Original thinkers will find AI a poor writing assistant

I’m working with a brilliant thinker with an utterly original idea about business. She’s been using AI to polish some of the documents she has created as she plans her book. And she complained to me that it doesn’t seem to be helping that much — in fact, it’s making the expression of her ideas less effective.

That made perfect sense to me, because Large Language Models are allergic to originality.

If you are original, AI thinks you are wrong

If you are writing about something well understood and well-known, AI is likely to be very helpful. Let’s imagine that, say, you wanted to write about the idea that marketers should measure the results of every action they take and carefully analyze what’s working.

Almost everybody believes that. And as a result, an AI trained on all the vast collection of material in the world on marketing knows that well. If you ask an AI to describe that principle in a few paragraphs, it will have no trouble. The result will be boring and ordinary, but well in line with accepted wisdom and best practices in marketing.

However, there’s very little demand for books about the conventional wisdom. It’s easily discoverable online. Even if you suck at Google searches, anyone could just ask any LLM to spit it out. Nobody needs a book like that.

There is demand for books that write about new, counterintuitive ideas. What happens if you ask an LLM to write about those ideas? Even if you attempt to draw it into your unique worldview, it will remain biased by the old, conventional ideas. You may be able to force it into writing something, but its “heart” won’t be in it. So the writing will be boring and unpersuasive.

You’re basically telling a right-handed swordsman to fence left-handed. Outside of the world of “The Princess Bride,” the results will be ineffective.

Where an AI could still be useful

You can still use LLMs to help you with your original idea.

You can ask them to find you studies that back up your perspective.

You can ask them to act as a devil’s advocate and poke holes in your thesis.

You can ask them if there is a better way to organize your ideas.

Just don’t ask them to write about your original idea. Since it’s your idea, your way of writing about it will be better than an AI can come up with.

Write what you believe. Because, when it comes down to it, AI won’t readily believe in your original, unconventional, counterintuitive idea that goes against all of its prior training.