Hi Kunlun, was this written with AI? (Sorry, couldn't resist the joke!)
So interesting that AI flagged Hemingway and King as AI. I wonder if it's because classic works were probably amongst the first used to train the models on writing, and perhaps the machines have mimicked them so often since that it now all feels AI. Almost like they've forgotten where their source of knowledge came from?
Yes, it has AI score is 7/10. At least I'm beating Hemingway on that front. 😄
Your point about training data is interesting, and I thought about that too. If these works were part of the training set, in theory the model should recognize them. In fact, when I tested it about other well-known Hemingway quotes, it correctly attributed them. For example this one “The best way to find out if you can trust somebody is to trust them.” ChatGPT responses "Score: 1 / 10 Reason: Clean, paradoxical phrasing typical of classic literary aphorisms. Widely attributed to Hemingway—very unlikely AI-generated."
So I don’t think this is about “forgetting the source.” It feels more like a pattern problem. The detector isn’t really recognizing authorship—it’s matching surface features: simplicity, structure, familiarity. And ironically, those are exactly the qualities that make writers like Hemingway and King so powerful. Their clarity starts to look “generic” to a system that sees everything as patterns.
Which is kind of the point I was trying to get at in the piece.
The experiment wasn’t really about testing how good AI is at detecting AI. It was more about how we, as humans, are starting to rely on those judgments. If a machine can misread Hemingway as “AI-like,” it raises a bigger question: what are we actually valuing when we read?
At the end of the day, writing isn’t just pattern—it’s intention, experience, meaning. And those are things detectors can’t really measure.
So maybe the better question isn’t “Was this written by AI?” but “Did this make me think or feel something real?”
Yes, I found your point on what it makes you think and feel very interesting, too! I don't know how to feel about AI when it comes to such things, I think I do value knowing that the thinking and feeling a piece might evoke in me actually came from another human thinking and feeling something as they wrote. That might be personal preference, though, and may well change as we spend more years living alongside AI work!
Thank you for researching, writing, and Substacking. One of the reasons that I am exploring Substack is to try to understand what the so-called, 'digital writing project' is.
Hi Kunlun, was this written with AI? (Sorry, couldn't resist the joke!)
So interesting that AI flagged Hemingway and King as AI. I wonder if it's because classic works were probably amongst the first used to train the models on writing, and perhaps the machines have mimicked them so often since that it now all feels AI. Almost like they've forgotten where their source of knowledge came from?
Hey Charlotte, happy to see your comment.
Yes, it has AI score is 7/10. At least I'm beating Hemingway on that front. 😄
Your point about training data is interesting, and I thought about that too. If these works were part of the training set, in theory the model should recognize them. In fact, when I tested it about other well-known Hemingway quotes, it correctly attributed them. For example this one “The best way to find out if you can trust somebody is to trust them.” ChatGPT responses "Score: 1 / 10 Reason: Clean, paradoxical phrasing typical of classic literary aphorisms. Widely attributed to Hemingway—very unlikely AI-generated."
So I don’t think this is about “forgetting the source.” It feels more like a pattern problem. The detector isn’t really recognizing authorship—it’s matching surface features: simplicity, structure, familiarity. And ironically, those are exactly the qualities that make writers like Hemingway and King so powerful. Their clarity starts to look “generic” to a system that sees everything as patterns.
Which is kind of the point I was trying to get at in the piece.
The experiment wasn’t really about testing how good AI is at detecting AI. It was more about how we, as humans, are starting to rely on those judgments. If a machine can misread Hemingway as “AI-like,” it raises a bigger question: what are we actually valuing when we read?
At the end of the day, writing isn’t just pattern—it’s intention, experience, meaning. And those are things detectors can’t really measure.
So maybe the better question isn’t “Was this written by AI?” but “Did this make me think or feel something real?”
Yes, I found your point on what it makes you think and feel very interesting, too! I don't know how to feel about AI when it comes to such things, I think I do value knowing that the thinking and feeling a piece might evoke in me actually came from another human thinking and feeling something as they wrote. That might be personal preference, though, and may well change as we spend more years living alongside AI work!
Thank you for researching, writing, and Substacking. One of the reasons that I am exploring Substack is to try to understand what the so-called, 'digital writing project' is.