(NOTE: I promise this is my last rant about AI for this year.)
I’m beginning to feel like those who used to make buggy whips near the end of the 19th century. A few automobiles were starting to appear, but no one seriously thought they would ever replace a good horse or wagon. Noisy and unreliable cars could barely keep up with a good steed and would be left behind at a gallop. They had less pulling-power than a single mule and ran on cleaning fluid you had to buy at the drugstore. The idea that someone would build a network of convenient gasoline stations across the country seemed impractical and certainly unimaginable. No one would have predicted that an entire industry (buggies and whips) would be obsolete by the next decade.
Why the reflection? It has to do with ChatGPT 4.0 and my story that ran in CDC Gaming on December 18 about Email Service Providers. I’ve been a writer since graduating from journalism school in the late 1970s. My process, and that of just about every writer, has been the same for years. First there is research through on-scene reporting, source material (books and news articles) and then interviews with subject-matter experts or pundits. That is followed by the first rough draft.
If deadlines permit, it is best to let a piece brew for a day or two and then do a self-edit. From there, the procedure is to share the draft with any interview subjects to make sure their quotes are accurate (sometimes that can be contentious). Occasionally, that generates some minor changes. Finally, there was a submission to a copy editor who always had plenty of corrections and suggestions.
But things slowly started to change. The first blow came with automated spell checkers that became widespread in the 1990s. They were incredibly helpful in spotting careless typos. We all love them. But at the same time, they were also frustrating and annoying since they often produced more errors than they found. We all hate them.
I was always amused that the earliest versions of WordPerfect would suggest that the word “Harrahs” should be replaced by “whore.” Even today, my computer keeps insisting that I change “steed” in the first paragraph of this story to “stead?”
Next came the internet, which put small newspapers on shaky financial ground. One of the first job categories to be chopped was that of “copy editor”. At one time, you could pretty much count on newspapers to be free of typographical errors. Not today. Most local papers, TV news shows, and internet blurbs are afloat in typos. Only a very few small operations, and most of the largest publications/networks, seem to employ decent copy editors.
Of course, the internet was not all bad. Basic research was greatly simplified once you managed to sort out fact from fiction. And that was helped along by reliable checking services like Snopes.com or vetted Wikipedia citations.
But now, there is ChatGPT (the acronym stands for Generative Pre-trained Transformer – whatever that means). The early versions were amusing. However, the current 4.0 iteration is both exciting and frightening.
As an experiment, I submitted my story on ESPs to the latest Chat and asked it to make improvements. It did, and it didn’t. My draft of 1,061 words was chopped to 604. Most copy editors would say that’s a good thing. If you can say the same thing with fewer words, that’s an improvement. But it also eliminated some of the “feeling” I meant to convey.
It suggested that my headline of “Many Tech Firms Think Casinos Are Despicable” should be changed to “The Unjust Stigma: Tech Firms’ Unwarranted Opposition to Casinos”. Not bad. But to me, headlines are meant to attract attention, as well as preview the story. I stayed with the original and several other notations, despite my tendency to get a bit wordy.
Other edits were just better than what I had written. So, I included a few of them. Chat also wanted to shorten or eliminate many of the quotes. I didn’t follow that recommendation since I felt verbatim quotes add authenticity and provide additional perspective.
The point is that I could have just used the ChatGPT v4.0 story written by those faceless AI machines networked around the country. It took them less than a second to create their work (my version took two+ weeks). And Chat made zero typos (my first draft had over 20).
When I tried a similar experiment two years ago with an earlier version of ChatGPT, it was almost comical. I asked it to write a story about the 1963-64 Studebaker Avanti (I was working on a project for the National Automobile Museum at the time). The story was filled with factual errors. The worst was that it wrote that the car had a rear-mounted engine, and that the Avanti II was also made by Studebaker. Both are major errors that any true automotive fan would spot. And there were lots of other basic literary problems that any decent writer would recognize.
But less than two years later, Version 4.0 wrote a decent story. Better than mine in some ways; maybe not in others. And I believe it would be impossible for anyone to determine which lines in my published story were from ChatGTP or Buddy Frank. And that’s scary.
Those in the casino world who are assigned to write press releases, budget justifications, or daily security incident reports may be inclined to celebrate. Automobiles took only two decades to obsolete horses as a practical mode of transportation. Chat and the other AI tools may do away with writers in the casino world, and everywhere else, in a fraction of that time. The jury is still out about whether that’s good or bad.
