US technology company OpenAI, developer of the artificial intelligence chat platform ChatGBT, has revealed the development of a program that can distinguish between machine-written text using artificial intelligence and human text provided by a real person, in response to growing concerns that inaccurate machine-written text is flooding the Internet. without being distinguishable from texts written by normal users.
According to Al-Bayan website, it is mentioned that GBT chat can imitate human speech in a new way, which raises concerns about many things, including the possibility of using it to cheat during school tests, preparing university research papers, or fabricating misleading information and publishing it on a large scale. .
Despite its best efforts, OpenAP admitted in an online message that its new tool still performs poorly.
GBT Chat said that the new classification and review tool is not completely reliable, noting that it is conducting tests and the tool was able to correctly identify the nature of the text in 26% of cases.
At the same time, the new program said about texts written by users that they are automated texts in 9% of cases, so it recommends at the present time not to rely on the program mainly to evaluate texts.
The company added that it will continue to develop and update this program and hopes to provide improved methods in the future.
OpenAI said the need for software to recognize the use of AI in text writing is clear given the potential to use AI to automate disinformation campaigns, use AI tools for school cheating, and use AI conversational techniques as a human.