Unraveling the Hidden Environmental Impacts of AI Solutions for Environment

Anne-Laure Ligozat,Julien Lefèvre,Aurélie Bugeau,Jacques Combaz
2022-04-22
Abstract:In the past ten years, artificial intelligence has encountered such dramatic progress that it is now seen as a tool of choice to solve environmental issues and in the first place greenhouse gas emissions (GHG). At the same time the deep learning community began to realize that training models with more and more parameters requires a lot of energy and as a consequence GHG emissions. To our knowledge, questioning the complete net environmental impacts of AI solutions for the environment (AI for Green), and not only GHG, has never been addressed directly. In this article, we propose to study the possible negative impacts of AI for Green. First, we review the different types of AI impacts, then we present the different methodologies used to assess those impacts, and show how to apply life cycle assessment to AI services. Finally, we discuss how to assess the environmental usefulness of a general AI service, and point out the limitations of existing work in AI for Green.
Artificial Intelligence,Computers and Society
What problem does this paper attempt to address?
This paper primarily explores the hidden environmental impacts of artificial intelligence (AI) solutions in environmental protection and proposes a method to assess these impacts. ### Research Background In recent years, with the advancement of AI technology, AI has been increasingly viewed as an important tool for addressing environmental issues, particularly in reducing greenhouse gas emissions. However, the deep learning community has gradually realized that training models with an increasing number of parameters requires a significant amount of energy, which in turn generates greenhouse gas emissions. Despite this, current research mainly focuses on the positive environmental effects of AI solutions, while neglecting the potential negative impacts. ### Paper Objectives The goal of the paper is to comprehensively assess the net environmental impact of AI solutions used for environmental protection (AI for Green). Specifically, the authors aim to: 1. **Review the types of AI impacts**: Including various direct and indirect environmental impacts that AI solutions may generate. 2. **Introduce assessment methods**: The paper proposes Life Cycle Assessment (LCA) as a comprehensive method to evaluate the direct environmental impacts of AI services. 3. **Evaluate environmental value**: Discuss how to assess the environmental value of an AI service designed for environmental purposes. 4. **Identify limitations**: Point out the limitations in existing work and emphasize that evaluating AI services solely based on technical potential may not adequately reflect real-world scenarios. ### Main Contributions 1. **Review of related work**: The authors review existing work on assessing the environmental impacts of AI and its limitations. 2. **Application of Life Cycle Assessment**: Introduce how to use Life Cycle Assessment to comprehensively evaluate the direct environmental impacts of AI services. 3. **Environmental value assessment**: Explore how to assess the environmental value of AI services specifically designed for environmental protection. 4. **Discussion of limitations**: Highlight that even with improvements to existing methods, the proposed assessment method can only demonstrate the technical potential of the service, which may not be fully realized in real life. In summary, this paper attempts to fill the gap in the AI for Green field regarding the comprehensive assessment of the environmental impacts of AI solutions, with the aim of promoting more sustainable AI applications.