OpenAI, the startup that created the ChatGPT text generator, launched a tool to identify text generated by artificial intelligence.
OpenAI said in a blog post that 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.
The classifier will label text as very likely, unlikely, unclear if it is possible or likely AI-generated.
The blog post said that the purpose of the AI Text Classifier is to foster conversation about the difference between human-written and AI-generated content. The results may help but should not be the sole piece of evidence when deciding whether a document was generated with AI. ChatGPT, which became popular online last year, is a free AI tool that can generate dialogue based on user prompts, and has gone viral for producing poems, recipes, emails and other text samples. The chat bot has passed graduate-level exams in multiple fields, including the final exam for the Master of Business Administration program at the University of Pennsylvania and exams for four law courses at the University of Minnesota. It was comfortably within the passing range of the medical licensing exam of the United States.
There are concerns among educators about the accessibility and capabilities of ChatGPT. New York City s Department of Education banned ChatGPT from school devices and networks earlier this month, citing concerns about the negative effects of student learning. A spokesman for the department said the tool can provide quick and easy answers to questions, but it doesn't build critical thinking and problem-solving skills. Some schools and colleges have considered amending their honor codes to address the rise of ChatGPT and other text generators.
That has spurred the creation of programs to detect AI-generated writing. Edward Tian, a senior at Princeton University, developed GPTZero late last year in an effort to combat AI plagiarism in academia. Copyleaks launched its own AI Content Detector for educational institutions and publishing earlier this month. The Giant Learning Model Test Room, a collaboration between the Harvard Natural Language Processing Group and the MIT-IBM Watson AI Lab in 2019, identifies AI-generated writing using predictive text.
The classifier of OpenAI has some limitations. The writing samples must be at least 1,000 characters, or about 150 -- 250 words. The tool isn't always accurate - AI generated text can be edited to evade detection tools, and the text classifier may misidentify both AI-generated and human-written samples.
OpenAI also acknowledged that the tool is trained using English text samples written by adults, so it may misidentify content written by children or in languages other than English.
OpenAI said it hasn't thoroughly assessed the effectiveness of the classifier in detecting content written in collaboration with human authors. To train the text classifier model, OpenAI used human-written text from a Wikipedia dataset, a WebText dataset and human demonstrations that were used to train InstructGPT, another language model. The company said it used balanced batches that contain equal proportions of AI-generated and human-written text to train the text classifier.
OpenAi said that the classifier may be very confident in a wrong prediction, because it hasn't been carefully evaluated on principle targets like student essays, chat transcripts or disinformation campaigns.
Because of these limitations, we recommend that the classifier be used only as one factor out of many when used as part of an investigation determining a piece of content's source, OpenAI said.