@liquiditygoblin
In an effort to try stop seeing so much slop I've been trying to train my own AI detection model. Found something incredibly interesting. for the most part LLM generated text and human written text are linearly separable. https://t.co/6uWEQxVi6O