This column written by actual human
I got my feelings hurt…by an internet robot. It’s my own fault. I’ve heard so much about ChatGPT, the language processing bot, and I had a deadline coming up. So I decided to let artificial intelligence write my column for me. You might think that’s cheating, but I thought it would be fun. I thought it would be easy. I thought I could get away with it.
Maybe it was a busy day at OpenAI, the company that created ChatGPT, or maybe it was user error. But I could have written my own column and painted my house in the time it took me to set up my account and figure out how to use it. I didn’t, but I could have.
First I got something called a gateway error. I’m pretty tech savvy so I immediately thought what the heck is a gateway error.
Then I got a message saying all systems were busy at the moment and would I mind taking a break and trying later. I never mind taking a break, so I went away for the weekend.
When I came back, I sat down at my computer, logged into ChatGPT and ... ta-da, my robot writing assistant was ready to work. That was lucky because I wasn’t.
There were two disclaimers, but neither of them worried me. ChatGPT confessed that it may occasionally generate incorrect information. I didn’t see that as a problem because I do, too.
It also admitted it has limited knowledge of the world and events after 2021. My limited knowledge goes further back than 2021 and I’ve never let that stop me.
So I typed in my request: Write a column about ChatGPT in the style of Dorothy Rosby. I was hoping it would sound like I wrote it — only better. I was hoping it would make me sound clever and witty. I was hoping it wouldn’t ask, “Who’s Dorothy Rosby?”
It didn’t. It just started writing and poof! It was done faster than you can say, “What does GPT stand for?”
I thought ChatGPT was going to be the best thing that’s happened to my writing career since the invention of spellcheck — until I read what it wrote. There were two problems with it. One, it was too short. You might not see that as a problem.
Two, it stunk. And that’s when it hit me. In order to write a column in my style, ChatGPT scoured the internet for other things I’ve written. And based on what it found there it wrote insipid drivel. What does that say about my style? Don’t answer that.
I noticed right off that the column included the word “quirky” eight times. I don’t use quirky that often. I don’t use any word that often. I have an extensive vocabulary — and a thesaurus.
And ChatGPT is not only inaccurate, it’s a liar. Writing as me, it started out by saying, “I want to introduce you to my friend ChatGPT.” It said we met while I was drinking coffee and lazing around one afternoon. I don’t drink coffee and if it were really my friend, it would know that. It was right about the lazing part though.
It said I was “impressed by its vast knowledge.” Was not! Okay, maybe a little. But it also said I was “charmed by its humor and witticism.” What witticism? It said we “engaged in delightful banter and a lively exchange of ideas.”
Did not! I made one request and it did all the talking—or writing. And delightful and lively are not the words that came to mind when I read what it wrote—in my style.
In the end the whole effort was a waste of time. Not only that, it made me mistrust AI even more than I did already. Plus now I’m questioning everything I’ve ever published online. And after all that, I still had to write my own darn column.
Dorothy Rosby is the author of I Didn’t Know You Could Make Birthday Cake from Scratch: Parenting Blunders from Cradle to Empty Nest and other books. Contact her at www.dorothyrosby.com/contact.