I Helped Invent Generative AI, and I Know When You're Using ChatGPT

Look, here's the deal. I was part of a team that invented and released the first commercially available automated content/natural language generation/generative AI platform. Back in 2010. I have a patent. I'm incredibly proud of it. And I also apologize. Because I predicted this would happen. But more importantly, the reason I bring it up is that once we got the technology working and lined up customers like Yahoo Fantasy Football and the Associated Press, I spent most of my time, for a long time, working on algorithms and models and code to make the content sound less like it was generated by a machine. I think we did a great job with this. Like, we had decent individualized jokes when the customer let us write decent individualized jokes. You have no idea how hard that is. I'm not falling for the banana in the mainframe. But having done it -- man, it pains me to tell you this. I know when you're using ChatGPT. And what's more, it's not hard to figure out. And in a lot of cases, it's kinda making people look foolish. Emails and Messages If you send me a personal email written with ChatGPT, we're not friends anymore. So I'll just move on to business emails. Hey, sales guy. Sending me a ChatGPT-generated email is the equivalent of sending me an email that starts with "Hi there, FIRST NAME LAST NAME !!!" Oops. But it's even worse than the templated garbage you normally send because it forces me to wade into a lot of peripheral nonsense that whomever you paid too much money to code the sales email generating software thinks might get everyone to buy that same ChatGPT sales email generating software. I was just deleting your spam when it snuck through my filter. Now I'm mad at you. Context, sales guy. Context is everything on the way to close. Did no one teach you this? And all you're doing by generating fake, unrelated context is wasting my time. Time kills deals. That one I'm sure you know. Reviews and Comments OK, let me turn off the anger, because I'm only angry on your behalf. I'm faking it. It's a little writing trick I use. Speaking of writing, when people use ChatGPT or some equivalent to comment on my columns or posts, I see it in a second and I totally think it's 100-percent OK. I would never, ever push back against someone or criticize someone who takes the time to comment on one of my columns -- positive, negative, or machine-written. Because something I did made you take the time, and it's the time that is most important to me -- taking the time to read what I wrote, taking the time to respond. You're doing that for me, and I am grateful. OK. Anger back on. Let's talk about reviews. If you're getting paid or otherwise sponsored to use ChatGPT to write reviews of products and services, you are doing something unethical. Stop it. I wrote about this in a column a while back when Sports Illustrated got caught using bots and avatars to create content. Having started my own automated content journey in sports, I saw no problem with this. Especially when we did something like recaps for Little League games. We made heroes out of kids and kept traveling moms and dads in the loop. But what I did see as a problem -- and what everyone missed, and why the hammer rightfully came down -- was that SI was using bots and avatars to write sponsored product reviews. I don't care whether the reviewer is using ChatGPT as a helper or is just straight up a bot, which obviously would never have had any contact with the product or service. If the reviewer can't conjure the words that describe the emotions and utility involved with actually using the product, they are not a reviewer. They are writing advertising copy, poorly, and they are lying to you. Articles and Posts No, ChatGPT, you can GTFOH. Words mean things. That's why we invented them. And if someone strings together just a few of them that don't mean anything, it's very easy to detect, and it's very obvious that someone's content is being written by a machine. I spent days, weeks, and months trying to hide word salad in our automated content platform. I got maybe 60-percent of the way there. Maybe. Résumés and Cover Letters I can't honestly call this a bad idea, because the whole hiring landscape is a quagmire of minefield and quicksand right now. But again, I can tell when someone has used ChatGPT to put together a résumé, and so can recruiters and HR people. What's more, it's antithetical to the purpose of both the résumé and the cover letter, since both are meant to show you took the time to align your skills and desires with the requirements and opportunities presented in the job requisition. Oh, and if you're using ChatGPT to write the job requisition itself, again, get all the way out of here. Personal Connection Matters One of the things I constantly hammered home to my team was to never forget the personal connection that is the purpose of any kind of content, including that written by a machine. Don't create content for content's sake, no matter how targeted or individualized or personalized it may be. I'd rather see a single sentence that means something rather than a bunch of paragraphs that don't. There is a time and place for generative AI. Our company was called Automated Insights, because we were automating insights, not words. Words were secondary to what we were doing. Those words were always meant to be read by people so that they could more easily understand the data behind them and make their own decisions. Today's generative AI isn't doing that. Not well anyway. It's being sold as a replacement for the personal connection that matters in many forms of communication. And when we forget why we invented words, well, we deserve all the word salad we can eat. Expert Opinion By Joe Procopio, Founder, TeachingStartup.com @jproco