Energy and Policy Considerations for Modern Deep Learning Research

By Emma Strubell, Ananya Ganesh, Andrew McCallum. 2020

901 citations recorded by the source corpus.

The field of artificial intelligence has experienced a dramatic methodological shift towards large neural networks trained on plentiful data. This shift has been fueled by recent advances in hardware and techniques enabling remarkable levels of computation, resulting in impressive advances in AI across many applications. However, the massive computation required to obtain these exciting results is costly both financially, due to the price of specialized hardware and electricity or cloud compute time, and to the environment, as a result of non-renewable energy used to fuel modern tensor processing hardware. In a paper published this year at ACL, we brought this issue to the attention of NLP researchers by quantifying the approximate financial and environmental costs of training and tuning neural network models for NLP (Strubell, Ganesh, and McCallum 2019). In this extended abstract, we briefly summarize our findings in NLP, incorporating updated estimates and broader information from recent related publications, and provide actionable recommendations to reduce costs and improve equity in the machine learning and artificial intelligence community.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety, Robustness & Interpretability

Citations in the corpus

High-Resolution Image Synthesis with Latent Diffusion Models (2021, background): “This has two consequences for the research community and users in general: Firstly, training such a model requires massive computational resources only available to a small fraction of the field, and leaves a huge carbon footprint [63, 83].”
Scaling to Many Languages with a Triaged Multilingual Text-Dependent and Text-Independent Speaker Veriﬁcation System (2021, background): “In scaling up to many languages, an opportunity is presented to improve efﬁciency (as inspired by [27]).”
A Practical Survey on Faster and Lighter Transformers (2021, background): “[109], this creates three major issues: (1) it stiles creativity as researchers and practitioners that do not have access to considerable resources are not able to experiment with Transformers, (2) it reinforces the łrich get richerž cycle where successful labs and companies receive more funding due”
SpeakerStew: Scaling to Many Languages with a Triaged Multilingual Text-Dependent and Text-Independent Speaker Verification System (2021, background): “In scaling up to many languages, an opportunity is presented to improve efficiency [31].”
Challenges in Deploying Machine Learning: A Survey of Case Studies (2020, background): “[49] regarding the impact the training of ML models has on the environment.”
Systematic Rectification of Language Models via Dead-end Analysis (2023, background): “We are fully aware that many more potential harms exist in such LLMs (Sheng et al., 2019; Huang et al., 2020; Strubell et al., 2020).”
Contrastive Learning Using Graph Embeddings for Domain Adaptation of Language Models in the Process Industry (2025, unknown): “…training domain-specific LMs from scratch (Lee et al., 2019; Beltagy et al., 2019; Huang et al., 2019), or domain-adaptive continual pretraining (DAPT) using smaller amounts of domain text data when labeled domain task-specific data is limited (Gururangan et al., 2020; Strubell et al., 2020).”