There have been quite a few posts in our WordPress community recently about ChatGPT, artificial intelligence chatbots, and whether or not such chatbots have the potential, at some point, to be good enough to take the place of us human bloggers.
I was curious about the AI-generated content, ChatGPT, and chatbots, and whether they are all that they are cracked up to be. So I downloaded an app called Genie to my iPhone. Genie is an OpenAl ChatGPT powered by GPT-3. I opened up the app and typed “Write a story with the words premonition, trance, vinyl, scud, sock, and trendy.” Here’s the post that the chatbot generated.
I also posed this question, “What would happen if Al surpassed human intelligence?” on this week’s Provocative Question prompt, which generated some interesting responses. One such response was this one from blogger Kajmeister. It’s long, but informative, interesting, and if you are at all intrigued about AI chatbots, well worth the read.
One interesting experiment Kajmeister conducted was to run a few paragraphs of my entirely chatbot-generated post through an AI detector model to see if it picked up on the AI-generated content. And the results were interesting.
bullshit AI detector determined that the probability of those paragraphs having been written by a chatbot were 99.98%. Excellent detecting work, AI detector.
So, I decided to take two recent posts of mine that I actually wrote without the assistance of a chatbot, and run them through this AI detector. Here are the results below.
99.98% “real” on the first sample, but only 89.19% “real” on the second sample. Maybe my post comparing Dunkin’ Donuts to Krispy Kreme doughnuts was a bit more robotically written. Or maybe the AI detector was pissed that I prefer Krispy Kreme to Dunkin’ Donuts.
Next on my list of things to waste time doing is to write a post and try to see if I can fool the AI detector into thinking it was written by a chatbot. Yeah, that’s the ticket!