Shulou(Shulou.com)11/24 Report--

CTOnews.com, September 10 (Xinhua) 2023 is the year of artificial intelligence, from chatbot ChatGPT to songs written by AI for Grammy Awards, to Hollywood actors and writers protesting against the excessive use of automation technology by big studios.

CTOnews.com noted that a recent study by the University of South Florida in the United States shows that the text generated by AI may no longer be distinguishable from human writing. The study authors invited linguistics experts from the world's top linguistic journals to participate in the project, but even such professionals find it difficult to recognize what AI wrote. Overall, they can only correctly distinguish 39% of the cases.

Matthew Kessler, a scholar in the Department of World languages at the University of South Florida, and J. Elliot Kassar collaborated to have 72 linguists review a series of abstracts and determine which were written by humans and which were generated by AI. Specifically, each participating expert must examine four writing samples. As a result, none of the experts could correctly identify all four samples, and 13% of the experts got them all wrong.

Based on these results, the authors have to conclude that most modern professors cannot distinguish between content written by students themselves and content generated by AI. The researchers speculate that software may need to be developed in the near future to help professors identify what AI writes.

Linguists try to use some reasons to judge writing samples, such as identifying certain language and stylistic features. However, in the end, these methods basically failed-resulting in an overall correct recognition rate of only 38.9%.

All in all, Kessler and Kassar have concluded that chatbots like ChatGPT can indeed write short articles like most humans (and in some cases better). AI doesn't usually make grammatical mistakes. However, the study authors point out that human beings still have an advantage in long writing.

"for long text, AI has been shown to produce hallucinations and fictional content, making it easier to recognize that it was generated by AI." Kessler concluded.

Kessler hopes that this work will facilitate discussion on the urgent need to establish clear ethical and guiding principles for the use of AI in research and education. The study is published in the journal Applied Linguistics Research methods.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.