GPT&ME

Proofreading

I asked ChatGPT to proofread a story. It was originally published in 2019 by Reflex Fiction.

There are three errors that I purposely inserted to see if The Machine would catch them. Strike-through text highlights the mistakes in the story.

I apologize in advance. The story is quite sad but 100% made up. Pure fiction.

Yesterday’s Pictures

The boy beams when finished. Beams. Like the face of God’s son whose name I no longer invoke. For eleven years he’s smiled whenever he sees me. Smiles when I feed him yogurt. Smiles as I clean up shit and vomit. Gabe is a happy child. Happiest when he finished a drawing.

“A fire truck,” I say, holding the crisp, thick sheet of paper. My son cannot draw a straight line. Each streak trembles or spikes because him hands shake badly. But Gabe has adapted, bundles jabs of color like a Impressionist. A fire truck, without doubt.

“Which ones?” I ask. The color of pushpins are important to Gabe, for any artist choosing a frame. He holds up the red crayon because he’s long past talking. The tube beneath his chin put an end to that.

I take down the grazing horse and put up the fire truck. There’s only a small stretch of bear wall in this tiny room, above the child-sized desk piled high with yesterdays’ pictures.

I want the exhibition to end and hate myself for it.

Okay, ChatGPT came through with flying colors, using the correct words finishes, his and bare. Two things: 1) Mistakes were not pointed out separately. I had to review the corrected reprint to check the work; 2) Since The Machine has a reputation for making mistakes, I had to proofread the story again. Seems redundant to me, unless you’re a horrible proofreader to begin with.

If you’d like to hear this story read in all its pre-GPT glory, check out the Micro podcast.

Leave a comment