It was October 2022. I was attending a VetPartners conference in Portland, Oregon, when a friend and colleague of mine asked, “So, what do you think of this ChatGPT thing?”
“I haven’t heard much about it,” I responded. “What is it?”
With “It can do this” and “It can do that” and “It’s so fast and smart,” my colleague proceeded to tell me all about ChatGPT and its super-human abilities—especially when it comes to writing. A few minutes later, I found myself on the OpenAI website, entering my first prompt into ChatGPT.
Write a blog post about the importance of brushing a dog’s teeth.
Immediately, words began to appear on my screen. It was as if a human were on the other side of the chat message, thinking quickly about what to write and then typing it.
As pet owners, we all want to ensure that our furry friends are healthy and happy. However, one area that is often overlooked is dental hygiene…
The words just kept coming. Within 30 seconds, I was staring at a blog post written by a chatbot.
As the owner of a veterinary writing company, I was terrified. But I was also in awe. How could something that takes a writer about two hours be done by a chatbot in less than a minute? What does this mean for the future of writers and editors? What does this mean for my company?
Since that day, I’ve learned a lot about artificial intelligence (AI). I’ve written articles for industry publications about AI’s potential applications in veterinary medicine. I’ve signed one client’s contract addendum prohibiting the use of large language models like ChatGPT. I’ve met with other clients curious about how we might use AI to write more content in less time—and for less money. I’ve attended webinars, listened to podcasts, watched news stories, and read articles.
“AI won’t replace humans, but humans who use AI will replace those who don’t use AI.” This has been a common theme I’ve come across during my 16 months of AI research and experimentation.
To avoid being replaced by a copywriting agency that uses AI to increase efficiency and productivity, I asked the Rumpus Writing team to begin playing around with ChatGPT. We’ve experimented using it to brainstorm ideas, get inspiration for titles or section heads, create outlines, and even write starter “skeleton” blog posts that we can beef up and make our own.
I started an AI Slack channel, where our team discusses their AI wins, frustrations, and lessons learned—and often posts ridiculous ChatGPT responses so we can all get a laugh during our workday. We created an AI folder, where we can share documents that include our ChatGPT prompts, the bot’s responses, and how we would edit those responses so our work would still be creative, original, medically accurate…ours.
Read on for a glimpse into what the Rumpus Writing team has learned during our AI adventures.
ChatGPT’s response length is limited
While there isn’t an exact word count limitation to a single ChatGPT response, we’ve found that most of the chatbot’s responses are fewer than 500 words. That first blog post I asked ChatGPT to write back in 2022 was only 351 words. Today, I asked ChatGPT 4 to write an 800-word blog post, and its response was 549 words. However, ChatGPT is a conversational chatbot, so you can ask it to elaborate on a topic or add sections to a previously generated response, and it will do so.
ChatGPT includes too much fluff and not enough substance
Remember those two-page, double-spaced essays you had to write in middle school? To meet the length requirement, many of us would slip in unnecessary filler and fluff, but skimp on the meat of the topic. Sometimes, that’s what it feels like ChatGPT is doing. We’ve found that most of its responses are full of adverbs like “extremely” and “very,” but missing necessary or relevant information.
A veterinarian friend of mine recently shared a survey she sent to her clients. I could tell immediately that she used ChatGPT to write the introduction. This is an actual paragraph from the five-paragraph—yes, you read that right—survey intro:
Your feedback is immensely valuable to us as we strive to continuously improve our services and better meet the needs of our cherished clients and their furry family members. We truly appreciate your participation in this survey and look forward to incorporating your insights into our wellness packages.
As a writer, my soul dies a little when I read that. I’d rewrite the entire thing, but even a nonwriter could take out the unnecessary fluff—immensely, continuously, cherished, truly—and make it more digestible for the reader.
And, while ChatGPT includes too much of the unnecessary, it sometimes leaves out too much of the necessary. When it comes to pet owner-facing content about pet health, we’ve found that most of the chatbot’s responses are vague and lack substance.
I recently asked ChatGPT to write a blog post about winter weather safety for pets. The response included the following section:
Understand Your Pet’s Cold Tolerance
Different pets have varying levels of tolerance to cold weather. Factors like age, coat type, body fat stores, activity level, and overall health can influence how well your pet can withstand the cold. For instance, breeds with thick coats are generally more cold-tolerant, but no pet should be left outside for long periods in cold weather.
I followed up by asking ChatGPT which dog breeds are more sensitive to cold temperatures. It supplied 10 breeds, along with a short description of each, which I used to improve the bot’s original response.
ChatGPT is often wrong
Chatbots can produce incorrect information—referred to as “hallucinations”—and they can be quite convincing when doing so. According to a recent New York Times newsletter written by David Leonhardt:
A lawyer representing Michael Cohen, the onetime fixer for Donald Trump, recently submitted a brief to a federal court that mistakenly included fictitious court cases. As it turns out, a Google chatbot had invented the cases.
ChatGPT can help the average person with mundane writing tasks
Simple emails, job ads, employee handbooks, letters of recommendation… ChatGPT can help with any of these basic writing tasks. But remember to closely review and edit anything AI produces.
Our team has learned that if your first prompt doesn’t produce the results you’re looking for, don’t hesitate to start over with a different prompt or ask the bot to elaborate on or revise its original response. Some of us have spent more time editing ChatGPT’s responses than we would have spent writing something from scratch. Click here to watch how this process works.
ChatGPT’s voice can be inconsistent
Some of the bot’s responses sound too robotic and cold. Other times, it seems like ChatGPT is trying too hard to not sound like a robot.
One Rumpus writer recently posted this in our AI Slack channel:
I asked ChatGPT to write a letter from a pet’s perspective about holiday safety, and this is the beginning: “Dear Furry Friends and Their Human Companions, Woof, meow, chirp, and whatever other sounds we use to communicate our thoughts and feelings!”
Another Slack conversation between two Rumpus team members:
Rumpus Writer 1: I love when I ask for a rewrite to make an introduction more interesting or entertaining and this is what I get: “Ahoy, pet enthusiasts! Living the fur-filled life with your four-legged pals is undoubtedly an adventure, but every now and then, our furry comrades find themselves in situations that ruffle their fur – vet visits, grooming sessions, or just a good ol’ pawdicure. Fear not, intrepid pet owners!…”
Rumpus Writer 2: Well, pirates need pet education, too.
At this point, large language models are shiny and new and exciting. They can be useful tools for busy people who don’t have a lot of time to spend on writing tasks. But, just like anything in its infancy, these bots need supervision and oversight. Since I first learned about ChatGPT, my initial awe has often been overtaken by curiosity, disappointment, and frustration.
No one knows what the future holds for large language models and other AI tools. Lawsuits have been filed. Politicians and others have called for regulation. All the while, these tools are learning, growing, and changing the way humans accomplish many tasks.
If you’ve played around with ChatGPT and have found its responses to be lacking in personality, substance, and medically accurate information, reach out to Rumpus Writing for help with your content. We’re still humans over here.