Abstract: Artificial Intelligence (AI) techniques with tools have been increasingly popular in software development in recent years. This paper investigates the impact of AI tools using ChatGPT as an AI tool in generating code and compares it with the code generated by humans in an open-source platform on GitHub. Four key quality metrics (cyclomatic complexity, line of code, technical debt, and code smell) are used to evaluate the accuracy factor. Three case studies at different degrees of complexity (easy, medium, and hard) are conducted to evaluate the performance and accuracy of tool-generated code against human-generated code. Findings indicated that the ChatGPT tool can significantly reduce code complexity, provide similar solutions in fewer lines of code, and reduce technical debt and code smell in complex tasks with fewer lines of code. Minor differences were observed in the easiest and medium tasks, with similarities on several metrics. This paper highlights the ability of AI tools to generate code that complies with software quality factors that improve the software development process, particularly in complex scenarios, and improve accuracy and efficiency while reducing human errors.

Keywords: AI Tools, Software Accuracy, Software Quality, ChatGPT, Code Generated, Software Metrics

Pages: 185-193