Like many others, earlier this year I signed up for an OpenAI account to take the much-ballyhooed ChatGPT for a few spins around the querying block.
My first request on that January 19th online excursion was for a “dad joke,” which it swiftly obliged with this little number:
Q: Why did the tomato turn red?
A: Because it saw the salad dressing.
I chuckled.
I texted the Q & A to family and a few friends, none of whom knew how I’d come upon the quip. One pal’s reply was telling: “Now THAT’s a dad joke.”
OK, so it seems like a legit dad-joke generator, I thought. Because ChatGPT had been trained on a dataset of 300 billion words, I would hope so.
A few moments later, wearing my hat as senior editor of a trade magazine, I asked for a 500-word Q & A feature story on a topic relevant to that industry. To be abundantly clear: I would never employ what it churned out. I was simply looking to see what it would generate, as a learning experience.
In this case, ChatGPT produced a rather generic “vanilla” overview of the subject matter. The subject is esoteric enough that I wouldn’t be able to catch mistakes or do a great job of judging the overall quality of the answer.
However, I can envision there coming a time when I might use such an AI-conceived piece as a step in my reporting process.
For example, one potential avenue could be prompting a source to view the AI-generated copy as a starting point from which they could add, subtract, revise and otherwise tailor the communication to “make it their own.” In some cases, that could involve an extensive overhaul; in other instances, maybe some nips and tucks.
I have only visited the site a few times since then, just poking around a bit. Nothing too remarkable to report. But then, earlier tonight and eager for the start of the Major League Baseball season, I asked ChatGPT to tell me who led the American League in triples for three particular years between 1972 and 1982. Of course, there are plenty of spots online that would reliably provide that info, such as Baseball Reference. But I figured these were easy, fat pitches right down the middle of the plate. Instead, ChatGPT whiffed on all three soft tosses, swinging more wildly with each inquiry.
First, it erroneously claimed Rod Carew led the league in ’72 when, in fact, he had done so in ’77 for the Minnesota Twins. OK, it mixed up the year. I figured I’d give it a little help by typing in the question for 1977 – so that it would bounce back with Carew’s name there. Instead, it inexplicably spit out Roy White, a New York Yankee who never came close to leading the league in triples.
Next ChatGPT’s credibility declined from bad to much, much worse: I asked who led the league in triples in ’82 – and this time, ChatGPT got it right, with the Kansas City Royals’ Willie Wilson. Only, it didn’t know when to quit. Like a pathological liar who can’t settle for telling a simple truth, the words kept coming, first asserting that Wilson led the league three times (when it was five times) and then throwing in pure fiction – that Wilson led the league in stolen bases during that 1982 season as well. In fact, his 37 stolen bases were sixth-most, and 93 behind the record 130 swiped that year by Oakland A’s speedster Rickey Henderson.
Makes me wonder how close (or far) its response had been to that hypothetical Q & A for my magazine.
Over time, I am sure the technology – and my relationship with it – will evolve. There are significant holistic concerns, on myriad fronts, including the ethical implications of AI, that must be navigated. But for this moment in time, I view ChatGPT (and its competitors in this space) somewhere between “friend” and “foe.” It’s a tool that can be hugely helpful, as well as massively destructive. In that respect, I suppose, it’s much like the countless array of tools that have preceded it, going back to the dawn of civilization.
An `On This Day’ Inside Edge Double-Play:
On this day in 1964, people across the country and world learned about the devastating earthquake that struck Alaska the day before.
On this day in 1999, I was enjoying my second day of self-employment. Today is Day 8,767. Thanks for sharing a piece of it with me.
If you have enjoyed reading this and any other installments of The Inside Edge, please consider clicking on the “like” button, leaving a comment or sharing this with others who might also enjoy it.