OpenAI has announced the launch of a new AI classifier that can distinguish between AI-written and human-written text. This classifier aims to inform mitigations for false claims that AI-generated text was written by a human, such as running automated misinformation campaigns, using AI tools for academic dishonesty, and positioning an AI chatbot as a human.
Accuracy and Limitations:
Although the classifier is not fully reliable, it has shown promising results in detecting AI-written text. In evaluations on a “challenge set” of English texts, the classifier correctly identifies 26% of AI-written text as “likely AI-written,” while incorrectly labelling the human-written text as AI-written 9% of the time. However, the classifier’s reliability typically improves as the length of the input text increases.
The classifier is not entirely reliable, especially for short texts below 1,000 characters, and it performs significantly worse in other languages and on code. Therefore, it should not be used as a primary decision-making tool but as a complement to other methods of determining the source of a piece of text.
The classifier is based on a language model fine-tuned on a dataset of pairs of human-written text and AI-written text on the same topic. The dataset was collected from a variety of sources believed to be written by humans, such as the pretraining data and human demonstrations on prompts submitted to InstructGPT.
OpenAI is making this classifier publicly available to get feedback on whether imperfect tools like this one are useful. They recognize that identifying AI-written text has been an important point of discussion among educators, and they have developed a primary resource on the use of ChatGPT for educators, outlining some of the uses and associated limitations and considerations.
While this new AI classifier is not fully reliable, it is a step forward in detecting AI-generated text and can inform mitigations for false claims that AI-generated text was written by a human. As this technology continues to develop, it is important to recognize its limitations and use it in conjunction with other methods to determine the source of a piece of text. OpenAI welcomes feedback from educators and other groups as they continue to broaden their outreach and engagement with affected communities.
