OpenAI, one of the leading artificial intelligence research organizations, and the maker of GPT-4 and ChatGPT has discontinued its AI text-detection tool, known as the AI Classifier. The tool was designed to distinguish between text written by a human and text generated by AI, but it was found to have a low rate of accuracy. The decision to deactivate the tool was quietly made, with the announcement appearing as an update to the original blog post that introduced the tool.
AI Classifier: Promise and Performance
Introduced in January 2023, the AI Classifier promised to aid educators and protect academic integrity by detecting content created using generative AI, such as OpenAI's ChatGPT. However, after six months of operation, the tool was shut down due to its disappointing performance. OpenAI revealed that the tool suffered from significant accuracy issues, rendering it unreliable in identifying AI-generated content. The tool correctly identified only 26% of AI-written text as “likely AI-written”, while it incorrectly labeled human-written text as AI-written 9% of the time.
Limitations and Future Plans
The AI Classifier struggled with shorter texts containing fewer than 1,000 characters and often misclassified human-written content as AI-generated. Classifiers based on neural networks performed poorly when dealing with data outside their training set. Despite these limitations, OpenAI remains committed to enhancing its tools and incorporating feedback. The organization is currently researching more effective techniques for text provenance and is determined to develop mechanisms that allow users to discern whether audio or visual content is AI-generated.
OpenAI stated, “We are working to incorporate feedback and are currently researching more effective provenance techniques for text, and have made a commitment to develop and deploy mechanisms that enable users to understand if audio or visual content is AI-generated.”
Impact on the Education Sector
The education sector had shown keen interest in an accurate AI detection tool, especially since the launch of ChatGPT. Educators had expressed concerns about students potentially using the AI chatbot to produce essays and academic papers. OpenAI acknowledged these concerns and stated that identifying AI-written text has been an important point of discussion among educators. While OpenAI removed it´s AI Classifier, there are other companies that promise to provide accurate AI detection, such as originality.ai which brand themselves as the “Most Accurate AI Content Detector & Plagiarism Checker”. Given OpenAI´s own top-notch AI expertise and technical resources at hand, such tools might also suffer from similar or even worse problems as the for now discontinued AI Classifier.