Towards a Quantitative Evaluation of the Relationship between Performance and Environmental Sustainability of Artificial Intelligence Algorithms
This work addresses the relationship between the performance and environmental sustainability of artificial intelligence (AI) algorithms. Although it is widely recognized that the adoption of AI technology is fundamental in various fields, ranging from healthcare to industry and entertainment, a quantitative assessment on an operational scale of the environmental impact of training and validating AI algorithms is still an open issue. In order to address this aspect, in this work, the first steps towards a metrology-based analysis are investigated with a two-fold aim: (i) to outline a methodology for evaluating AI algorithms also considering the consequent greenhouse gas emissions, and (ii) to better understand how to continue improving their classification performance in a non-harmful way for the environment.
Paper
Full text
Towards a Quantitative Evaluation of the Relationship between Performance and Environmental Sustainability of Artificial Intelligence Algorithms
Semantic Scholar · Environmental Science · 2024
Abstract
This work addresses the relationship between the performance and environmental sustainability of artificial intelligence (AI) algorithms. Although it is widely recognized that the adoption of AI technology is fundamental in various fields, ranging from healthcare to industry and entertainment, a quantitative assessment on an operational scale of the environmental impact of training and validating AI algorithms is still an open issue. In order to address this aspect, in this work, the first steps towards a metrology-based analysis are investigated with a two-fold aim: (i) to outline a methodology for evaluating AI algorithms also considering the consequent greenhouse gas emissions, and (ii) to better understand how to continue improving their classification performance in a non-harmful way for the environment.