ChatGPT 52% Inaccurate in Software Engineering Responses, Study 

ChatGPT generator generates almost 50% incorrect responses to most of the questions related to software engineering.

OpenAI’s ChatGPT generator generates almost 50% incorrect responses to most of the questions related to software engineering, raising concerns about its accuracy and reliability, according to a study. 

The study conducted by researchers from Purdue University aims to analyze 517 Stack Overflow questions by using ChatGPT generator and found that over half of the answers (52%) to the programming related prompts were inaccurate. Moreover, a considerable number of the answers (77%) were considered extremely long. 

One key element stood out in this study, is the way users reacted to GPT’s answers. Despite identifying some inaccuracies in the responses that were generated by ChatGPT, users still prefer the answers of the Microsoft-backed chatbot, over those given by Stack Overflow 39.34% of the time.  

This choice is due to ChatGPT providing comprehensive and clear language style, which users found logical and insightful, leading them to believe that the responses were correct. 

Speaking of the Errors Made 

When it came to the errors made by the chatbot, users were only able to identify them when they were obvious. In cases where these responses were not easily verified or were more complex, users failed to notice the inaccuracies and underestimated how serious was this error, affecting their workflow. Surprisingly, at some point, even when the answer had an obvious error, users considered it as correct. 

To address these issues, the study emphasized the necessity for better communication regarding the correctness and accuracy of AI-generated answers. Although the most popular language model mentions that it can make mistakes, researchers believe that this notification is not enough, recommending the inclusion of a disclaimer with the answers, specifying the level of incorrectness and uncertainty. 

It also stressed that it is important to critically assess AI-generated content, especially in fields requiring high precision such as software engineering. 

Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.