 |
| OpenAI launches AI text recognition tool |
On Tuesday, OpenAI announced a new tool for categorizing and recognizing text written by AI tools such as ChatGPT.
AI Technologies explained in a blog post, "AI Text Classifier is a finely tuned (GPT) model that predicts the probability that a piece of text will be generated by AI from various sources such as: (ChatGPT)."
The launch of the new OpenAI tool comes after several universities and schools banned the company's popular chatbot (ChatGPT) due to its ability to perform student tasks such as writing book reviews and articles, and even doing programming tasks.
According to reports from BusinessInsider, ChatGPT has been banned in schools in US cities such as New York, Seattle, Los Angeles, and Baltimore, and universities in France and India have also banned access to the tool from school computers.
While teachers can use OpenAI's new AI text grading tool to check if students are cheating, the company warns that the tool should not be used as "the only evidence" of academic dishonesty.
"Our classifier is not completely reliable," the company emphasized in its contribution. She added, “In the English-language text challenge group’s assessment, the classifier correctly identified 26% of the AI-written text as ‘probably AI-written’, while misclassifying 9% of the human-written texts written for AI.”
"The reliability of our classifier increases with the length of the input text," OpenAI said. The classifier's ability can improve over time, especially as it is trained on more data.