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

CTOnews.com, August 8 (Xinhua) according to a study by Purdue University in the United States, ChatGPT, an artificial intelligence chat robot developed by OpenAI, has an error rate of more than half in answering software programming questions and can fool 1/3 of the questioners.

The team analyzed ChatGPT's answers to questions on 517 Stack Overflow sites and assessed their correctness, consistency, comprehensiveness and simplicity. They also conducted a linguistic and emotional analysis of the responses and conducted a questionnaire survey of 12 volunteer participants. The study concluded that although the language style of ChatGPT's answer is good, 52% of the answers are wrong and 77% are lengthy. Participants can only be identified if the error in the answer is obvious. Otherwise, they will be misled by ChatGPT's friendly, authoritative and meticulous language style.

Stack Overflow is a website that provides question and answer services for programmers and is considered to be an important source for developers to get code snippets and solve problems. However, the site has also faced problems in recent years, such as declining traffic, declining content quality and declining user trust. According to a report by SimilarWeb in April, monthly traffic on the site has fallen by 6% since January 2022, down 13.9% in March, and ChatGPT may be one of the reasons for the decline. Stack Overflow denied SimilarWeb's assessment and said its traffic had been affected by a surge in search algorithms and ChatGPT interest.

OpenAI admitted on its ChatGPT website that its software "may generate inaccurate information about people, places, or facts." As of CTOnews.com 's release, the company has yet to respond to Purdue University's research.

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.