Answers are grounded in this paper and cite their sources.
Abstract Generative artificial intelligence (GAI), a subset of artificial intelligence, requires substantial computational hardware resources for data processing and model training. However, the electronic-waste (E-waste) toll of GAI remains underexplored and overlooked. Here, we propose a Computational Power-driven Material Flow Analysis (CP-MFA) model to measure GAI-related E-waste generation, with a specific focus on large language models. By quantifying server requirements and E-waste generation of GAI under different scenarios, we find that this emerging waste stream will grow at a rapid pace (16 million tons cumulative waste by 2030) with deleterious environmental impacts. Accordingly, we call for an early adoption of circular economy measures among server manufacturers and data center operators. This study reveals significant hardware-linked environmental implications in the context of GAI boom.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: