Detecting LLM-Generated Texts with “Classical” Machine Learning