Large Language Models (LLMs), such as GPT-4, represent a significant advancement in contemporary Artificial Intelligence (AI), demonstrating remarkable natural language processing, customer service automation, and knowledge representation capabilities. However, these advancements come with substantial energy costs. The training and deployment of LLMs require extensive computational resources, leading to escalating energy consumption and environmental impacts. This paper explores the driving factors behind the high energy demands of LLMs through the lens of the Technology Environment Organization (TEO) framework, assesses their ecological implications, and proposes sustainable strategies for mitigating these challenges. Specifically, we explore algorithmic improvements, hardware innovations, renewable energy adoption, and decentralized approaches to AI training and deployment. Our findings contribute to the literature on sustainable AI and provide actionable insights for industry stakeholders and policymakers.

Machine-learning solutions for pattern classification problems are nowadays widely deployed in society and industry. However, the lack of transparency and accountability of most accurate models often hinders their safe use. Thus, there is a clear need for developing explainable artificial intelligence mechanisms. There exist model-agnostic methods that summarize feature contributions, but their interpretability is limited to predictions made by black-box models. An open challenge is to develop models that have intrinsic interpretability and produce their own explanations, even for classes of models that are traditionally considered black boxes like (recurrent) neural networks. In this article, we propose a long-term cognitive network (LTCN) for interpretable pattern classification of structured data. Our method brings its own mechanism for providing explanations by quantifying the relevance of each feature in the decision process. For supporting the interpretability without affecting the performance, the model incorporates more flexibility through a quasi-nonlinear reasoning rule that allows controlling nonlinearity. Besides, we propose a recurrence-aware decision model that evades the issues posed by the unique fixed point while introducing a deterministic learning algorithm to compute the tunable parameters. The simulations show that our interpretable model obtains competitive results when compared to state-of-the-art white and black-box models.