CarbonReveal: Embodied Carbon Accounting with Retrieval-Augmented LLM for Computer Systems
Xiaoyang Zhang, Yucheng Bao, Taiqi Zhou, Dan Wang · 2024
Traditional carbon accounting methods, e.g., Life Cycle Assessment (LCA), heavily rely on extensive data collection and expert knowledge, which is labor-intensive and time-consuming. We develop a system named CarbonReveal to achieve automatic embodied carbon accounting for computer systems, and it can be easily extended to carbon accounting in other fields. CarbonReveal leverages retrieval augmented generation to enhance the capabilities of large language models (LLMs) in reliable and cost-effective carbon accounting. Our preliminary results show CarbonReveal has a 93.92% improvement compared to the state-of-the-art.