Comparing Proficiency of ChatGPT and Bard in Software Development

Generative language-based artificial intelligence models, like ChatGPT and Bard, have gained popularity for their impressive performance across various domains. These models hold the potential for contributing to software development tasks. This chapter presents a comparative study on the proficiency of ChatGPT and Bard in software development, examining their strengths and weaknesses and synergy in their combined use. To facilitate the comparison, we utilize an online tour reservation system (TORS) as a running example. The study found that ChatGPT and Bard have different focuses in requirement analysis and ChatGPT outperforms in design modeling, and they are comparable in implementation and error fixing. However, both models suffer from many inconsistencies particularly between design and implementation, which raises traceability concerns between phases. In both models, human involvement was required to address inconsistency issues and finalize the development.