So, we had this big discussion here recently about how well we can spot AI-generated text, even just after a year. But some of you were saying it’s not just about the AI, it’s also about how the person sets up the prompt. They were saying you can actually tweak these language models to make their output seem totally human.
I’m curious, if any of you are using other language models, what techniques have you found that really work to make the text sound more human?
There are a few strategies we can apply to make AI-generated text sound more humanlike. First, we establish the context—similar to telling the AI, “Hey, let’s talk about cats!” Then we introduce some personality. Maybe our AI evolves into a poetic squirrel or a quirky librarian. Imperfections also assist, such as when Copilot hesitates or stammers somewhat. We can also use anecdotes or follow-up questions to keep the conversation continuing. Varying phrase length and using emotional nuances make it more human-like. Remember, AI will not deceive everyone, but the path is intriguing.
You might request that it be less descriptive and more direct.
I find the Claude AI on Poe excels at creative writing, it even managed bypass most GPT detectors. Give it a try.