OpenAI writes a Twitter thread
Predictions about my future as a human writer — part 1
.
In a matter of weeks it became obvious to me that I was either going to start experimenting with OpenAI or be socially banned. You know, black listed in the sense of “What? You haven’t tried GPT yet? Who the f*** do you think you are? And you call yourself a writer…?!”
With my back to the wall I started playing around with the artificial intelligence’s API and tried to get the hang of it. Interestingly enough, I actually had a first revelation regarding the essence of AI some fifteen minutes into the process, but I’ll write about that in part 2 of this modest adventure tale. For now, I’ll focus on the only decent idea I finally had after getting the first introductions between GPT and myself out of the way.
.
A few months ago I wrote about the magic of threads and how Twitter threads change the art of storytelling. If you’ve ever created a Twitter thread you should know it’s not an effortless or straightforward task. If it were easier I would have probably written many of them. So… what if artificial intelligence could produce Twitter threads as easy as it is to heat dried corn kernels until they burst and turn into popcorn? As all writers and content creators I have my work cut out for me. Adding the thread versions of my articles to the mix is nothing but a fantasy. But just imagine how great it would be to turn all my articles, that were and are crafted with hard labor and countless hours of sweat, into Twitter threads by merely copy pasting them to the OpenAI API, thus immediately getting some awesome traffic and engagement-generating threads. By all means this will definitely become a game changer.
So I chose “Yes, but what about the banks?” — a recent article I wrote about regulating crypto, and was ready to let GPT do its wizardry on it. I asked GPT to convert the article into a Twitter thread, and in an effort to make the thread more Crypto Twitter compatible I’ve also asked GPT to make the writing a bit cynical (those were my exact words). It took me a couple of minutes to realize that this was not going to be a simple “Input article, Output thread” affair. No. More like a couple of hours of trial and error, of learning as I go along, of playing with the settings, working with only one or two paragraphs at a time (one cannot input a whole article into the mysterious machine) and so on. Get a result, resent it, delete it, try again, change my instructions (what I’m asking GPT to do), change the settings, hate the result, try again, and so on. And of course, most importantly — get mad at the software. You can’t expect the software to do a decent job if you don’t curse at it like it’s human. And this might come as a surprise, but it very quickly became clear that the AI could care less about my profanities.
.
To be honest, it was a tedious endeavor, although a satisfying one as well. After the fact it was as clear as day that this was a work in progress, both for myself and for this advanced and sophisticated piece of coding. And the result? Well… first of all you can judge for yourself. It might be a good idea to read the original article beforehand, and only then go for the artificially created thread. In short I will just say that for now I could’ve done a much better job writing the thread myself. I’m being gentle of course. I wouldn’t want to offend the AI. But let’s face it, the uniqueness of the article, its storytelling properties, its style of writing — all this has been lost in translation while the text was spending its time in the AI’s black box. One way or another, the instrument will get there and its ability to generate some human-like creativity that’s worthwhile is just a matter of time, I’m certain.
[]
I’d love to read your thoughts on this article. Please share them in the comments.
I would also appreciate it if you consider following me here. Thanks!
