OpenAI’s detection service cannot distinguish AI-generated work and has been taken down due to accuracy concerns.

OpenAI's detection service cannot distinguish AI-generated work and has been taken down due to accuracy concerns.

OpenAI Removes AI Text Detection Tool Due to Accuracy Concerns

OpenAI, the creator of the popular chatbot ChatGPT, has quietly removed its tool for detecting AI-generated text, citing concerns about its low rate of accuracy. The tool, which was released in January to help educators detect academic cheating, has faced limitations and should not be solely relied upon for decision-making.

The company acknowledged the importance of identifying AI-written text in education but recognized the need for more effective provenance techniques for text. In a recent update, OpenAI stated, “We are working to incorporate feedback and are currently researching more effective provenance techniques for text.”

Educators worldwide are dealing with the reality of generative AI being widely accessible to students. Rebecca Tan, a lecturer at the National University of Singapore, understands that the motivations behind students using AI tools, such as stress and the desire for shortcuts, are complex issues that cannot be solved solely by educators. It highlights a deeper problem in education that needs attention.

However, OpenAI’s tool is not the only one raising concerns about accuracy. Turnitin, a popular plagiarism-checker software used by schools, announced a new tool that supposedly detects similarities to AI writing. While it has reviewed over 65 million papers since its launch in April, a Washington Post investigation revealed that the software incorrectly identified more than half of the test cases used.

This inaccurate identification has led to several incidents where students were falsely accused of using AI to cheat by Turnitin. Outlets such as Rolling Stone, Times Higher Education, and USA Today have reported separate incidents of these false accusations.

Experts have previously highlighted that generative AI, like ChatGPT, exacerbates underlying issues in education. The overemphasis on tests and formulaic essays is a concern that goes beyond the use of AI to cheat in schools. Leah Henrickson, a lecturer at the University of Leeds, emphasizes that the real way to learn is through problem-solving, which no artificial intelligence can replace.

In response to Insider’s request for comment, OpenAI did not provide a response outside regular business hours. However, it is evident that the concerns surrounding AI-generated text and its impact on education require further exploration and consideration moving forward.