OpenAI, the creators of ChatGPT, have terminated support for their own AI detector as it consistently failed to produce accurate outcomes.
The detector developed by OpenAI was notorious for its poor performance in detecting AI-written content. From the beginning, the company acknowledged its low accuracy rate of only 26%, and this figure did not see significant improvement during its limited operation.
The announcement page for the AI classifier has been updated to reflect the decision, stating that as of July 20, 2023, the AI classifier is no longer available due to its unreliable performance. OpenAI is actively working to gather feedback and researching more effective provenance techniques for text. Additionally, the company has committed to developing and implementing mechanisms that allow users to discern whether audio or visual content has been generated by AI.
AI detectors, in general, have a history of producing unreliable results. Some AI models, like GPTZero, rely on speculation and patterns rather than concrete data, which can lead to inaccuracies.
The AI industry is exploring the possibility of watermarking text or adding additional data to improve reliability. This approach is similar to adding metadata to an image, such as camera model and location.
Due to issues with AI detectors, there have been cases where students were wrongly identified as cheaters, causing concerns in educational settings.
OpenAI has not disclosed a specific timeline for the reintroduction of a new classifier, but it’s worth noting that there may be a potential decrease in traffic due to students taking breaks for the summer.