#433 – What will AI teach you about your writing?

Many writers now add a disclaimer: "AI didn't write this." As if we couldn't tell what's AI-generated from human-created!

I recently asked ChatGPT to write the script for a video-course I'm creating, based on the outline I fed it. I thought it'd be straightforward: the machine would take my language and arrange it in blurbs I could say in each lesson.

I got stuck in the first blurb, refining, giving it new commands that'd lead to a more accurate reflection of my ideas: "that's not what I said," "that's not what this means."

Then it hit me: LLMs are predictive machines. They're based on how probable it is that certain words go together.

But "good" writing, the kind of writing that makes the hairs in the back of your neck stand is, precisely, unpredictable.

In 1989, when I was a Spanish Philology student, my friend Pablo, an Argentinian exile in Spain, told me that we philologists hated Julio Cortázar because he didn't write "correctly." We were on a beach, the Atlantic Ocean too cold to get into. He was lying on his belly, an open book in front of him; I was crosslegged, facing the sea, an open book on my lap. The sand felt cool and damp on my elbows, as I lay back to tell him, "not this philologist!"

Today, thinking about how AI writes (re: puts words together in predictable ways), it came to me that AI is like those philologists that hate that type of "incorrect" writing.

We humans, however, aren't moved by "correct" writing; we crave alive, true, imperfect writing. We read to learn how other humans experience life. But no machine can give us that: it doesn't have the human experience.

If you want to learn the power of your writing, ask AI to write something for you. It'll give you something correct, perfect, and sterile. Then, you'll see how your voice is irreplaceable.

What proof do you still need that your unique voice as a writer is needed?

Love,

Carolina