Green AI: métricas para eficiência energética e sustentabilidade dos algoritmos de inteligência artificial e data centers

Ivini Ferraz

Resumo
A inteligência artificial (IA) tem impulsionado avanços tecnológicos significativos em áreas como saúde, agricultura e energia. No entanto, seu crescimento exponencial apresenta desafios relevantes em relação ao impacto ambiental. Grandes data centers ao redor do Mundo tornaram-se grandes consumidores de energia e emissores de carbono, atingindo níveis comparáveis ao consumo energético de cidades ou até mesmo de países inteiros. O presente artigo explora publicações recentes na área de computação de alto desempenho (HPC). Foram analisadas e comparadas diferentes métricas, indicadores de desempenho energético, abordagens metodológicas e técnicas voltadas para a promoção de uma IA mais sustentável (Green AI). Os resultados indicam o surgimento de um campo científico historicamente paradoxal, que busca conciliar a eficiência energética e a sustentabilidade diante do aumento dos data centers e do uso intensivo de tokens, prompts, dados e algoritmos, necessários para o treinamento e a produtividade de modelos de IA.

Palavras-chave
Inteligência artificial; Green AI; Sustentabilidade; Data centers; IA responsável.

Abstract
Green AI: Metrics for energy efficiency and the sustainability of artificial intelligence algorithms and data centers. Artificial intelligence (AI) has been driving significant technological advancements in areas such as health, agriculture, and energy. However, its exponential growth presents relevant challenges regarding environmental impact. Large data centers around the World have become major energy consumers and carbon emitters, reaching levels comparable to the energy consumption of cities or even entire countries. This article explores recent publications in the field of high-performance computing (HPC). Different metrics, energy performance indicators, methodological approaches, and techniques aimed at promoting more sustainable AI (Green AI) were analyzed and compared. The results indicate the emergence of a historically paradoxical scientific field that seeks to reconcile energy efficiency and sustainability in the face of the increasing number of data centers and the intensive use of tokens, prompts, data, and algorithms required for training and productivity in AI models.

Keywords
Artificial intelligence; Green AI; Sustainability; Data Centers; Responsible AI.

DOI
10.21438/rbgas(2024)112914


Texto completo

 Baixar este arquivo PDF

 


Referências
Alcott, B. Jevonsapos; paradox. Ecological Economics, v. 54, n. 1, p. 9-21, 2005. https://doi.org/10.1016/j.ecolecon.2005.03.020

Berdardo, F.; Silva, G.; Gritz, M.; Ferro, M; Schulze, B. Avaliação do consumo de energia e o impacto da emissão de CO₂ para algoritmos de inteligência artificial. Petrópolis: Laboratório Nacional de Computação Científica, 2022.

Clemm, C.; Stobbe, L.; Wimalawarne, K.; Druschke, J. Towards Green AI: Current status and future research. arXiv Preprint, 2024. https://doi.org/10.48550/arXiv.2407.10237

Databricks. State of Data+AI: Data intelligence and the race to customize LLMs. San Francisco: Databricks, 2024. Disponível em: <https://www.databricks.com/sites/default/files/2024-06/state-of-data-ai-report.pdf. Acesso em: 30 set. 2024.

de Vries, A. The growing energy footprint of artificial intelligence. Joule, v. 7, n. 10, p. 2191-2194, 2023. https://doi.org/10.1016/j.joule.2023.09.004

Duncan, G.; Thorpe, D.; Syafiq, M.; Beets, K. Data Centers 2024 Global Outlook: AI and the green energy transition will bring new challenges and opportunities. Chicago: JLL, 2024. Disponível em: <https://www.jll.com.br/content/dam/jll-com/documents/pdf/research/global/jll-data-center-outlook-global-2024.pdf>. Acesso em: 10 set. 2024.

Dutta, S.; Innan, N.; Yahia, S.; Shafique, M. AQ-PINNs: Attention-enhanced quantum physics-informed neural networks for carbon-efficient climate modeling. arXiv Preprint, 2024. https://doi.org/10.48550/arXiv.2409.01626

EPE - Empresa de Pesquisa Energética. Balanço Energético Nacional (BEN): relatório síntese 2024 - Ano base 2023. Brasília: EPE, 2024. Disponível em: <https://www.epe.gov.br/sites-pt/publicacoes-dados-abertos/publicacoes/PublicacoesArquivos/publicacao-819/topico-715/BEN_Síntese_2024_PT.pdf>. Acesso em: 03 out. 2024.

García-Martín, E.; Rodrigues, C. F.; Riley, G.; Grahn, H. Estimation of energy consumption in machine learning. Journal of Parallel and Distributed Computing, v. 134, p. 75-78, 2019. https://doi.org/10.1016/j.jpdc.2019.07.007

Gartner. Hype cycle for emerging technologies highlights developer productivity, total experience, AI and security. Stanford: Gartner, 2024.

Global Energy Monitor. Global wind power tracker report. Covina: Global Energy Monitor, 2024. Disponível em: <https://globalenergymonitor.org>. Acesso em: 03 out. 2024.

Google. Relatório de sustentabilidade 2024. Mountain View: Google Inc., 2024. Disponível em: <https://www.google.com/sustainability/report2024>. Acesso em: 02 out. 2024.

IEA - International Energy Agency. Global data centre energy demand by data centre type, 2010-2022. Paris: IEA, 2020. Disponível em: <https://www.iea.org/data-and-statistics/charts/global-data-centre-energy-demand-by-data-centre-type-2010-2022>. Acesso em: 20 set. 2024.

InfoQ. Investimentos em IA para redução de emissões de carbono até 2030. Relatório, 2024. Disponível em: <https://www.infoq.com/>. Acesso em: 12 set. 2024.

Li, P.; Yang, J.; Islam, M.; Ren, S. Making AI less thirsty: Uncovering and addressing the secret water footprint of AI models. arXiv Preprint, 2023. https://doi.org/10.48550/arXiv.2304.03271

Lopes, A. Brasil concentra 75% dos investimentos previstos para data centers e inteligência artificial. Exame, 2024. Disponível em: <https://exame.com/inteligencia-artificial/brasil-concentra-75-dos-investimentos-previstos-para-data-centers-e-inteligencia-artificial/>. Acesso em: 29 nov. 2024.

Luccioni, A. S.; Viguier, S.; Ligozat, A.-L. Estimating the carbon footprint of BLOOM, a 176B parameter language model. Journal of Machine Learning Research, v. 24, p. 1-15, 2023.

Maslej, N.; Fattorini, L.; Perrault, R.; Parli, V.; Reuel, A.; Brynjolfsson, E.; Etchemendy, J.; Ligett, K.; Lyons, T.; Manyika, J.; Niebles, J. C.; Shoham, Y.; Wald, R.; Clark, J. The AI Index 2024 Annual Report. Stanford: AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, 2024. Disponível em: <https://aiindex.stanford.edu/wp-content/uploads/2024/05/HAI_AI-Index-Report-2024.pdf>. Acesso em: 10 set. 2024.

Microsoft. How AI can enable a sustainable future. Reino Unido: PWC, 2023. Disponível em: <https://www.pwc.com/gx/en/issues/analytics/assets/pwc-ai-analysis-sizing-the-prize-report.pdf>. Acesso em: 10 set. 2024.

Olabi, A. G.; Abdelghafar, A. A.; Maghrabie, H. M.; Sayed, E. T.; Rezk, H.; Al Radi, M.; Obaideen, K.; Abdelkareem, M. A. Application of artificial intelligence for prediction, optimization, and control of thermal energy storage systems. Thermal Science and Engineering Progress, v, 39, 101730, 2023. https://doi.org/10.1016/j.tsep.2023.101730

Schwartz, R.; Dodge, J.; Smith, N. A.; Etzioni, O. Green AI. arXiv Preprint, 2019. https://doi.org/10.48550/arxiv.1907.10597

Stamoulis, D.; Cai, E.; Juan, D.-C.; Marculescu, D. HyperPower: Power- and memory-constrained hyper-parameter optimization for neural networks. Proceedings of the 2018 Design, Automation, and Test in Europe Conference, Dresden, Germany, p. 19-24, 2018. https://doi.org/10.23919/DATE.2018.8341973

Stanford University. Artificial intelligence: Index report 2024. Stanford: Stanford University, 2024. Disponível em: <https://aiindex.stanford.edu/wp-content/uploads/2024/05/HAI_AI-Index-Report-2024.pdf>. Acesso em: 20 set. 2024.

Strubell, E.; Ganesh, A.; McCallum, A. Energy and policy considerations for deep learning in NLP. Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, p. 3645-3650, 2019.

Team Ciente. Red AI vs Green AI. Medium, 2024. Disponível em: <https://medium.com/@ciente/red-ai-vs-green-ai-d7c9e4c6e6af>. Acesso em: 30 set. 2024.

Uptime Institute. Global data center survey 2022: Resiliency remains critical in a volatile world. Singapore: Uptime Institute Intelligence, 2022. Disponível em: <https://uptimeinstitute.com/resources/research-and-reports/uptime-institute-global-data-center-survey-results-2022>. Acesso em: 20 set. 2024.

Verdecchia, R.; Sallou, J.; Cruz, L. A systematic review of Green AI. WIREs Data Mining and Knowledge Discovery, v. 13, n. 4, e1507, 2021. https://doi.org/10.1002/widm.1507


 

ISSN 2359-1412