Raw LLM Responses

Most people can't tell the difference anymore between an article written by an LLM vs. written by a human (this was even a benchmark in the original GPT papers from OpenAI). We're way past the uncanny valley. The problem is that quite obviously some people write like AI and always have way before it was a thing - that text is part of the training data. All these tools that promise to detect AI generated content also barely work better than a coin flip and needlessly accuse students of something they didn't do. Now you also need to think about avoiding a certain style (dashes, bullet points, colorful adjectives) to be not accused of using AI, although there's nothing wrong with that particular style. Other than being overused in the responses of the default style and default initial prompt of ChatGPT.