Abstract
Artificial Intelligence (AI) systems, particularly large language models, have shown remarkable advancements, revolutionising various fields across industries. However, the sustainability of building large AI models with billions of parameters has become a subject of concern due to their significant environmental and social impact. The training of such models consumes enormous amounts of water and energy and emits substantial carbon emissions, contributing to climate change as data centres heavily rely on fossil fuels. This article summarises the current situation and explores the benefits and challenges of large AI models, emphasising the environmental impact and proposing strategies towards sustainability. Special attention is given to the social challenges, including accessibility, job displacement, biases, and data privacy concerns. Finally, the article advocates for the formulation of green and good AI practices standards for the future. To achieve sustainability, regulations are suggested to ensure transparency and accountability while promoting innovation-friendly frameworks. The authors see that while there is more progress in technology and infrastructure to address environmental impacts, social impacts are more neglected, and they are arguing for more detailed regulation as a solution.
