ChatGPT, an OpenAI chatbot that has taken the world by storm over the past few months, can generate fake text, raising concerns among teachers in education circles that students might misuse it for homework. In order to detect AI-generated text, Stanford researchers introduced DetectGPT, which is based on the specific tendency of large language models (large language models, LLM) to generate text, so that it can determine whether the text is generated by AI. The researchers call their method zero-shot and plan to release their code and data in the future.
This article is transferred from: https://www.solidot.org/story?sid=73991
This site is only for collection, and the copyright belongs to the original author.