OpenAI, the artificial intelligence company that developed ChatGPT, is offering a tool for detecting if English language text was written by AI or a human (Editor’s Note: it’s not really that hard for a literate human to make that determination now, but presumably the AI will get better).
The new “classifier” tool was released to the public Tuesday, though OpenAI says the tool is far from infallible and is asking for user feedback. The new tool was first reported by Semafor.
The tool “may help, but should not be the sole piece of evidence” used to determine if any particular piece of text is AI-generated. Humans can edit AI writing to make it sound more human, and less robotic, which can evade the classifier without much problem.
The tool correctly identified 26% of AI-written text as “likely AI-written,” but incorrectly thought human-written text was actually AI-generated 9% of the time. Which might say more about the human writers than the accuracy of the classifier. Educators, students, and parents are invited to submit their feedback online.
ChatGPT, released late last year, and was hailed for its remarkably conversational tone. A flurry of articles have been written, presumably by people, speculating about whether AI chatbots could replace human writing, if lazy students will use it for term papers (thereby appearing smarter than they might actually be), how it might be used for fraud, and whether writers, already generally penurious, will become entirely redundant.
ChatGPT was trained on a massive amounts of text databases and has successfully taken several mock academic exams, and written high-level essays. Some independent developers developed tools for detecting if text was written by ChatGPT.
Even though OpenAI has discontinued its AI detection tool, there’s still a pretty big demand for such tools out there. There are other tools such as this AI Detector, which you can use to check whether the text is AI-generated or not.