It Would be so FINE to See Your Face at My Door

Musing about AI and my writing voice.

Because I wrote this week about using AI to create teacher’s lesson plans, I thought I’d also post this essay I wrote about AI. It expresses why I still don’t use AI to write my stories.

—

Whenever I think about using AI-LLMs to write something, Carole King’s song, So Far Away, starts going through my mind, and I can’t stop singing it. It begins:

So far away
Doesn't anybody stay in one place anymore?
It would be so fine to see your face at my door

It’s the “fine” that stops me. Why did King use that word?

A common phrasing would say it this way:

“It would be so nice…”

“It would be so great…”

But those don’t work? Why?

Phonics - The Sounds of Words

Here’s my college phonics class coming in to explain:

Fine works because F is a fricative, a rough sounding consonant made by forcing air between the teeth and lips, and that echoes the friction, or the anguish of this story. Combine that with the I, a vowel sound made with the tongue high in the mouth that takes a lot of tension to hold it there, and which also is a vowel you can hold for a long time. And finally the N sends the sound through your nose, creating a whiny sound because maybe she’s whining just a bit in this song?

It would be so FINE to see your face at my door.

And I’m captivated all over again by the power of a person’s writing voice.

How you say something is everything.

If King had said, “It would be so nice…” the song wouldn’t resonate. Nice has the whiny N and the tense I sounds, but the word doesn’t hold out for emphasis. “Nice” just droops here.

Dictionary.com notes:

Fine: of superior or best quality; of high or highest grade.

First recorded in 1250–1300; Middle English ffyn, fin, fyin, from Anglo-French, Old French, from Medieval Latin finus “pure, fine,” from Latin fīnis “end, utmost limit, highest point” (as in fīnis bonōrum et malōrum “the highest good and evil”)

As in: It would be the ultimate joy to see your face at my door. The meaning of the word expresses the longing of the singer.

It would be so fine to see your face at my door.

Compare the synonyms

Fine has the connotations of costly (A fine wine), or delicate (a fine cotton fabric).

Nice has the connotations of pleasant or agreeable (a nice visit—but sorta generic), or suitable or proper (That was not a nice remark.)

Great has the connotations of large in size (a great fire), or exceptionally outstanding (a great event—but also sorta generic).

“Fine to see your face” is a sharper, less generic expression.

Did Carole King consciously go through the options for word choice? Not likely. I don’t know if she has the grounding in phonics to understand that the F is a fricative sound and so on. Probably not. She just listens, holds the words in her mouth and heart as she says them and feels whether it’s the right word or not.

Her experience in writing lyrics helped her reject the common words of nice and great and reach for an uncommon word: fine. Because it’s unexpected, the word elevates the line of the song into something memorable.

Then add in the music: the word fine is on a higher pitch note and it’s a longer note, hanging in the air before the lyrics continue “to see your face at my door.” The melody emphasizes the “fine.”

And I think about a generative AI-LLM: its task, on the minute level, is to predict the next word. That’s it. Even the more sophisticated AI-LLM’s which “think” are still predicting the next word.

Carole King was doing a uniquely human thing as she wrote the lyrics. She was reaching for words to express the regret of someone leaving and not seeing them again. She built it word by word, reaching for the next word by bringing her humanity, her experience, her voice to the task.

Humans don’t predict the next word; we reach for it.

Voice. That’s why I can’t use AI-LLMs. Carole King chose the word that seemed right to her, that fit her intentions because she listened to her writing voice. AI-LLMs can’t choose words that seem right to ME. It only predicts the next word—generically.

Does the AI-LLM know about phonics? How words are created in the mouth and how they sound? Does it know about the “wave in the mind,” as Ursula LeGuin referred to the phrasing of words across a passage?

No.

I know that there are ways to input my previous writing and ask the AI-LLM to use my writing voice as a model for writing. But really? All that—and much, much more—is already in my brain. It seems a waste of time, especially for results that most AI-assisted writers still say needs a lot of editing.

Others say, just wait. The AI-LLMs are learning so fast, it’s just a matter of time.

And I keep singing: It would be so FINE to see your face at my door.