Stanford researchers created DetectGPT, a tool to help teachers and others identify content generated by ChatGPT and similar large language models (LLMs).
ChatGPT has been a topic of concern since the introduction of the chatbot. OpenAI, the maker of ChatGPT, after the controversy, confirmed that they are working on a tool to detect ChatGPT-generated content.
What Is DetectGPT And How Does It Work?
DetectGPT is a tool designed to detect ChatGPT content and other similar tools. Researchers found that text generated by ChatGPT LLMs “occupy negative curvature areas of the model’s log probability function.”
The researchers then developed new criteria to determine if a passage was generated using an LLM. This allows DetectGPT to use log probabilities to determine if a particular LLM generated a passage. The researchers claim that DetectGPT has 95% accuracy.
Concerns raised by educational institutions around the globe about students using these tools for homework and papers have led to the need for DetectGPT.
The team behind DetectGPT will soon release more information about this ChatGPT content detection tool.