Academic Papers related to Artificial Intelligence Resource Use

The growing carbon footprint of artificial intelligence (AI) has been undergoing public scrutiny. Nonetheless, the equally important water (withdrawal and consumption) footprint of AI has largely remained under the radar. For example, training the GPT-3 language model in Microsoft's state-of-the-art U.S. data centers can directly evaporate 700,000 liters of clean freshwater, but such information has been kept a secret. More critically, the global AI demand is projected to account for 4.2-6.6 billion cubic meters of water withdrawal in 2027, which is more than the total annual water withdrawal of 4-6 Denmark or half of the United Kingdom. This is concerning, as freshwater scarcity has become one of the most pressing challenges. To respond to the global water challenges, AI can, and also must, take social responsibility and lead by example by addressing its own water footprint. In this paper, we provide a principled methodology to estimate the water footprint of AI, and also discuss the unique spatial-temporal diversities of AI's runtime water efficiency. Finally, we highlight the necessity of holistically addressing water footprint along with carbon footprint to enable truly sustainable AI.

Published In: arXiv

Published On: 2023-04-06

Authors:

• Li, Pengfei
• Yang, Jianyi
• Islam, Mohammad A.
• Ren, Shaolei

Throughout 2022 and 2023, artificial intelligence (AI) has witnessed a period of rapid expansion and extensive, large-scale application. This accelerated development raises concerns about the electricity consumption and potential environmental impact of AI and data centers. This commentary reveals that the AI enthusiasm of 2022 and 2023 has put the AI server supply chain on track to deliver a more significant contribution to worldwide data center electricity consumption in the coming years.

Published In: Joule

Published On: 2023-10-01

Authors:

• de Vries, Alex

Recent progress in hardware and methodology for training neural networks has ushered in a new generation of large networks trained on abundant data. These models have obtained notable gains in accuracy across many NLP tasks. However, these accuracy improvements depend on the availability of exceptionally large computational resources that necessitate similarly substantial energy consumption. As a result these models are costly to train and develop, both financially, due to the cost of hardware and electricity or cloud compute time, and environmentally, due to the carbon footprint required to fuel modern tensor processing hardware. In this paper we bring this issue to the attention of NLP researchers by quantifying the approximate financial and environmental costs of training a variety of recently successful neural network models for NLP. Based on these findings, we propose actionable recommendations to reduce costs and improve equity in NLP research and practice.

Published In: arXiv

Published On: 2019-06-05

Authors:

• Strubell, Emma
• Ganesh, Ananya
• McCallum, Andrew

Creating efficiency in AI research will decrease its carbon footprint and increase its inclusivity as deep learning study should not require the deepest pockets.

Published In: Communications of the ACM

Published On: 2020-12-01

Authors:

• Schwartz, Roy
• Dodge, Jesse
• Smith, Noah A.
• Etzioni, Oren

Deep learning (DL) can achieve impressive results across a wide variety of tasks, but this often comes at the cost of training models for extensive periods on specialized hardware accelerators. This energy-intensive workload has seen immense growth in recent years. Machine learning (ML) may become a significant contributor to climate change if this exponential trend continues. If practitioners are aware of their energy and carbon footprint, then they may actively take steps to reduce it whenever possible. In this work, we present Carbontracker, a tool for tracking and predicting the energy and carbon footprint of training DL models. We propose that energy and carbon footprint of model development and training is reported alongside performance metrics using tools like Carbontracker. We hope this will promote responsible computing in ML and encourage research into energy-efficient deep neural networks.

Published In: arXiv

Published On: 2020-07-06

Authors:

• Anthony, Lasse F. Wolff
• Kanding, Benjamin
• Selvan, Raghavendra

The advent of cloud computing has provided people around the world with unprecedented access to computational power and enabled rapid growth in technologies such as machine learning, the computational demands of which incur a high energy cost and a commensurate carbon footprint. As a result, recent scholarship has called for better estimates of the greenhouse gas impact of AI: data scientists today do not have easy or reliable access to measurements of this information, which precludes development of actionable tactics. We argue that cloud providers presenting information about software carbon intensity to users is a fundamental stepping stone towards minimizing emissions.

Published In: 2022 ACM Conference on Fairness Accountability and Transparency

Published On: 2022-06-21

Authors:

• Dodge, Jesse
• Prewitt, Taylor
• Tachet des Combes, Remi
• Odmark, Erika
• Schwartz, Roy
• Strubell, Emma
• Luccioni, Alexandra Sasha
• Smith, Noah A.
• DeCario, Nicole
• Buchanan, Will

Machine learning (ML) requires using energy to carry out computations during the model training process. The generation of this energy comes with an environmental cost in terms of greenhouse gas emissions, depending on quantity used and the energy source. Existing research on the environmental impacts of ML has been limited to analyses covering a small number of models and does not adequately represent the diversity of ML models and tasks. In the current study, we present a survey of the carbon emissions of 95 ML models across time and different tasks in natural language processing and computer vision. We analyze them in terms of the energy sources used, the amount of CO2 emissions produced, how these emissions evolve across time and how they relate to model performance. We conclude with a discussion regarding the carbon footprint of our field and propose the creation of a centralized repository for reporting and tracking these emissions.

Published In: arXiv

Published On: 2023-02-16

Authors:

• Luccioni, Sasha
• Hernandez-Garcia, Alex

Progress in machine learning (ML) comes with a cost to the environment, given that training ML models requires significant computational resources, energy and materials. In the present article, we aim to quantify the carbon footprint of BLOOM, a 176-billion parameter language model, across its life cycle. We estimate that BLOOM's final training emitted approximately 24.7 tonnes of CO2eq if we consider only the dynamic power consumption, and 50.5 tonnes if we account for all processes ranging from equipment manufacturing to energy-based operational consumption. We also study the energy requirements and carbon emissions of its deployment for inference via an API endpoint receiving user queries in real-time. We conclude with a discussion regarding the difficulty of precisely estimating the carbon footprint of ML models and future research directions that can contribute towards improving carbon emissions reporting.

Published In: arXiv

Published On: 2022-11-03

Authors:

• Luccioni, Alexandra Sasha
• Viguier, Sylvain
• Ligozat, Anne-Laure

As ever larger language models grow more ubiquitous, it is crucial to consider their environmental impact. Characterised by extreme size and resource use, recent generations of models have been criticised for their voracious appetite for compute, and thus significant carbon footprint. Although reporting of carbon impact has grown more common in machine learning papers, this reporting is usually limited to compute resources used strictly for training. In this work, we propose a holistic assessment of the footprint of an extreme-scale language model, Noor. Noor is an ongoing project aiming to develop the largest multi-task Arabic language models-with up to 13B parameters-leveraging zero-shot generalisation to enable a wide range of downstream tasks via natural language instructions. We assess the total carbon bill of the entire project: starting with data collection and storage costs, including research and development budgets, pretraining costs, future serving estimates, and other exogenous costs necessary for this international cooperation. Notably, we find that inference costs and exogenous factors can have a significant impact on total budget. Finally, we discuss pathways to reduce the carbon footprint of extreme-scale models.

Published In: Proceedings of BigScience Episode #5 -- Workshop on Challenges & Perspectives in Creating Large Language Models

Published On: 2022-01-01

Authors:

• Lakim, Imad
• Almazrouei, Ebtesam
• Abualhaol, Ibrahim
• Debbah, Merouane
• Launay, Julien

Large language models (LLMs) have exploded in popularity due to their new generative capabilities that go far beyond prior state-of-the-art. These technologies are increasingly being leveraged in various domains such as law, finance, and medicine. However, these models carry significant computational challenges, especially the compute and energy costs required for inference. Inference energy costs already receive less attention than the energy costs of training LLMs-despite how often these large models are called on to conduct inference in reality (e.g., ChatGPT). As these state-of-the-art LLMs see increasing usage and deployment in various domains, a better understanding of their resource utilization is crucial for cost-savings, scaling performance, efficient hardware usage, and optimal inference strategies. In this paper, we describe experiments conducted to study the computational and energy utilization of inference with LLMs. We benchmark and conduct a preliminary analysis of the inference performance and inference energy costs of different sizes of LLaMA-a recent state-of-the-art LLM-developed by Meta AI on two generations of popular GPUs (NVIDIA V100 & A100) and two datasets (Alpaca and GSM8K) to reflect the diverse set of tasks/benchmarks for LLMs in research and practice. We present the results of multi-node, multi-GPU inference using model sharding across up to 32 GPUs. To our knowledge, our work is the one of the first to study LLM inference performance from the perspective of computational and energy resources at this scale.

Published In: 2023 IEEE High Performance Extreme Computing Conference (HPEC)

Published On: 2023-09-25

Authors:

• Samsi, Siddharth
• Zhao, Dan
• McDonald, Joseph
• Li, Baolin
• Michaleas, Adam
• Jones, Michael
• Bergeron, William
• Kepner, Jeremy
• Tiwari, Devesh
• Gadepally, Vijay

Recent years have seen a surge in the popularity of commercial AI products based on generative, multi-purpose AI systems promising a unified approach to building machine learning (ML) models into technology. However, this ambition of 'generality' comes at a steep cost to the environment, given the amount of energy these systems require and the amount of carbon that they emit. In this work, we propose the first systematic comparison of the ongoing inference cost of various categories of ML systems, covering both task-specific (i.e. finetuned models that carry out a single task) and 'general-purpose' models, (i.e. those trained for multiple tasks). We measure deployment cost as the amount of energy and carbon required to perform 1,000 inferences on representative benchmark dataset using these models. We find that multi-purpose, generative architectures are orders of magnitude more expensive than task-specific systems for a variety of tasks, even when controlling for the number of model parameters. We conclude with a discussion around the current trend of deploying multi-purpose generative ML systems, and caution that their utility should be more intentionally weighed against increased costs in terms of energy and emissions. All the data from our study can be accessed via an interactive demo to carry out further exploration and analysis.

Published In: arXiv

Published On: 2023-11-28

Authors:

• Luccioni, Alexandra Sasha
• Jernite, Yacine
• Strubell, Emma

In this work, we study LLMs from a carbon emission perspective, addressing both operational and embodied emissions, and paving the way for sustainable LLM serving. We characterize the performance and energy of LLaMA with 1B, 3B, and 7B parameters using two Nvidia GPU types, a latest-generation RTX6000 Ada and an older-generation T4. We analytically model operational carbon emissions based on energy consumption and carbon intensities from three grid regions --- each representing a different energy source mix, and embodied carbon emissions based on chip area and memory size. Our characterization and modeling provide us with an in-depth understanding of the performance, energy, and carbon emissions of LLM serving. Our findings highlight the potential for optimizing sustainable LLM serving systems by considering both operational and embodied carbon emissions.

Published In: ACM SIGEnergy Energy Informatics Review

Published On: 2024-12-01

Authors:

• Nguyen, Sophia
• Zhou, Beihao
• Ding, Yi
• Liu, Sihang

The rapid advancement of generative AI has heightened environmental concerns, particularly regarding carbon emissions. Our framework, Sprout, addresses these challenges by reducing the carbon footprint of inference in large language models (LLMs). Sprout introduces "generation directives" to guide the autoregressive generation process, achieving a balance between ecological sustainability and high-quality outputs. By employing a strategic optimizer for directive assignment and a novel offline quality evaluator, Sprout reduces the carbon footprint of generative LLM inference by over 40% in real-world evaluations, using the Llama model and global electricity grid data. This work is crucial as the rising interest in inference time compute scaling laws amplifies environmental concerns, emphasizing the need for eco-friendly AI solutions.

Published In: Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing

Published On: 2024-01-01

Authors:

• Li, Baolin
• Jiang, Yankai
• Gadepally, Vijay
• Tiwari, Devesh

Climate change is profoundly affecting nearly all aspects of life on earth, including human societies, economies, and health. Various human activities are responsible for significant greenhouse gas (GHG) emissions, including data centers and other sources of large-scale computation. Although many important scientific milestones are achieved thanks to the development of high-performance computing, the resultant environmental impact is underappreciated. In this work, a methodological framework to estimate the carbon footprint of any computational task in a standardized and reliable way is presented and metrics to contextualize GHG emissions are defined. A freely available online tool, Green Algorithms (www.green-algorithms.org) is developed, which enables a user to estimate and report the carbon footprint of their computation. The tool easily integrates with computational processes as it requires minimal information and does not interfere with existing code, while also accounting for a broad range of hardware configurations. Finally, the GHG emissions of algorithms used for particle physics simulations, weather forecasts, and natural language processing are quantified. Taken together, this study develops a simple generalizable framework and freely available tool to quantify the carbon footprint of nearly any computation. Combined with recommendations to minimize unnecessary CO2 emissions, the authors hope to raise awareness and facilitate greener computation.

Published In: Advanced Science

Published On: 2021-06-01

Authors:

• Lannelongue, Loic
• Grealey, Jason
• Inouye, Michael

The information communication technology sector will experience huge growth over the coming years, with 29.3 billion devices expected online by 2030, up from 18.4 billion in 2018. To reliably support the online services used by these billions of users, data centres have been built around the world to provide the millions of servers they contain with access to power, cooling and internet connectivity. Whilst the energy consumption of these facilities regularly receives mainstream and academic coverage, analysis of their water consumption is scarce. Data centres consume water directly for cooling, in some cases 57% sourced from potable water, and indirectly through the water requirements of non-renewable electricity generation. Although in the USA, data centre water consumption (1.7 billion litres/day) is small compared to total water consumption (1218 billion litres/day), there are issues of transparency with less than a third of data centre operators measuring water consumption. This paper examines the water consumption of data centres, the measurement of that consumption, highlights the lack of data available to assess water efficiency, and discusses and where the industry is going in attempts to reduce future consumption.

Published In: npj Clean Water

Published On: 2021-02-15

Authors:

• Mytton, David

Much of the world's data are stored, managed, and distributed by data centers. Data centers require a tremendous amount of energy to operate, accounting for around 1.8% of electricity use in the United States. Large amounts of water are also required to operate data centers, both directly for liquid cooling and indirectly to produce electricity. For the first time, we calculate spatially-detailed carbon and water footprints of data centers operating within the United States, which is home to around one-quarter of all data center servers globally. Our bottom-up approach reveals one-fifth of data center servers direct water footprint comes from moderately to highly water stressed watersheds, while nearly half of servers are fully or partially powered by power plants located within water stressed regions. Approximately 0.5% of total US greenhouse gas emissions are attributed to data centers. We investigate tradeoffs and synergies between data center's water and energy utilization by strategically locating data centers in areas of the country that will minimize one or more environmental footprints. Our study quantifies the environmental implications behind our data creation and storage and shows a path to decrease the environmental footprint of our increasing digital footprint.

Published In: Environmental Research Letters

Published On: 2021-06-01

Authors:

• Siddik, Abu Bakar
• Shehabi, Arman
• Marston, Landon
• Koningstein, Ross
• Masanet, Eric

Although energy and water are intertwined, the analysis of energy consumption in data centers (DCs) has largely ignored water needs. In this study, we utilize established DC efficiency metrics, power usage effectiveness (PUE) and water usage effectiveness (WUE), to analyze the water and energy use in two colocation DCs (DC1 and DC2) located in the greater Phoenix area. These DCs use distinct cooling systems: DC1 uses air-cooled chillers and DC2 uses both water-cooled chillers and evaporative cooling. We also present a method for analyzing the water-energy tradeoff in DCs as a useful tool for DC designers and managers to evaluate different cooling systems. The results highlight that although DC1’s average PUE is nearly 13% higher compared to DC2's average PUE, its source WUE is 66% lower than DC2. In addition, DC1’s PUE is more affected by seasonal changes (variation of dry-bulb temperature) , while DC2's WUE has a larger seasonal variation. Our cooling system analysis indicates that hybrid evaporative cooling has the least power consumption, while air-cooled chillers have the most water use. Utilizing free cooling and evaporative cooling when outside air condition permits (∼40% per year), results in substantial power and water savings in Phoenix DCs.

Published In: Resources, Conservation and Recycling

Published On: 2022-06-01

Authors:

• Karimi, Leila
• Yacuel, Leeann
• Degraft-Johnson, Joseph
• Ashby, Jamie
• Behl, Madhur
• Taha, Ahmad F.
• et al

The internet and associated Information and Communications Technologies (ICT) are diffusing at an astounding pace. As data centers (DCs) proliferate to accommodate this rising demand, their environmental impacts grow too. While the energy efficiency of DCs has been researched extensively, their water footprint (WF) has so far received little to no attention. This article conducts a preliminary WF accounting for cooling and energy consumption in DCs. The WF of DCs is estimated to be between 1047 and 151,061 m3/TJ. Outbound DC data traffic generates a WF of 1–205 liters per gigabyte (roughly equal to the WF of 1 kg of tomatos at the higher end). It is found that, typically, energy consumption constitues by far the greatest share of DC WF, but the level of uncertainty associated with the WF of different energy sources used by DCs makes a comprehensive assessment of DCs’ water use efficiency very challenging. Much better understanding of DC WF is urgently needed if a meaningful evaluation of this rapidly spreading service technology is to be gleaned and response measures are to be put into effect.

Published In: Sustainability

Published On: 2015-08-01

Authors:

• Ristic, Bora
• Madani, Kaveh
• Makuch, Zen

The demand for data center solutions with lower total cost of ownership and lower complexity of management is driving the creation of next generation datacenters. The information technology industry is in the midst of a transformation to lower the cost of operation through consolidation and better utilization of critical data center resources. Successful consolidation necessitates increasing utilization of capital intensive ldquoalways-onrdquo data center infrastructure, reduction in the recurring cost of power and management of physical resources like water. A 1MW data center operating with water-cooled chillers and cooling towers can consume 18,000 gallons per day to dissipate heat generated by IT equipment. However, this water demand can be mitigated by appropriate use of air-cooled chillers or free cooling strategies that rely on local weather patterns. Water demand can also fluctuate with seasons and vary across geographies. Water efficiency, like energy efficiency is a key metric to evaluate sustainability of the IT ecosystem. In this paper, we propose a procedure for calculation of water efficiency of a datacenter and provide guidance for a management system that can optimize IT performance while managing the tradeoffs between water and energy efficiency in conventional datacenters

Published In: 2009 IEEE International Symposium on Sustainable Systems and Technology

Published On: 2009-05-01

Authors:

• Sharma, Ratnesh
• Shah, Amip
• Bash, Cullen
• Christian, Theddeus
• Patel, Jay

The onsite water use of data centers (DCs) is becoming an increasingly important consideration within the policy and energy analysis communities, but has heretofore been difficult to quantify in macro-level DC energy models due to lack of reported water usage effectiveness (WUE) values by DC operators. This work addresses this important knowledge gap by presenting thermodynamically-compatible power usage effectiveness (PUE) and WUE values for a wide range of U.S. DC archetypes and climate zones, using a physics-based model that is validated with real-world data. Results enable energy analysts to more accurately analyze the onsite energy and water use of DCs by size class, cooling system type, and climate zone under many different operating conditions including operational setpoints. Sensitivity analyses further identify the variables leading to best-achievable PUE and WUE values by climate zone and cooling system type-including operational set points, use of free cooling, and cooling tower equipment and operational factors-which can support DC water- and energy-efficiency policy initiatives. The consistent PUE and WUE values may also be used in future work to quantify the indirect water use of DCs occurring in electrical power generating systems.

Published In: Research Square Platform LLC

Published On: 2021-08-03

Authors:

• Lei, Nuoa
• Masanet, Eric

Water, the key to life, is one of the earth's abundantly available elements. Yet access to freshwater is extremely limited and unevenly distributed, especially made worse with today's de-teriorating climate conditions, growing population, and aging water infrastructure [5]. The global water scarcity problem calls for every industry sector to actively contribute to water sustainability efforts. To that end, data centers with their scale and critical role in today's digital infrastructures, can, and also should, lead by example. Unfortunately, however, while data center operations achieved remarkable progress over the years in energy and carbon efficiency improvements, the water footprint of data centers has remained under the radar for much of the past decade. To that end, our goal in this paper is to shed light on data center water consumption and operational strategies that can improve data center sustainability

Published In: 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)

Published On: 2024-07-01

Authors:

• Islam, Mohammad A

Accurate reporting of energy and carbon usage is essential for understanding the potential climate impacts of machine learning research. We introduce a framework that makes this easier by providing a simple interface for tracking realtime energy consumption and carbon emissions, as well as generating standardized online appendices. Utilizing this framework, we create a leaderboard for energy efficient reinforcement learning algorithms to incentivize responsible research in this area as an example for other areas of machine learning. Finally, based on case studies using our framework, we propose strategies for mitigation of carbon emissions and reduction of energy consumption. By making accounting easier, we hope to further the sustainable development of machine learning experiments and spur more research into energy efficient algorithms.

Published In: Journal of Machine Learning Research

Published On: 2020-01-01

Authors:

• Henderson, Peter
• Hu, Jieru
• Romoff, Joshua
• Brunskill, Emma
• Jurafsky, Dan
• Pineau, Joelle

From an environmental standpoint, there are a few crucial aspects of training a neural network that have a major impact on the quantity of carbon that it emits. These factors include: the location of the server used for training and the energy grid that it uses, the length of the training procedure, and even the make and model of hardware on which the training takes place. In order to approximate these emissions, we present our Machine Learning Emissions Calculator, a tool for our community to better understand the environmental impact of training ML models. We accompany this tool with an explanation of the factors cited above, as well as concrete actions that individual practitioners and organizations can take to mitigate their carbon emissions.

Published In: arXiv

Published On: 2019-10-21

Authors:

• Lacoste, Alexandre
• Luccioni, Alexandra
• Schmidt, Victor
• Dandres, Thomas

The computation demand for machine learning (ML) has grown rapidly recently, which comes with a number of costs. Estimating the energy cost helps measure its environmental impact and finding greener strategies, yet it is challenging without detailed information. We calculate the energy use and carbon footprint of several recent large models-T5, Meena, GShard, Switch Transformer, and GPT-3-and refine earlier estimates for the neural architecture search that found Evolved Transformer. We highlight the following opportunities to improve energy efficiency and CO2 equivalent emissions (CO2e): Large but sparsely activated DNNs can consume <1/10th the energy of large, dense DNNs without sacrificing accuracy despite using as many or even more parameters. Geographic location matters for ML workload scheduling since the fraction of carbon-free energy and resulting CO2e vary ~5X-10X, even within the same country and the same organization. We are now optimizing where and when large models are trained. Specific datacenter infrastructure matters, as Cloud datacenters can be ~1.4-2X more energy efficient than typical datacenters, and the ML-oriented accelerators inside them can be ~2-5X more effective than off-the-shelf systems. Remarkably, the choice of DNN, datacenter, and processor can reduce the carbon footprint up to ~100-1000X. These large factors also make retroactive estimates of energy cost difficult. To avoid miscalculations, we believe ML papers requiring large computational resources should make energy consumption and CO2e explicit when practical. We are working to be more transparent about energy use and CO2e in our future research. To help reduce the carbon footprint of ML, we believe energy usage and CO2e should be a key metric in evaluating models, and we are collaborating with MLPerf developers to include energy usage during training and inference in this industry standard benchmark.

Published In: arXiv

Published On: 2021-04-21

Authors:

• Patterson, David
• Gonzalez, Joseph
• Hölzle, Urs
• Le, Quoc
• Liang, Chen
• Munguia, Lluis-Miquel
• Rothchild, Daniel
• So, David
• Texier, Maud
• Dean, Jeff

Machine Learning (ML) workloads have rapidly grown in importance, but raised concerns about their carbon footprint. Four best practices can reduce ML training energy by up to 100x and CO2 emissions up to 1000x. By following best practices, overall ML energy use (across research, development, and production) held steady at <15% of Google's total energy use for the past three years. If the whole ML field were to adopt best practices, total carbon emissions from training would reduce. Hence, we recommend that ML papers include emissions explicitly to foster competition on more than just model quality. Estimates of emissions in papers that omitted them have been off 100x-100,000x, so publishing emissions has the added benefit of ensuring accurate accounting. Given the importance of climate change, we must get the numbers right to make certain that we work on its biggest challenges.

Published In: arXiv

Published On: 2022-04-11

Authors:

• Patterson, David
• Gonzalez, Joseph
• Hölzle, Urs
• Le, Quoc
• Liang, Chen
• Munguia, Lluis-Miquel
• Rothchild, Daniel
• So, David
• Texier, Maud
• Dean, Jeff

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 require 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

Published In: Sustainability

Published On: 2022-01-01

Authors:

• Ligozat, Anne-Laure
• Lefevre, Julien
• Bugeau, Aurelie
• Combaz, Jacques

While there is a growing effort towards AI for Sustainability (e.g. towards the sustainable development goals) it is time to move beyond that and to address the sustainability of developing and using AI systems. In this paper I propose a definition of Sustainable AI; Sustainable AI is a movement to foster change in the entire lifecycle of AI products (i.e. idea generation, training, re-tuning, implementation, governance) towards greater ecological integrity and social justice. As such, Sustainable AI is focused on more than AI applications; rather, it addresses the whole sociotechnical system of AI. I have suggested here that Sustainable AI is not about how to sustain the development of AI per say but it is about how to develop AI that is compatible with sustaining environmental resources for current and future generations; economic models for societies; and societal values that are fundamental to a given society. I have articulated that the phrase Sustainable AI be understood as having two branches; AI for sustainability and sustainability of AI (e.g. reduction of carbon emissions and computing power). I propose that Sustainable AI take sustainable development at the core of its definition with three accompanying tensions between AI innovation and equitable resource distribution; inter and intra-generational justice; and, between environment, society, and economy. This paper is not meant to engage with each of the three pillars of sustainability (i.e. social, economic, environment), and as such the pillars of sustainable AI. Rather, this paper is meant to inspire the reader, the policy maker, the AI ethicist, the AI developer to connect with the environment—to remember that there are environmental costs to AI. Further, to direct funding towards sustainable methods of AI.

Published In: AI and Ethics

Published On: 2021-08-01

Authors:

• van Wynsberghe, Aimee

In the rapidly evolving realm of artificial intelligence, deploying large language models (LLMs) poses increasingly pressing computational and environmental challenges. This paper introduces MELODI - Monitoring Energy Levels and Optimization for Data-driven Inference - a multifaceted framework crafted to monitor and analyze the energy consumed during LLM inference processes. MELODI enables detailed observations of power consumption dynamics and facilitates the creation of a comprehensive dataset reflective of energy efficiency across varied deployment scenarios. The dataset, generated using MELODI, encompasses a broad spectrum of LLM deployment frameworks, multiple language models, and extensive prompt datasets, enabling a comparative analysis of energy use. Using the dataset, we investigate how prompt attributes, including length and complexity, correlate with energy expenditure. Our findings indicate substantial disparities in energy efficiency, suggesting ample scope for optimization and adoption of sustainable measures in LLM deployment. Our contribution lies not only in the MELODI framework but also in the novel dataset, a resource that can be expanded by other researchers. Thus, MELODI is a foundational tool and dataset for advancing research into energy-conscious LLM deployment, steering the field toward a more sustainable future.

Published In: arXiv

Published On: 2024-07-04

Authors:

• Husom, Erik Johannes
• Goknil, Arda
• Shar, Lwin Khin
• Sen, Sagar

Recent advancements in Generative AI (GenAI), particularly Large Language Models (LLMs), have led to significant increases in computational power and energy demands. These models are not only becoming larger and more complex but are also being widely deployed across various platforms, reaching millions of users and developers. This widespread use has escalated the overall energy consumption and environmental impact associated with these technologies. To address these concerns, we introduce EcoLogits, an open-source Python package designed to evaluate the environmental footprint of using GenAI models. EcoLogits estimates energy consumption and multi-criteria environmental impacts of GenAI API requests using a bottom-up Life Cycle Assessment methodology, considering both direct energy use and hardware production impacts

Published In: Journal of Open Source Software

Published On: 2025-07-09

Authors:

• Rince, Samuel
• Banse, Adrien

Data centers are energy-intensive infrastructures that generate, manage, and store information for our interconnected society. Models based on Artificial Intelligence (AI), such as ChatGPT, are increasingly accessible, leading to significant energy consumption and associated carbon emissions. Assessing the environmental footprint of Generative AI (GenAI) is essential for evaluating its sustainability and promoting responsible AI development. In this work, a comprehensive environmental assessment of GenAI systems was performed – which includes both training and inference phases – using a life cycle assessment (LCA) approach. Prior studies have primarily focused on server-level assessments or energy consumption analyses. In contrast, this work considers the full lifecycle of data centers and evaluates environmental impacts across complete architectural configurations, offering a broader and more integrated perspective. Finally, multiple data center architectures are compared, from edge systems to AI dedicated infrastructures. Two simulation-based use cases are presented: (1) A 20-year simulation comparing different data center architectures across three indicators – total emissions, emissions per year, and emissions per installed IT MW. For a subset of these architectures, emissions per Floating-Point Operations Per Second (FLOPS) are also included to assess performance efficiency – considering that FLOPS estimations can only be done on GPU based data center architectures; (2) A focused simulation comparing the environmental footprint of three large language models – GPT-4o, LLaMA 3.1405B, and DeepSeek V3 – to quantify trade-offs between benchmark performance and environmental impact. By expanding the scope of assessment and incorporating varied use cases, this work aims to inform strategies for minimizing the environmental costs of GenAI while advancing sustainable AI development.

Published In: Applied Energy

Published On: 2026-03-01

Authors:

• d'Orgeval, Adrien
• Thirion, Benoit
• et al

In recent decades, artificial intelligence has undergone transformative advancements, reshaping diverse sectors such as healthcare, transport, agriculture, energy, and the media. Despite the enthusiasm surrounding AI’s potential, concerns persist about its potential negative impacts, including substantial energy consumption and ethical challenges. This paper critically reviews the evolving landscape of AI sustainability, addressing economic, social, and environmental dimensions. The literature is systematically categorized into “Sustainability of AI” and “AI for Sustainability”, revealing a balanced perspective between the two. The study also identifies a notable trend towards holistic approaches, with a surge in publications and empirical studies since 2019, signaling the field’s maturity. Future research directions emphasize delving into the relatively under-explored economic dimension, aligning with the United Nations’ Sustainable Development Goals (SDGs), and addressing stakeholders’ influence.

Published In: Analytics

Published On: 2024-03-01

Authors:

• Dhiman, Rachit
• Miteff, Sofia
• Wang, Yuancheng
• Ma, Shih-Chi
• Amirikas, Ramila
• Fabian, Benjamin