Can AI Help You Write a Children's Book?

No. Well, maybe it can do outlining and grammar. Maybe.

My grandsons were over a while ago, and when they got bored, we played with A.I. Here’s what I learned, here are some programs I’m looking at, and here’s what I predict for the future of A.I.

What Software to Use? And How?

Various LLM (Large Language Model) AI programs are available right now, and the list changes almost daily. To skim the surface, not even attempting to be all-inclusive: ChatGPT is the best known, but there’s also Gemini from Google, Microsoft’s Copilot, and Claude from Anthropic. Every major player has an AI program available. So, what do I use them for?

Write a picture book. I’ve tried prompting various programs to write a children’s book, and the results are awful. It doesn’t understand the genre of children’s books - at all! AI experts will say that the results will depend on the prompt. Correct. If you just prompt, “Write a children’s book about pigs,” you’ll get an awful rip-off of the Three Little Pigs folk tale. But even with lengthy, detailed prompts, it still writes bad manuscripts. Not enough to even bother with as a rough draft. YMMV.

Creative Language. So, I wondered if the AI might help me with some creative language. Like metaphors, rhyming words, etc. I asked a couple AIs to give me a list of nonsense words that fit an anapestic meter (da da DUM). ChatGPT couldnt’ do it, giving me a list of iambic (da DUM) nonsense words instead. Claude did a better job, understanding the poetic task better. So, I started looking at Claude in more depth.

Outlining. One thing about Claude is that it allows you to create a project and within a project to unload documents. I uploaded a document, “The Hero’s Journey,” a plot template that explained the Hero’s Journey in some detail. Then, I created a “Story Bible,” that listed main characters, setting details, and the basic plot problems. I asked Claude to refer to the Hero’s Journey and use the Story Bible to create a plot outline for a 50,000 word children’s book. (For those knowledgeable of A.I.—of course, the prompt was much more detailed and complex.)

That was interesting. The plot was cliched, of course. The LLMs can ONLY predict what the next word should be! That’s it. They aren’t thinking, not really. Only predicting the next word. So, of course, the plot was cliched.

But then, I asked Claude to evaluate the plot. Using the Hero’s Journey document as a plot template, list three weaknesses and three strengths of the plot. Interesting. It was able to evaluate it’s own plot and sure enough, the weaknesses were weak and the strengths were strong.

For each weakness, I asked for ten possible alternatives to the weak plot point. And ten more. And sometimes, ten more.

None of them were good, they were all cliched. But, a couple made me think of something new and creatively different. I changed the plot to include my new ideas.

Understand: the AI didn’t suggest anything great, but its suggestions prompted something in my mind to make a unique connection. I may have gotten to that point on my own, but using AI to throw out possibilities made that something click faster. And then, I asked the program to repeat the evaluation, finding three weaknesses again. And repeated until I was happy.

The revised outline has sat a couple weeks as I’ve been very busy. I’ll get back to it soon, and with fresh eyes, I wonder what I’ll think? Will it be a strong enough outline to write from, or will it still need more work? Claude can’t tell me that. As the author, I’m the only one to make that determination. But Claude, with its project feature, has been interesting.

In the meantime, I reread Kate diCamillo’s fabulous book, The Beatryce Prophecy. I tried to imagine the prompts that would result in this book, which tells things sideways, includes too many narrative sections, and has a twisty plot line. No way could A.I. write this book. Literature like this is outside the realm of possibility for current A.I programs. And that’s what I want to write, something that touches the human heart. I won’t be asking any A.I. program to write my stories - instead, I’ll do it myself.

For me, with the current state of A.I., the most it can help me do is revise an outline. Claude acted as a critique partner, giving me information on strengths and weaknesses of my plot. The program and I were on the same page, since my prompt made it reference a plot template for the Hero’s Journey. But that’s about it. It can’t write my story for me because it would be poorly written, cliched, IMHO.

Research Material. And then, I went on a trip to Europe and part of that trip included visiting every Copernicus site I could in Poland. We went to Torún (birthplace), Krakow (university), and Frombork (where Copernicus worked as a cleric for the Catholic church for 40 years). In the course of research, I’ve found many resource documents full of facts and dates. Enter Google’s NotebookLM. This tool allows you to upload documents, or to reference websites, and create a sort of database of information. Then, you can ask questions and it will tell you which document has the relevant information. For example, you might ask: give me a timeline of important events in Copernicus’s life. Each timeline entry will be footnoted with the places in the resources that documents the answers.

Wow. This is an experimental A.I., but how nice to not have to dig through ten documents to find the one piece of information you need. This seems a great, reasonable use of A.I.

Caution: The NotebookLM still says: “NotebookLM may still sometimes give inaccurate responses, so you may want to confirm any facts independently.” So, you should check all the facts again.

Grammar. For several years, I’ve used Prowriting Aid to correct grammar in a piece of writing. I am most definitely NOT a Grammar Witch, and I do a lousy job of copyediting. I simply don’t SEE errors, even when I’m told there’s one in a certain sentence. I need a copyeditor to correct things for me. I like Prowriting Aid because it looks for things like repeated words (I keep saying, “a little bit.”), helps me with all the commas (I’m horrible at them), and gives me a visual of sentence lengths. When I’ve had interns, I asked them to run all blog posts through the program. It’s a fantastic grammar solution for me.

The Future of A.I.

However, A.I. is coming. My three grandsons LOVE playing with Claude—and we’re not even talking about AI to create images such as MidJourney.

They asked Claude to create screenplays. Our prompt included characters, plot lines, and limited the story to only 1000 words. In seconds, Claude spit out new stories. We prompted it to revise in certain ways, and laughed at the new versions.

One character was supposed to be inarticulate, unable to speak anything but gibberish. Here are a couple of the phrases Claude has Gibberish Guy saying:

• Hakuna matata! Yippee ki-yay!

• GABBA GABBA HEY!

• Wakanda forever!

We died laughing at the phrases used out of context! Juxtaposing Lion King (Hakuna matata!) and cowboys (Yippee ki-yay!) was hilarious. They were so obviously wrong that they were funny.

But through it all, the boys loved playing with A.I. and creating stories, screenplays, and jokes. They had found a new toy and it was fascinating to them. They were trying to figure out how to integrate it into their classes.

Using A.I. will be a normal part of their lives. They quickly learned how to give longer, more complex prompts to get the results they wanted. We can rant and rave, and wait for the courts to rule on the copyright issues inherent in the programs—but we can’t stop A.I. It’s here and it’s going to be a normal part of our grandchildren’s lives.

A.I for Architects or Writers?

However, I recently talked with a couple architects about how they use A.I. to write about their creative projects. They love how the A.I. can churn out a description of a building project.

When I read the same material, all I want to do is edit because it’s overwritten. It needs to be cut, edited for length and detail. Focused. So, are the architects wrong? No. It’s just that the A.I. will always be more useful for certain tasks and certain audiences. Engineers, architects, and so on who traditionally struggle with words, will love that the words flow smoothly. But writers will understand that the writing needs to be edited for many reasons: audience, clarity, brevity, and so on.

A.I. can’t write my picture books. It doesn’t understand the genre, it doesn’t know how to be creative, it doesn’t understand the dual audience of children and adults, it doesn’t have my background in reading children’s literature. It just won’t work.

Sorry, A.I. You can’t take MY job. But you’ll find your place in this world.

What’s your experience with A.I.?