Recently, a research group at Stanford University has proposed a new method called DetectGPT, which aims to be one of the first tools to combat machine-generated text in higher education. The method is based on the principle that text generated by LLM typically hovers in a specific region of the negative curvature region of the model’s log probability function. With this finding, the team developed a new metric for determining whether text is machine-generated and does not require training artificial intelligence or collecting large datasets to compare texts.

This approach, called “zero-shot” learning, allows DetectGPT to detect machine-written text without the need to know what artificial intelligence tools are used to generate the text. It operates in contrast to other methods that require training “classifiers” and real and generated paragraph datasets.

The team tested DetectGPT on a dataset of fake news articles, and it outperformed other zero-times learning methods in detecting machine-generated text. The team claims substantial improvements in detection performance and suggests that DetectGPT may be a promising method for scrutinizing machine-generated text.

In summary, DetectGPT is a new method for detecting machine-generated text that takes advantage of the unique features of LLM-generated text. It is a zero-learning method that does not require any additional data or training, making it an efficient and effective tool for identifying machine-generated text. As the use of LLM continues to grow, the importance of a corresponding system for detecting machine-generated text will become increasingly critical. detectGPT is a promising approach that could have a significant impact in many fields, and its further development could be beneficial to many fields.

The post In response to students generating papers via ChatGPT, Stanford University launches DetectGPT appeared first on TechGoing.