How I Use ChatGPT
And additional thoughts on the rapid rise of AI technology
I am just over four years into writing The Sunday Morning Post, and my main regrets have mostly been around wishing I had more time in the week to research and write, but such is life with a full-time job, three busy kids, and a wife whom I am rather fond of and to whom I would like to stay married.
One regret that does stand out is that I haven’t written in detail much about AI yet. It is the topic of our time. In the early part of 2023, I was researching and reading about AI pretty heavily, and I was eager to write about it here. ChatGPT had been launched to the general public just a few months prior in November 2022, and there was a sudden whirlwind of activity and interest. I had stubs of several different articles on AI-related topics in my drafts folder, but I never finished any of them off.
Part of the reason why AI has been such a challenging topic to write about is that it is evolving so fast. Never in the history of human events has a technology become so ubiquitously used in such a short span of time — cars, TVs, computers, smart phones — none of them compares to the rapid rise of AI technology.
ChatGPT today boasts over 800 million weekly users and handles over 1 billion inquiries a day. Per recent polling from the Pew Research Center, 34% of U.S. adults have used ChatGPT, with 58% of those under age 30 having used it. I suspect a year from now, well over half of adults will have used the program, and in the years to come, the use of ChatGPT and other programs like it will have become as ubiquitous as Googling things has been over the past 20 years.
If you’re reading this and thinking, “Huh, AI. Not for me,” you might be surprised to know how much AI has already been implemented in so many of the tools and apps most people use on a daily basis. Google Maps predicts traffic and can reroute you in real time. Autocorrect now incorporates predictive text to help people write more efficiently and eloquently. Streaming services like Netflix and Spotify personalize recommendations based on what their AI algorithms predict you will want to watch. Even a traditionally conservative industry like banking is getting in on the AI wave, with automated fraud detectors that can sense when something is amiss in your accounts and that will text, email, or call automatically to alert you. Companies of all types now commonly use AI chatbots, and the consumer often doesn’t even realize they are talking to a bot. In many ways, the bots are better than humans because they can be programmed to be more patient and informative than even the best of human operators. They also “work” 24/7 and don’t require salary and benefits, including vacation and sick time.
These are just simple examples. AI is being used in all kinds of ways from writing code to creating graphics and artwork to writing all types of communications including newsletters. I could even have ChatGPT or another program like it write an entire Sunday Morning Post article if I wanted it to (I don’t —more on that below).
Big Questions
Collectively, we will have to confront some significant questions in the years ahead about the ways AI is changing the ways we work, socialize, and communicate with one another. Will AI cost jobs or create them? Will AI eventually replace human workers altogether? And if so, is that five years out or fifty? What will society look like if that happens?
One fascinating and somewhat terrifying question is what will happen as AI becomes more powerful than both its users and programmers, and what if that AI technology is not aligned with our collective human values. In a hypothetical experiment from a 2014 book by Nick Bostrom, Superintelligence: Paths, Dangers, Strategies, an AI device is given the command of maximizing paperclip production. The device ends up destroying humanity because 1) human beings could shut down the machines that make the paperclips so therefore must be eliminated and 2) humans take resources that could otherwise be used for paperclip production.
It’s one thing to ponder about paperclip production, but what if the AI is manufacturing the weapons of war? What if it purposely creates a virus, either technological or biological? What chance would humans have? Questions like this have led some to ponder whether there needs to be an “AI kill-switch” so that things could be shut down if necessary. But who decides when that is? And who is in control when the kill-switch needs to be pushed? And how could a hypothetical kill-switch shut down everything? These questions are both fascinating and unsettling.
A Brave New World
All that being said, it is worth remembering that plenty of people have predicted humanity’s doom at the outset of other major technological advancements, including during the Industrial Revolution and with the rise of both cars and computers, among many other advances. Even the humble bicycle generated its fair share of moral panics. Per Pessimists Archives, “When the bicycle debuted in the 1800s, it was blamed for all sorts of problems—from turning people insane to devastating local economies to destroying women’s morals.”
With that in mind, we should refrain from kneejerk negativity about AI technology despite the aforementioned fears; neither the future outlook nor the current landscape for AI use need be entirely pessimistic. In fact, people are doing some pretty cool things with AI, leading to improvements in quality of life for many on a macro scale, and professional and personal advancement for many on an individual level — those who know how to use it, at least.
There is a quote that floats around tech circles that captures a wise sentiment or a cautionary warning depending on your perspective. The line is: “AI won’t replace you. A human using AI will.” In other words, AI is not going to eliminate humans from the workforce, but as AI is incorporated into all things, people won’t be able to keep up or accomplish the necessary tasks of their jobs if they don’t know how to use AI, eventually leading them to be laid off or reassigned to other tasks (whatever those might be in an AI-dominant world). Further, the implication is that there won’t be any jobs that don’t incorporate AI.
No industry is really immune from this, nor is anyone at any level. In many ways, those who have advanced professionally the farthest especially in white-collar settings like finance and law might be the most vulnerable. There is already evidence that this is happening as AI is able to compile and analyst vast reams of data that many in professional settings have spent their entire careers becoming the experts on.
This is akin to the transition that the white-collar working world, in particular, experienced first through the Computer Revolution of the 1980s and 1990s, and then subsequently with the Internet Wave in the late 1990s into the 2000s. Workers who didn’t adapt were transitioned into other jobs or let go. The same thing will happen with AI over the next decade. Barely a job exists today that will not be impacted by AI in the years ahead.
With that in mind, I thought it would be worth sharing how I use AI myself. I am certainly no expert, and I am always trying to learn more. And keep in mind I am primarily talking about ChatGPT here, which is not only the most well-known and prototypical example of an AI tool that regular people can use, but it is really the only one I do use (other than where AI is baked into apps and other tools we all use). There are dozens of other AI tools out there now that I don’t honestly really know much about. I also mostly use ChatGPT for the Sunday Morning Post and for personal inquiries; I don’t really use it in my banking job as I’m reluctant (and ethically restricted and regulatorily limited) to input information into a program that could compromise my clients’ personal or financial information.
Here are the main ways I currently use ChatGPT for my writing and personal use:
Research Assistant. Over the past year, I have had ChatGPT open in a secondary window on my computer next to my Substack writing page on nearly every article I have written. As I write, I ping ChatGPT with all types of questions, and the program finds data, references, and other insights from around the web far quicker than I could do so myself. For example, in my article a few months ago about Mark Carney being elected Prime Minister of Canada, I asked ChatGPT the following questions:
“Has Mark Carney ever held political office before?”
“What is the process of winning an election for Prime Minister in Canada?”
“What are a few things Mark Carney has said about Donald Trump?”
Each week, I use the ChatGPT responses to help organize my own thoughts, and often to stimulate further thinking and brainstorming.
A Google Replacement. I love Google. I think it is perhaps the most influential company that has ever been built. But the information-gathering abilities of ChatGPT for certain types of searches far surpasses what I can find in Google. Google now feels cluttered, disorganized, and overly monetized to me, which is a strange sensation as Google has always been the most organized and keenly accurate search engine out there. But now I find that inputting questions into ChatGPT is much more informative and efficient, and the whole user interface feels cleaner and smoother than using Google.
Proofreading. If readers of The Sunday Morning Post have noticed there are not as many typos and grammatical errors in these digital pages of late, the credit lies entirely with ChatGPT. The last thing I do before posting a new article is run my entire “final” draft through ChatGPT, asking the program to proofread for typos and grammar. Almost always, it identifies minor errors, which I then am able to easily correct. In the early year or two of The Sunday Morning Post, errors were more common (I am frequently writing this late-night Saturday, and my faculties are not always at their optimal best). ChatGPT has helped me eliminate most of those issues. Oh, and by the way, it does this in about 5 seconds for a 1,500-word article. If I were paying a professional proofreader, I would not only need to have my articles done by, say, Thursday or Friday for a Sunday morning time of publishment, but it might require several hours of back-and-forth with the proofreader and perhaps $100-$200 in cost per piece. As it stands now, I can use ChatGPT for free and do the proofreading Saturday night at midnight if I want.
General Banter and Back-and-Forth. This one is more for personal use, or sometimes as I am developing ideas for articles — ChatGPT sometimes feels like a friend who is always ready to brainstorm with you. Just the night before this article was written, my wife and I were playing cribbage and sharing a bottle of rosé wine (her summer favorite). I asked ChatGPT “does rosé wine have anything to do with roses or is it just because of the color.” She responded (I don’t know why I tend to think of ChatGPT as a “she,” but I just do):
Rosé wine is called rosé because of its pinkish color — the word “rosé” is French for “pink.” The name refers to the wine’s appearance, which falls between red and white on the color spectrum. It’s made from red (or black) grapes, but unlike red wine, the grape skins are only in contact with the juice for a short period — typically a few hours to a couple of days. This limited skin contact gives rosé its signature blush hue and lighter body. The term became widely adopted due to the influence of French winemaking, especially from regions like Provence, which is known for producing world-class rosés. Today, the name “rosé” is used globally to describe this style of wine, even when made outside of France.
Okay, now we know — these are no traces of actual rose petals in rosé wine.
I mentioned above that I could ask ChatGPT to write an entire article for me. I have played around with this for the sake of experimentation, and it is scary how good it is. The reasons why I don’t are that it feels intellectually dishonest, and, believe it or not, one of the reasons I write these articles is to work my actual brain, and asking an AI tool to write it for me eliminates that benefit to myself. So I have resisted the temptation to call in the bots to write or at least finish off an article for me, even though, as mentioned, it is uncanny how effectively ChatGPT can adopt my own voice and round out an unfinished piece into something more than workable if not a completely finalized version.
That is my one big takeaway from using ChatGPT: it is so good. So good. Scary good. The writing and analysis are often better than I can do myself, and I fashion myself as a pretty good writer. I suspect that a fairly large amount of online written content is already AI-produced, and over the next few years, even more will be. On the one hand, some of this is mindless drivel, awkwardly written. But a lot of it is good, and the AI technology is only likely to get better with tone and nuance. The future for writers, thinkers, and journalists of all types is, in my opinion, somewhat up in the air at the moment. I don’t write The Sunday Morning Post as my primary source of income, but if I were a paid writer as my main gig, I would be worried. We are not too far off (we may be there now) from readers being unable to distinguish between AI-produced content and human output. And the particularly troubling thing is that the AI-content, in many cases, is better than what a human can do.
How to Learn ChatGPT
There are plenty of resources out there to teach people how to use AI. I mentioned above that I really only use ChatGPT, and I keep telling myself I should probably learn some other programs. I am particularly interested in the creation of graphics and design work, although I also have some feeling of obligation or even guilt that I should be using actual human beings for this type of work instead of programs that have been “trained” on human content. There are some ethical questions at play, in my opinion, about using AI-created content, when that content has been typically developed by scraping (or “learning”) from reams of human-produced content out there. There is an analogy here about asking an office worker to train their replacement.
My main recommendation to the layperson who is not that familiar with ChatGPT or AI technology in general, however, is to simply dive in. Create an account through ChatGPT and start asking it questions. Almost any questions will do. Ask about recommendations on a new car, or ideas for places to visit when you are in Massachusetts next month. Ask about recipes, or how many teaspoons are in a tablespoon. A couple of weeks ago, I told ChatGPT where I was staying outside of Washington D.C., and asked about the easiest way to get to Capitol Hill. Instantaneously, she had an entire trip agenda planned for me with recommendations on how to get there the most efficiently and effectively. I’ve asked ChatGPT while standing inside of a Target how to interpret youth t-shirt sizes. I have asked health questions. I have asked for podcast recommendations. I have asked for workout ideas for myself, but also examples of running exercises for my up-and-coming running children. I could go on and on.
I am not saying everyone reading this needs to become a ChatGPT pro, but AI technology is here to stay. It will only be more ubiquitous with each passing year (and, at the moment, each passing month). We are living through a seismic shift in the way humans operate that, I think, will eventually lead to questions about what it even means to be human. Yes, we are still a ways away from that, but this is happening fast.
By the way, upon completing my final-ish draft of this week’s piece, as mentioned above, I put the entire article into ChatGPT to proofread. Here is what she told me:
Your piece is thoughtful, engaging, and reflective of your unique voice—it reads very naturally and conversationally. Below is a proofread and lightly fact-checked version, preserving your tone while correcting grammar, clarity, and factual precision. Suggested changes are explained afterward.
I am not writing out the entire re-draft here, but hopefully in what you read above, you didn’t identity any typos, grammatical mistakes, or other errors.
I’ll come back to the AI topic, broadly defined, in the weeks ahead. Thanks for reading today my, almost entirely, hand-crafted column.
Ben Sprague lives and works in Bangor, Maine as a Senior V.P./Commercial Lending Officer for Damariscotta-based First National Bank. He previously worked as an investment advisor and graduated from Harvard University in 2006. Ben can be reached at ben.sprague@thefirst.com or bsprague1@gmail.com.
The Sunday Morning Post is always free, never pay-walled, and contains no ads. To learn how you can support this work and keep these articles free from clutter, become a paid supporter here or click to read more here.
Addendum
In the interest of transparency and for entertainment and further context on today’s article, below is the cut and paste of the actual ChatGPT proofreading of today’s piece. The corrections below, as flagged by ChatGPT, were all implemented in the draft you read above. This what it the proofreading looks like following the prompt I inputted, “Please proofread and fact-check for grammar and typos.” By the way, if I don’t specify that I only want her to proofread for grammar and typos, she proofreads in overly complex ways that include moving content around and adjusting the overall flow of the piece. Sometimes that is too much, which is why I prompt the program to just proofread specifically for grammar and typos as I don’t want big sweeping changes at the 11th hour. These were the typos she found in my almost-final draft:
✅ Identified Issues:
1. Incorrect verb tense
“I haven’t wrote in detail about AI yet.”
✅ Correct: “I haven’t written in detail…”
2. Article/subject-verb agreement
“Streaming services like Netflix and Spotify personalize recommendations based on what its AI algorithms…”
✅ Correct: “…based on what their AI algorithms…” ("services" = plural)
3. Redundant or awkward phrasing
“don’t get require salary and benefits…”
✅ Correct: “…don’t require salary and benefits…”“There are just simple examples.”
✅ Suggested: “These are just simple examples.” (more natural)“An entire article Sunday Morning Post article…”
✅ Correct: “An entire Sunday Morning Post article…”
4. Spelling
“ubiqutious”
✅ Correct: “ubiquitous”
5. Punctuation/usage
“the program finds data, references, and other insights from around the web far quicker…”
✅ Style tip: “more quickly” is more formal and grammatically correct than “far quicker,” though the latter is acceptable in informal writing.
6. Consistency/clarity
“...don’t get require salary and benefits...”
Appears to be a typo mashup between "don't get" and "don't require"
✅ Correct: "...don’t require salary and benefits..."
Have a great week, everybody!