Comparative Analysis of Large Language Models in Source Code Analysis

This article is a summary of a study focusing on Artificial Intelligence (AI) based source code analysis amidst the complexity of software development and rapidly evolving technological needs. The study evaluates analyses conducted to improve code quality, detect errors, and perform code optimization by examining the potential impacts of AI in software development processes. The time spent on research and experiments for detecting and resolving errors in the software development process has been a constant source of concern. In this context, the results of using unoptimized source code often lead to outputs that directly affect complex and maintenance costs. The topic has been extensively addressed in the literature as a comprehensive subject known as AI, Code Intelligence (CI), and Programming Language Processing (PLP) and has been the focus of various surveys and application studies. The article suggests that the use of AI could be a potential solution to increase efficiency and minimize errors in software development processes. In the study, two different AI tools, namely ChatGPT and Gemini, were used to address problem resolution. Two different models, GPT4 and Gemini, were included in the analysis process. JavaScript was the preferred language for obtaining source code, which was sourced from the GitHub platform.