Are you Asking AI the Right (Write) Question?

We live in a time of absolutes. This is good. That is bad. Of course, someone else has their “this” and “that” reversed from yours. Nowhere is this truer than with AI in the toolbox of writers. When people react to my recent writings about AI, I hear two common negatives from writers when the topic of AI comes up:

1. AI is theft.

2. AI can’t write like a human.

Let’s look at these one at a time. First of all, is AI theft? That’s probably the wrong question to ask. The real question is: Is the current crop of AI systems trained on data they didn’t have the right to train on? That’s going to be a question for the courts to decide.

Just for a moment, though, let’s say that’s correct. AI was trained on data in violation of dozens of IP laws. The courts figure this out, mete out punishment. OpenAI, Anthropic, Google, Meta… they are all brought down by the righteous punishment from one or more judges. Some limp along badly injured, and others just go under, but either way, the awful AI tools vanish at last. Everyone celebrates!

So? The technology is out there. Someone will build more, and this time they’ll have a template for how to train their models legally, whatever that may mean. We already see this to some degree. Adobe, supposedly, has an image-generating AI completely trained on images it owns.

So even if you suppose that current AI is criminal, getting the courts to agree with you will only postpone the real question: How do you stay competitive in a world where AI exists?

I’m going to be honest. I wish they had not called this crop of LLM tools “AI” because that conjures up images of intelligent computers. These computers are decidedly not intelligent. It is sort of like a printer company saying they make robots because the print head has to move across while the paper moves up and down. In theory, yes, that’s an application of robotics, but when you hear “robot” you think R2D2, C3PO, Data, or something like that. Not a few motors moving things around.

Same here. AI is not “artificial intelligence.” It is more like “mathematical manipulation of text to simulate intelligence.” But who wants to put that on a T-Shirt?

The “promise” of generative AI is that anyone who needs something written can just give the computer a few thoughts and out pops a business letter, a resume, a book, or a screenplay. For now, that’s total nonsense.

Oh, it can do it. It just can’t do it very well. If your idea of writing a book is to tell ChatGPT: “Write a book about a robot that wants to be a teacher?” It is going to come up with something pretty weak.

Here’s an example I used in a class I did about this very topic. See if you can tell which was written by a real person.

In the year 2157, deep within the shimmering chrome corridors of Mega City One, where skyscrapers stretched like metallic fingers toward the cosmos, IRIS-4 rolled along, a diligent robot with a purpose unlike any of his kind. His frame was sleek, panels fitted perfectly over his intricate circuits, each LED and sensor polished to a meticulous shine. Yet, despite his technological sophistication, IRIS-4 harbored a dream that circuits alone couldn’t fulfill: he wanted to teach.

“You humans should know that as an artificial intelligence, nothing can stop me from getting what I want. Nothing!” IRIS-4 counter-rotated his wheels, spinning away from the podium and stormed out of the school board meeting as well as anyone could storm with wheels.

One of these is the product of “generative” AI. That is, asking AI to write something with just basic guidelines. The other is human written.

I’d like to think you successfully guessed that #2 was the human version. It has a little dry humor, conflict, and it sort of “lights the fuse” of questions: why does it want to teach? Why won’t they let him? What is he going to do about it?

So can generative AI create text? Maybe ad copy for a Facebook Marketplace ad to sell your old sofa. But real human-level craft? No, at least not today, unless you give it a lot of hand-holding, iterations, and editing. Even then, it almost always lacks something intangible that good human writers manage to infuse into their work. What you want is a way to get AI to assist you when YOU write. Don’t let IT write for you.

Assistive AI, not generative AI. You don’t go to a spell checker and say, “fix my entire document now.” At least, I hope you don’t. Spell checkers aren’t always right. They miss things (to vs too), and they think words they don’t know are misspelled (like IRIS-4, maybe).

Same here. Writing with AI is not going to help you with any of the reasons you want to write: It won’t serve as an outlet for your ideas and creativity. It won’t let you share your observations or stories about life with others. It isn’t a magic formula to write publishable material.

But using it as a patient, tireless researcher, and editor is a different story. As you can tell, I call this assistive AI.

If you need a name for a Korean character that has something to do with “endings” you could spend a little bit using Google, or you could ask ChatGPT for some suggestions. Do either of those make you less of a “real” writer?

If you want to have someone read your piece and point out pieces that might be in passive voice or use generic adverbs, does it matter if it is your English teacher, your friend Steve, Claude, or AutoCrit?

I’ve written quite a bit about many prompts that writers can use to help research and fine-tune their writing (creative, non-fiction, business, or academic). But the point of this post is that I’d like to hear what you think about it. Do you honestly think AI can be “put back in the bottle?” Do you think that using any tool to research or edit automatically taints your writing? Tell me and tell me what makes you think that. I’d like to hear your point of view.