Grammar Moses: You can fool some of the people some of the time

Judging by the first response I received after last weekend's column in which I went up against the ChatGPT artificial intelligence human interface, the next assignment I considered giving ChatGPT was to “write a letter of resignation admitting I am of less use than a 1970s-era Texas Instruments calculator.”

David Borck, who has been reading my column from its infancy, wrote: “Seems that YOU are Example A, but truthfully, it is hard to tell.”

Of the two “humorous” column items about gerunds, I wrote Example B (in 5 minutes) while ChatGPT “compiled” its column items in 40 seconds.

But given that 23 of the 28 people who responded guessed correctly, I “tore up” that letter of resignation even before I could ask ChatGPT to “write” it for me.

“The jokes weren't funny in Example A,” wrote Sue Knisely.

“Example A was confusing, repetitive and didn't summarize the key points in a well written way,” wrote Linda Puetz. “So, I hope Example A was written by a poorly developed ChatGPT in its infancy. However, I do envision the day when the AI software is much more highly developed and capable of excellent writing, having been programmed and monitored by expert humans.”

Cheryle Staats said, “Example A seems to think you must keep telling the reader how funny the topic is, whereas Example B is just funny and I didn't have to be told that.”

Sarah Poeppel also picked correctly. Saying that, “Example B is written in the style that I associate with your writing for this column, and Example A is not in that style. Example A is not as well written as Example B. Example A is written as though the 'author' is trying too hard to be humorous and to make the point. It is not written in a style that seems natural; it is contrived.”

Ken Valentine said Example A is choppy and doesn't match my writing style to which he has grown accustomed.

“I would never expect you to use the word 'cheesy' in a column or joke,” he said. “Example B flows smoothly, the examples are unforced, and it's generally more pleasant to read. This is the sort of prose I expect from an experienced writer.”

Scott Zapel wasn't fooled for a second. “The lede told me right off the bat,” he wrote. “'Today, I want to talk about gerunds.' It has a schoolboy simplicity. The Example B version is much 'punchier.'”

Lucy Delmonico couldn't quite put her finger on it, but she said Example A didn't sound like me. “Maybe because Example B sounds more personal. Would AI be shopping at Costco on Saturdays? And would instructing it to write a funny column result in shivering, worrying and cat vomit? These sound more like you!” she said. “I did learn something about gerunds. More from B than A, which means B was written by you.”

Kurt Stauff graciously pointed out the frequent use of incomplete sentences in Example A: “That is something I would rarely, if ever, expect from you. Example B makes use of more concise explanations and a better sense of humor (congratulations!).”

Mary Lou Wehrli wrote that Example B “has your voice. B opens with tight language of an editor. B has your usual personal touches, making the reader feel part of Grammar Moses' world. B has your personal humor. I hope the cat is better. B is for Best.”

Rick Dana Barlow wagered Example A was the AI because it followed a textbook skeleton sentence structure with little variation and that it “overdoes the conversational tone with a conclusion that is more greeting card-stand-up monologue routine. However, Example B demonstrates organic creativity apparently out of range and reach of a machine centering on 'logic.' Example B also finishes with human-conjured humor that a series of codes could not re-create.”

Jamie Smith thought I might have tried to sneak one by you by purposely writing something stilted and “adolescent.”

But she chose correctly.

“Example A is AI for the simple reason that it sounds too much like an adolescent trying too hard to be funny,” she said.

Bruce Steinberg's email opened with: “If you wrote Example A, I'd cancel my DH subscription.”

So, I guess more was at stake here than my continued employment.

“The ChatGPT produced a result without specifics, empathy, humor, charm or warmth. It read like an outline,” Bruce wrote.

Karen Daulton Lange said, “Your humor is more subtle, and Choice A tried way too hard to be funny.”

Even Lisa Tuttle, who said she just discovered my column, wasn't fooled. “After careful consideration, I have decided the personal touch of wiping up cat vomit and shopping on Saturdays at Costco must come from your own horrifying experiences,” she said.

Horrifying, indeed.

But as the saying and the headline go, you can fool some of the people some of the time. The AI tricked Dave Wcislo, who wrote, “I hope you're Example A in this week's article. You're way more amusing than ChatGPT. (I hope.)”

Gulp.

Ken Klein also was mistaken.

“Example A is the real Jim Baumann,” he wrote. “Example A is a much warmer style, and I can't imagine ChatGPT using a phrase like, 'Yeah, those guys.' Example B seems almost too perfect and nonhuman.”

What I'm taking from that, Ken, is that I'm almost too perfect. Thanks.

Sheila Barrett noted that “no self-respecting generative, pretraind transformer would use 'verb-ing' once, no less three times.”

Perhaps ChatGPT hasn't learned self-respect yet.

Paul Sill, an adjunct professor at DePaul University, said he has been tinkering with ChatGPT, too.

“My guess is that ChatGPT wrote Example B in your column today. I think the slightly more witty Example A writing was done by you, the human,” he said.

Rita Boserup didn't buy that an AI would use the word “cheesy.”

Lastly, Mike Piorkowski, whom I must have disagreed with vociferously at some point, chose correctly but had this to say: “Example A is AI. It's much wittier and more interesting than your boring Example B.”

Hey, you can't win them all.

What all of this tells me is that to come across as a human columnist, it's always best to write from the heart, personalize it and never, ever, ever explain a joke.

*End of transmission*

Write carefully!

• Jim Baumann is vice president/executive editor of the Daily Herald. You can buy Jim's book, “Grammar Moses: A humorous guide to grammar and usage,” at grammarmosesthebook.com. Write him at jbaumann@dailyherald.com and put “Grammar Moses” in the subject line. You also can friend or follow Jim at facebook.com/baumannjim.