ChatGPT and Me
I developed a fondness for ChatGPT. Until the day it lied to me.
Artificial Intelligence is all the rage. Whoever would have guessed it? AI? That’s some techie, nerdy thing, right? It’s only when ChatGPT put new meaning in the word Chat. A computer with whom you can converse. Sure, we’ve been doing it with voice assistants like Siri and Alexa, but telling Alexa to turn on the lights when you want a room illuminated and asking ChatGPT to write a poem is startling different.
Of course, as a writer and a technologist, I had to take AI for a spin. I’ve been using smart tools for some time: ProWritingAid and Grammarly for proofing my work before we branded them as AI. Now, I’ve been putting ChatGPT to the test. No, I haven’t asked it to write for me. I’ve been using it as a super Google, where I can be more vague about what I’m looking for.
Our publishing company listed a new book about several men taking a road trip. So, I told ChatGPT, “List books written in the past ten years about men on a road trip.” It produced a valid list of ten titles. I have a character in my latest novel who quotes Shakespeare and had a scene in which he talks about being in a play, so I commanded, “Provide quotes from Shakespeare about a play about to start.” Out came a suitable list to choose from.
I was on a roll.
Then came the time I was establishing characters for a novel I’m beginning. It takes place in Brazil. I asked ChatGPT for Brazilian women’s first names. Out came a list that seemed right. Same is true when I asked for last names. I like to ensure I’m not using a real person’s name, so I googled first and last name combinations. I was frustrated to be unable to find any combination that was not already taken.
So, I finally had the idea of getting to the point. ChatGPT is smart, right? It would be up for the challenge. I asked Chat to provide Brazilian woman’s names that do not appear on Facebook. It responded, “Here are some more fictional Brazilian woman's first and last names that do not appear on Facebook:” and provided a list of ten names. I was happy.
But then I paused.
I cannot say what made me doubt Chat. Perhaps it was what I’ve read about AI systems having ‘hallucinations,’ making stuff up. I checked Facebook. Chat lied to me. Each name was on Facebook.
I know about the huge data source ChatGPT has from a multitude of sources. I doubt it had checked with Facebook. But why did it have to lie?
Perhaps when I asked it “What is the difference between tomato sauce and tomato paste,” its answer could not be trusted?