tool to help you rewrite/rephrase those portions of the manuscript flagged as being AI-generated. One such site is Quillbot. Quillbot is free, iirc, but you are limited to the number of words you can input at a time. Other apps like Grammarly, etc., can help here as well.

Even then, you’re not finished. You need to do a final editorial pass and let your beta readers have a go at it.

Here’s my concern if we, as writers, don’t follow these steps: we’ll find ourselves caught up in the same situation so many writers did when Amazon started cracking down on low content books. People tried to take advantage of the system (let’s be honest, they wanted to rig the system) and it backfired not just on them but on a lot of folks who got caught up by the bots. Writers who had done nothing wrong and who had no ill-intent. Books were removed from the store and some accounts were frozen and even removed.

Don’t run that risk. Use AI as a tool, not as a replacement for the work of being a writer. We are creators, not copy and pasters.

And now I’ll step off my soapbox and go find more coffee. Until late

I created the featured image using Midjourney AI.