Comparative Study of Code Generation by AI Models

Artificial Intelligence has become integral part of modern software development as different AI assistants help write code, debug them and allow quick push to production. AI-based Code generation tools like: GitHub Copilot, OpenAI CodeX, ChatGPT, Codenine, etc are on the rise. However, these different AI models have various limitations such as high false positives and incorrect suggestions, incomplete code, security and vulnerability risks, ethical and license issues, and limited debugging abilities. In this paper, we perform comparative analysis of present AI-based code generation models and highlight their technical limitations based on fair performance metrics.