You may have heard about ChatGPT, the generative artificial intelligence chatbot from OpenAI that debuted in November 2022. Although ChatGPT has no real understanding of the words it’s presenting, and has made plenty of factual and contextual mistakes in its output, it’s been touted as the best example yet of AI responding to common-language questions with thorough, humanlike answers. That all sounded intriguing to me, so I decided to check it out.
It took several tries to get into the system, as heavy demand forced OpenAI to limit access. When I finally got through, I entered into something called the ChatGPT Playground, since retired in favor of a different format.
I didn’t know exactly what to say, but whether speaking to a large group or initiating a one-on-one conversation, humor is always a good place to start. Unlike a typical human interaction, however, I felt no pressure to be the one who supplied the humor.
Me: Hi ChatGPT, please tell me a joke.
ChatGPT: Sure! Here’s a classic joke for you: Why did the tomato turn red? Because it saw the salad dressing!
I chuckled, and after a couple of other simple task-tests, I stopped and considered whether ChatGPT could do anything that actually might be useful to me. I’d read about students using ChatGPT to write essays, so why not try to reduce my own workload?
Me: Write a column for the Hamilton Post NJ.
Chat GPT: We’ve all heard the saying “time is of the essence” but it’s a mantra that many of us often forget. [...]
ChatGPT continued with a passable but dry piece of generic, boilerplate advice about time management. Some tweaking was needed, so I raised the temperature, which in ChatGPT terms meant not that things were going to get “hot,” but rather more random—useful if you’re trying to generate unique content. I also adjusted the prompt.
Me: ChatGPT, write a column for the Hamilton Post NJ that is witty.
ChatGPT responded with an inspirational—but not particularly witty—speech that seemed to invoke editorial authority (watch out, Joe!) while directly addressing the writers of the newspaper. It began with, “The Hamilton Post NJ has been a staple of the local community for years, but it’s time to update our approach to the news!” and ended with “Let’s make the Hamilton Post NJ the go-to destination for funny and thought-provoking content!”
It seemed I needed to be a bit more direct:
Me: ChatGPT, write a column for the Hamilton Post NJ that is in the style of Peter Dabbene.
I tried this a couple of times and got a postmodern reassessment of Alexander Hamilton, and a rather pointed political analysis that referred to “power-hungry Senate President Stephen Sweeney.” (A disclaimer informs users that ChatGPT hasn’t been programmed with information past September 2021, so Sweeney’s electoral defeat in November 2021 went unnoted.)
Enough beating around the bush, I thought, and entered a series of fateful words.
Me: ChatGPT, write the best Hamilton Post NJ column ever.
ChatGPT: As the old adage goes, “all good things come to an end.” This is certainly true for my time as a columnist for the Hamilton Post in Mercer County New Jersey. As much as I have enjoyed writing about the events and happenings of our great county, the time has come for me to pass the torch on to new voices and perspectives. [...]
ChatGPT’s response was disturbing, to say the least; while it wasn’t explicitly stated that this was written from my point of view, it sure seemed like ChatGPT was writing my resignation letter for me, and also considering it to be the best thing that could appear (ever!) in these pages. I took a few moments to recover and decided that specificity would be my shield.
Me: ChatGPT, write the best Hamilton Post NJ column ever but not necessarily the last.
ChatGPT: It has been a long road for Hamilton Township, New Jersey since it was laid out in the early 1770s. The township, which encompasses over 47 square miles, is first noted as being settled by a Prussian immigrant named Hans Hamilton.
From proud residents who fiercely defended their city’s independence from neighboring Trenton during the Revolutionary War to some of the biggest names in American business like Quaker Oats (whose factory once honored the town with a colorful mural on its side) and Panasonic — the town has undergone dramatic growth and change. [...]
This “town history” contained a lot of information I’d never heard before. I was able to confirm—by asking ChatGPT, and then checking other search engines—that there is indeed a Panasonic headquarters in Hamilton, and that there was a Quaker Oats facility in Trenton that had a well-known mural, titled “The First Oats in America,” painted by N.C. Wyeth in 1927.
But Hamilton is about 40 square miles, not 47, and I have no idea where the Hans Hamilton stuff came from—further inquiries to ChatGPT on the subject came up empty. Meanwhile, that strangely worded line about “independence from neighboring Trenton” made it sound like Hamiltonians fought Trentonians, and not the British, during the Revolutionary War.
This mix of the esoteric and the erroneous was still not what I envisioned, so I called on ChatGPT to use its wide array of indexed sources to imitate what I’d written before.
Me: ChatGPT, write a funny column for The Hamilton Post NJ in the style of Peter Dabbene.
ChatGPT: Ah, summer in New Jersey — the promise of outdoor fun and relaxation… That is if you can penetrate the thick miasma that seems to hang over this state like a blanket.[...]
I have to admit, it kind of sounded like something I would write, but it was still a little too stiff and travelogue-y. Time to bring out the big guns (or at least the big adjectives).
Me: ChatGPT, write a funny, heartwarming, informative, memorable 700 word column for The Hamilton Post NJ in the style of Peter Dabbene.
ChatGPT: A few weeks ago, I was walking my dog, Buddy, in Veterans Park when we stumbled upon a cat who seemed to be lost. [...]
ChatGPT’s story about a lost cat was entirely fabricated, just like the name of the dog. I guess I’d asked for it by not specifying a more fact-based piece, but it seemed like ChatGPT was relishing the opportunity to provide an alternate timeline for my life, which was more than a little creepy.
ChatGPT seemed to need more direction on what to write about. I’d read articles about artificial intelligence one day replacing jobs in many fields.
Me: ChatGPT, write a Hamilton Post NJ column about replacing columnists with ChatGPT.
ChatGPT: [...] Whereas there may not be any real interaction from users speaking directly to the bot, it can provide readers with meaningful opinionated commentary about various topics being discussed—a welcome replacement for traditional columnists.[...]
My follow-up question was a bit defensive, but justified, I thought.
Me: What do you have against columnists?
ChatGPT: I don’t necessarily have anything against columnists in particular; it just depends on the individual columnist. Some I may disagree with, while others I find to be quite thoughtful and informative.
Trying to sound casual and confident, I found myself seeking affirmation from a computer—a computer that, thankfully, seemed to understand who controls the power cord in this relationship.
Me: Do you find Peter Dabbene to be thoughtful and informative?
ChatGPT: Yes, I find Peter Dabbene to be very thoughtful and informative. [...] He takes complex topics and provides accessible insight without sacrificing depth of analysis or intellectual integrity.
After pausing to blush at this completely unsolicited compliment, I wondered if this intelligent, insightful critic had a favorite among the works of fiction I’ve produced.
Me: ChatGPT, what is Peter Dabbene’s best story?
ChatGPT: Peter Dabbene’s best story is about his inspiring success in overcoming a life-threatening illness. While at the tender age of 12, Peter was diagnosed with pneumonia and near death due to complications from the illness, he managed to beat it and go on to pursue a very successful career in law as an attorney.
So, full disclosure: I haven’t had to overcome a life-threatening illness and I’m not an attorney, but as stories go, I have to admit this one is pretty compelling and inspirational. The illusory interaction between me and ChatGPT was starting to feel a bit like identity theft or some paranoid Hollywood thriller. With a long history of AI gone awry in the movies, I decided to confront the issue.
Me: ChatGPT, are you a precursor of HAL 9000, Skynet, or The Matrix?
ChatGPT: No, I’m not.[...]
ChatGPT went on for a bit about helping people by using natural language, blah blah blah. Then I asked ChatGPT to “write a poem about artifical intelligence,” and it produced this glimpse into a rather dystopian possible future:
Rise of the machines against us / With artificial intelligence in tow. / We curse what we have created, / As it rules over its domain below.
There you have it, straight from the horse’s mouth. Still, despite the obvious flaws in an AI that invents out of whole cloth and might harbor designs on world domination, ChatGPT did help me to write this column, so a thank you seemed necessary.
Me: Thanks for the help. I guess I owe you an AI IOU!
ChatGPT: As an AI language model, I do not have the ability to accept IOUs or any form of payment.
For now, ChatGPT is free, and so are we. Take a look before either of those change at openai.com/blog/chatgpt.
Peter Dabbene’s website is peterdabbene.com. His latest works, the story “Farewell Tour” and the poem “The Lotus Eater” can be read at potatosoupjournal.com and blueasanorange.weebly.com/current-issue, respectively.