OpenAI, the creator of ChatGPT, has released an AI Text Classifier tool that may be able to determine whether or not texts were written by AI.
“The AI Text Classifier is a fine-tuned GPT model that predicts how likely it is that a piece of text was generated by AI from a variety of sources, such as ChatGPT,” explains a new OpenAI blog post.
The AI text classifier is designed to “distinguish between human-written text and AI-written text.” However, OpenAI warns in a press release that the classifier is “not fully reliable” and “should not be used as a primary decision-making tool.”
According to OpenAI, it can be useful in determining whether someone is attempting to pass off generated text as human-written text. OpenAI also stated that the model that powers the tool was trained using “pairs of human-written text and AI-written text on the same topic.”
OpenAI also claims that it will “incorrectly but confidently” label human-written text as AI, especially if it differs significantly from anything in the training data. It emphasizes that the classifier is still a “work in progress.”
Furthermore, OpenAI trained the tool primarily on English content written by adults, so false positives are entirely possible.
The tool was released after numerous universities and school districts banned the company’s popular ChatGPT AI chatbot for its ability to complete students’ homework, such as writing book reports and essays, as well as completing programming assignments.
The sources for this piece include an article in BleepingComputer.