College professors are expressing concern as OpenAI discontinues its AI plagiarism detection tool, AI Classifier. The decision was quietly made, with OpenAI simply taking down the landing page for AI Classifier and adding a brief update to their January blog post announcing the discontinuation. OpenAI cited the tool’s low accuracy rate as the reason behind this move. With only a 26% correct identification rate and 9% false positives, it is not surprising that educators are worried about the drawbacks of unreliable AI plagiarism detectors.
The education sector is particularly affected by the discontinuation of AI Classifier. Students have become increasingly reliant on generative AI, such as OpenAI’s ChatGPT, causing academic institutions to search for ways to combat AI-generated plagiarism. AI detection software was initially seen as a significant tool in this fight, but the discontinuation of AI Classifier raises concerns about the efficacy of such tools.
Marc Watkins, a professor at the University of Mississippi, believes that AI Classifier’s shutdown highlights a larger issue with AI detection software. He points out that the University of Pittsburgh recently stopped using AI detection tool Turnitin due to accuracy concerns. Watkins suggests that AI detection software may not work as effectively as expected.
The loss of faith in AI detection technology is evident, as shown by a recent Twitter poll in which only 15.3% of respondents believed that consistently accurate AI detectors could be created. The fact that OpenAI, the company behind ChatGPT, struggles to discern its own AI-generated text raises doubt about the reliability of other detection tools.
Inaccurate AI detection has serious consequences, extending beyond missing a few AI-generated papers. A UC student experienced this firsthand when an AI plagiarism detector wrongfully accused them of submitting an essay written without any outside influence. The tendency for false positives is not random, as non-English speaking students are more likely to be flagged due to the framework used by AI detectors.
While the discontinuation of AI Classifier may not have a significant tangible impact, it does highlight the overall inadequacy of current technological measures against AI-generated content.
Watkins advises against using any existing detection tools and OpenAI’s retreat from classification software further complicates the situation.
It is essential to provide balanced coverage of the topic and consider multiple viewpoints. Despite the discontinuation of AI Classifier, the education sector’s concerns about AI-generated plagiarism remain unresolved. The search for effective AI detection solutions continues, as educators strive to navigate the challenges posed by AI-generated content.
