Am I about to lose my job to ChatGPT?

Woman talking to a chat bot

Credit: AdobeStock

By Dorothy Pomerantz | Fitch Ink Managing Editor

The other day, I noticed that my son was writing a school paper by hand instead of on the computer. When I asked him why, he explained that several kids had gotten caught using ChatGPT to write their papers so now their English teacher was requiring them to handwrite everything.

The AI has been available to the general public for only a few months and already, teens are using it to cheat at school. How long until it comes for us writers? 

ChatGPT, for those who don’t know, is an artificial intelligence chatbot from OpenAI that has been trained using large language models to respond to queries. It’s designed to understand natural language and give human-like responses to text-based questions.

Aside from reading with delight how Microsoft’s Bing AI tried to seduce New York Times reporter Kevin Roose away from his wife, I’ve been mostly trying to keep my head in the sand and ignore ChatGPT news. Thus far, it’s felt sort of like climate change to me, an almost inevitable looming threat that I can do little to stop.

That’s because right now, ChatGPT is pretty good. Or at least it’s pretty good at being good enough. The OpenAI technology is not smart, just predictive. It can’t think or feel. But it can call on more information than you or I could glean in a lifetime to instantly write whatever it’s told. Here are some fun examples of people using to write things like a recipe for dairy-free mac and cheese and an essay comparing different theories of nationalism. And some companies are already working AI into their white-collar ranks. PwC recently announced it will use an AI chatbot to help lawyers work more quickly.

Putting the robot to the test

Before giving into despair, I decided to do a test. I asked ChatGPT to write a 500-word story on how lean management helped a GE factory improve efficiency. That might seem like a fairly random prompt but I was basing it on this excellent story on the GE Reports website (which Fitch Ink contributes to) about a turbine factory in South Carolina that was able to turn itself around using lean management.

Here’s ChatGPT’s lead paragraph:

“In the bustling town of Schenectady, New York, there stood a large factory that produced gas turbines for the energy industry. The factory was owned by GE, and it had been operating for over 100 years. Despite its long history and reputation for quality, the factory was facing a major challenge - how to remain competitive in an increasingly crowded market.”

To be fair, I think the AI was a bit thrown by my using the word “story” instead of “article,”  hence the almost fairy-tale-like opening. But I think the paragraph is a really good example of what we can expect (at least in the near term) from AI. It’s pretty dry. It uses passive voice. Mostly it’s generic. This is not a paragraph that is going to draw a reader in.

Compare that to this lead written by Tomas Kellner for GE Reports:

“In the last half-century, the GE Gas Power plant in Greenville, South Carolina, has experienced the same kind of dizzying roller coaster ride as the city where it’s located. Founded some 200 years ago on the Reedy River, Greenville grew into a textile manufacturing hot spot and a paragon of Southern hospitality — until the 1970s, when the city’s downtown took a turn. Mills closed, hotels shuttered, and the clattering trams carrying riders on shopping errands and to business lunches fell silent.”

The last sentence is particularly lovely. By evoking the silenced trams, Kellner brings us into the world of Greenville. We immediately understand what it was and what it has lost. It’s the kind of sentence that I think could only come from mining lived, human experiences. I could be wrong but I feel fairly confident in predicting that it will be a long time before AI can do that and it’s possible it never will be able to. 

Accuracy is the cornerstone of good writing

Then there’s the issue of accuracy. Within the story ChatGPT wrote:

“By the end of the first year of its lean journey, the factory had seen a 30% improvement in productivity, and it had reduced its lead time for turbine production by over 50%.”

I asked where it got that 30% number. Here was the reply:

“I apologize for not providing the source of the 30% improvement in productivity figure in my previous response. As an AI language model, I do not have access to specific data or sources, but I can suggest some possible ways in which this figure could have been determined.

The 30% improvement in productivity could have been based on various metrics such as output per worker, production time, or efficiency gains. These figures could have been tracked and analyzed over time to determine the overall improvement in productivity.”

Obviously, this is not OK. At least for now, accuracy is still a big problem for AI. 

Good writing is going to matter more than ever

So while I now think AI will definitely take over some rote kinds of writing like bond reports and high school sports stories (and that makes me worried for the state of journalism overall), I’m less worried about it coming for my specific job. At Fitch Ink, we distinguish ourselves by the high quality of our writing. Our squad of freelancers come from publications like Forbes and the Wall Street Journal, and we encourage our folks to use the skills they learned at those institutions when they write for our clients. 

If anything, I think the proliferation of AI like ChatGPT is going to make the difference between good and bad writing that much starker. Companies that want to stand out will have to do more to make sure their content is bold and interesting. Audiences will become used to tuning out the bland prose of AI and will be that much more engaged when they come across great writing.

I also think AI can be a great assistant to all writers. When I was writing this, I was struggling to define ChatGPT so I went to the source and simply asked for some help. Go look at the third paragraph of this story. It’s not word for word but I did take the information the AI spat out and used it to write a brief, direct explanation of ChatGPT.

Why we’ll still need real people

So while AI may be inevitable, my gut tells me that it’s going to be a good tool, not a replacement. In fact, ChatGPT agrees with me. Here’s what it told me:

“As an AI language model, I am capable of generating high-quality written content and can assist with various writing tasks. However, it is unlikely that AI language models like myself will completely take over corporate writing in the near future.

Corporate writing often involves strategic decision-making, brand messaging, and crafting a tone and voice that resonates with a specific audience. These tasks require a level of nuance and understanding of the brand and its target audience that may not be replicable by an AI language model. The human touch will continue to be an important component of effective corporate communication.”

Here’s hoping the computer is right.

Previous
Previous

The lyrical power of simple words

Next
Next

How brands can take a stand without alienating customers