As information technology usage escalates, so does its energy consumption and carbon footprint, primarily driven by data centers. This trend is expected to intensify with the increasing reliance on big data, digitization, and large AI models. Addressing these environmental concerns, our study explores the concept of green coding, focusing on the sustainability of code generated by AI language models and tools. We introduce a "green capacity" metric for code, assessing its sustainability from development to deployment. This metric is applied to compare human-generated code with code produced by the AI models across various complexity levels. Additionally, we extend our investigation to a practical case study involving the upgrade of a cloud-based web application for an automotive OEM, managed by a developer using the AI-assisted coding tools. Our methodology encompasses the entire lifecycle of the code, including the embodied carbon during its generation. The findings reveal that AI-assisted code generation tools hold the potential to reduce the environmental impact of software development. This research provides empirical evidence supporting the integration of automated AI tools in software development, marking a step forward in achieving greener and more sustainable coding practices.
Session:
Software Defined Vehicle I
|
| 14:40 - 15:10