[Paper Review] A Systematic Review of Green AI

Roberto Verdecchia, June Sallou+1 more|arXiv (Cornell University)|Jan 26, 2023

Green IT and Sustainability35 citations

TL;DR

This paper conducts a systematic literature review of Green AI, analyzing 98 primary studies to map trends, methods, and maturity in the field.

ABSTRACT

With the ever-growing adoption of AI-based systems, the carbon footprint of AI is no longer negligible. AI researchers and practitioners are therefore urged to hold themselves accountable for the carbon emissions of the AI models they design and use. This led in recent years to the appearance of researches tackling AI environmental sustainability, a field referred to as Green AI. Despite the rapid growth of interest in the topic, a comprehensive overview of Green AI research is to date still missing. To address this gap, in this paper, we present a systematic review of the Green AI literature. From the analysis of 98 primary studies, different patterns emerge. The topic experienced a considerable growth from 2020 onward. Most studies consider monitoring AI model footprint, tuning hyperparameters to improve model sustainability, or benchmarking models. A mix of position papers, observational studies, and solution papers are present. Most papers focus on the training phase, are algorithm-agnostic or study neural networks, and use image data. Laboratory experiments are the most common research strategy. Reported Green AI energy savings go up to 115%, with savings over 50% being rather common. Industrial parties are involved in Green AI studies, albeit most target academic readers. Green AI tool provisioning is scarce. As a conclusion, the Green AI research field results to have reached a considerable level of maturity. Therefore, from this review emerges that the time is suitable to adopt other Green AI research strategies, and port the numerous promising academic results to industrial practice.

Motivation & Objective

• Understand the characteristics and maturity of Green AI research.
• Identify main topics, artifacts, and phases studied in Green AI.
• Assess methodology, industry involvement, and tool provisioning in Green AI research.

Proposed method

• Automated search across Google Scholar, Scopus, and Web of Science using a targeted title query.
• Manual application of predefined inclusion/exclusion criteria to select primary studies (I1-I4; E1-E6).
• Bidirectional snowballing (backward and forward) to reach theoretical saturation.
• Two-phase data extraction to build a structured framework (definition, study type, topic, domain, data type, artifact, phase, strategy, dataset size, energy savings, industry involvement, intended reader, tool availability).
• Data synthesis using constant/open coding to harmonize extracted concepts.

Experimental results

Research questions

• RQ1 What are the characteristics of Green AI state-of-the-art research? (Overall landscape and definitions.)
• RQ2 What publication trends, topics, domains, and artifacts characterize Green AI literature?
• RQ3 What is the level of industry involvement and practical tool provisioning in Green AI studies?
• RQ4 What data types, algorithms, and dataset sizes are commonly studied in Green AI?

Key findings

• The review covers 98 primary Green AI studies, with the field showing substantial growth since 2020 (76% of papers published since 2020).
• Energy savings are reported up to 115%, with savings over 50% being relatively common.
• Most studies focus on the training phase and on neural networks, with image data being the most used data type.
• Industrial involvement exists but is limited (about 23%), and Green AI tool provisioning is scarce.
• Monitoring, hyperparameter tuning, model benchmarking, and deployment are the dominant topics, with data-centric and emissions-related topics underrepresented.
• The field shows a mature state suitable for porting academic results to industrial practice.