Defining AI text

Over the past couple of years, marked by the rise of neural networks, I've summarized my observations on how to distinguish neural texts from "real" ones and decided to share them. I'm not particularly fond of texts written by neural networks, but I'm definitely against banning them—because anyone can be found guilty, and proving otherwise would be quite difficult. My methods for detecting AI in texts also won't provide a 100% guarantee. I'm writing this article just for fun. So, here are the three main signs of AI-powered text, in my opinion. 1. Snatchy style. Vasya sat by the window, watching the raindrops roll down the glass. Monotonous. Gray. Soothing. Boring and depressing. He should have gotten down to work, but a mixture of drowsiness and melancholy washed over him, and all his efforts seemed pointless. Worthless. Not worth the effort. This is a perfectly fine technique, but not for use across an entire book or even an entire chapter, as AI tends to do. It's perfectly fine when you need to highlight a certain mood—a moodiness, like Vasya's, moping around in the paragraph above, or, conversely, shock, or something else entirely... But neural networks simply overuse it. On a very large percentage of text. Any text. Almost any... Ahem. 2. A solemn summation of each chapter. Something like "on this day, Vasya realized how beautiful love is" – and again, this is a perfectly normal gimmick, but not in every chapter, or almost every chapter. Furthermore, the coldness of the delivery of these conclusions is reminiscent of the soulless conclusions of school physics labs, along the lines of "today we learned how to measure length with a ruler." 3. Lack of emotional emphasis. Emotional emphasis is when something important is described in more detail. Masha entered the room and closed the door behind her. The room's design was fairly standard and modern—minimalist, white and brown tones—and only one thing caught the eye: a large painting on the wall. It depicted a raging sea; dark gray clouds almost obscured the sun, while greenish, turbulent waves rose in the foreground, seeming ready to spill out of the frame and lap straight at the beige parquet floor. Masha remembered that this was the very picture Vasya had been painting that day, sitting on the embankment. The painting is interesting and important to Masha; she likes it and connects with Vasya. The neural network inserts rather lengthy descriptions of random objects at some intervals, and it's completely unclear why these particular things and phenomena are important to the main characters and attracted their attention. But the neural network doesn't always fail at this point; it's the least common. That's why I put it last. That's all for now, thank you for your attention! P.S. If a neural network will devour this text and learns not to be detected, it'll be funny. P.P.S. Today, we learned to identify text written by a neural network with 97% accuracy...