ChatGPT has been getting a lot of press lately as an example of how generative AI can handle tasks ranging from writing computer code to drafting articles. Depending on whom you believe, ChatGPT poses an existential threat to Google’s dominant search engine, could signal the end of take-home tests, and might put writers and editors out of business.
I’m in no position to weigh in on the first two items, but I have a vested interest in the third one. So I spent a little time diving into ChatGPT to understand its uses and limitations. Here’s how it fared on a few different content requests:
Writing fact-based articles: Good
I started with a straightforward request: a presentation of publicly available facts.
Query: Write a three-paragraph bio of Dick Butkus.
Richard Marvin ‘Dick’ Butkus is a former American football linebacker who played for the Chicago Bears of the National Football League (NFL) from 1965 to 1973. He is widely considered one of the greatest and most intimidating linebackers in the history of football. Born in Chicago, Illinois, Butkus played college football at the University of Illinois, where he was a two-time consensus All-American. He was selected by the Bears with the third overall pick in the 1965 NFL Draft and went on to play for the team for nine seasons.
Butkus was known for his physical and aggressive style of play, and he was a dominant force on defense for the Bears. He was named to the NFL Pro Bowl eight times and was a first-team All-Pro selection six times. He was also named to the NFL’s All-Decade Team for the 1960s and 1970s. In addition to his football career, Butkus has also worked as an actor, appearing in television shows and movies such as The Bear and The Replacements.
After retiring from football, Butkus became a successful businessman and philanthropist. He has served on the board of directors for several companies and has also worked as a television commentator for NFL games. He has also been involved in numerous charitable causes, including the Dick Butkus Foundation, which supports programs that promote health and fitness for children.
It’s not surprising that ChatGPT nailed this task. Articles created by AI tools have been in circulation for more than a decade. If you recently read a preview or recap of a sporting event, there’s an even chance the author was a machine. Articles that follow a similar format and draw on stats, quotes, and other data are an ideal use of generative AI. This type of article wouldn’t be appreciably better if written by a human, but it would take exponentially longer. Bravo, ChatGPT.
Constructing arguments: Meh
Next, I wanted to see how the tool would do in creating a viewpoint based on existing sources.
Query: Write a two-paragraph introduction for an article on why gentrification in Chicago’s Pilsen neighborhood is bad.
Chicago’s Pilsen neighborhood has long been a vibrant, diverse community with a rich cultural history. However, in recent years, gentrification has threatened to change all of that. As property values and rents have increased, many long-time residents have been forced to leave the neighborhood, unable to afford the rising costs of living. This has led to the displacement of families, the loss of cultural diversity, and the erosion of the tight-knit community that has always defined Pilsen.
Additionally, the influx of new residents and businesses often leads to the loss of small, locally-owned establishments that have served the community for decades, replaced by more expensive and less connected alternatives. The impact of gentrification on Pilsen has been profound and negative, and it is important to address this issue in order to preserve the unique character of the neighborhood.
This introduction is missing several elements of distinctive, insightful articles: a sharp perspective (I’d argue the one above could apply to just about any gentrified neighborhood; for example, it contains no mention of the ethnic background of Pilsen’s residents), a unique angle on the issue (think of how much more effective it could be by starting with a profile of a resident of the neighborhood), and facts and figures that reinforce what’s at stake (the rise in property taxes or the median income of longtime residents, for instance).
But you can see why teachers are so worried; the intro is right in line with school writing assignments. This style of article is more representative of SEO campaigns that aim to churn out a steady stream of content featuring selected keywords. My issue with such articles is they ultimately aren’t that interesting and often provide little value to the reader.
So feel free to use generative AI to write these articles—just don’t expect people to read much past the first couple paragraphs.
Composing creative works: Please, don’t
Some have heralded generative AI’s ability to draw on available sources to actually be creative, such as writing a passage in the style of the King James Bible. So I gave it a try.
Query: Write song lyrics in the style of Pete Townshend.
Verse 1:
I wake up in the morning, feeling tired and drained
The weight of the world, it feels like it’s on my shoulders again
But I can’t give up, no matter how hard it gets
I’ve got to keep on fighting, with everything I’ve got left
Chorus:
Cause I’m a survivor, I won’t let life bring me down
I’ll keep on going, round and round
I’ll face each day with courage and with grace
I won’t let anything stand in my way, no matter the place
Verse 2:
I know there will be times, when I’m feeling low
When the road ahead seems long, and the path is hard to know
But I won’t let fear, control my fate
I’ll rise above it all, and I won’t hesitate
Chorus:
Cause I’m a survivor, I won’t let life bring me down
I’ll keep on going, round and round
I’ll face each day with courage and with grace
I won’t let anything stand in my way, no matter the place
Bridge:
I know that I’m not alone, in this journey through life
There are others by my side, through the struggles and strife
We’ll lift each other up, and we’ll face each day with hope
Together we’ll rise above, and we’ll find a way to cope
Chorus:
Cause we’re survivors, we won’t let life bring us down
We’ll keep on going, round and round
We’ll face each day with courage and with grace
We won’t let anything stand in our way, no matter the place
That’s truly awful. It has none of the signature elements of Townshend’s best songs—such as spirituality and London Mod culture. (This pablum might work for Bon Jovi, however. Quit hanging out with Robert Kraft and book some studio time, Jon!)
The goal of any content should be to provide insight and shed some light on the authors’ viewpoint. You’re not going to get that from ChatGPT yet. And it makes sense: if generative AI is scanning all relevant content, any uniqueness would naturally be washed away.
Conducting research: A great starting point
The most valuable use of the current iteration of ChatGPT for content is research, often one of the most time-consuming steps of writing.
Query: What are the top five examples for how AI can make businesses more productive?
1. Automation of repetitive tasks: AI can automate routine and repetitive tasks, freeing up time for employees to focus on more value-added activities. For example, AI can be used to process invoices, generate reports, or respond to customer inquiries.
2. Data analysis and decision-making: AI can analyze large amounts of data quickly and accurately, enabling businesses to make more informed decisions. For example, AI can be used to predict customer behavior, identify market trends, or optimize supply chain operations.
3. Personalization and customization: AI can personalize and customize products and services, improving the customer experience and driving loyalty. For example, AI can be used to recommend products or services based on a customer’s browsing history or to generate personalized content for marketing campaigns.
4. Predictive maintenance: AI can predict when equipment is likely to fail and schedule maintenance before it does, improving equipment uptime and reducing downtime.
5. Process optimization: AI can identify inefficiencies in business processes and suggest improvements, increasing productivity and reducing costs. For example, AI can be used to optimize routes for delivery trucks or to streamline manufacturing processes.
This result provides a valuable starting point for education and further research. Rather than having to read ten articles to gain context, a writer could use ChatGPT to provide this result in seconds and use it as direction for a deeper dive. Still, a couple of issues: First, it doesn’t provide original sources. Second, one of the most important decisions a writer or editor makes is prioritizing research insights to develop an informed perspective. When the machine does that work, it could close off an important and necessary part of the process.
***
Generative AI is a powerful tool, but as with any technology, it’s just one part of the puzzle. The more valuable opportunity is figuring out how to combine the expertise of humans and the nearly endless capacity of machines to process information.
My colleagues and I plan on exploring these opportunities throughout the year and will share our findings.