Scientific Paper · 2019-06-05
Energy and Policy Considerations for Deep Learning in NLP
Training large neural networks improves NLP accuracy but requires significant computational resources, leading to high financial and environmental costs.
ENVIRONMENTAL DOMAIN
Energy & climate
Electricity demand, grids, emissions, and climate systems.
Browse all 821 resources →Counts reflect provisional metadata classification, not measured environmental impact. A record may belong to multiple topics.
Topics in this domain
View this topic through four lenses
Share of the 821 records in this view. Pathways can overlap; bars do not add to 100%.
Explore the connected collections
Research papers and scholarly literature on impacts, applications, and methods.
Scientific Paper · 2024-10-15
Power Hungry Processing: Watts Driving the Cost of AI Deployment?
Generative AI models consume significantly more energy and emit more carbon than task-specific models, raising concerns about their environmental impact.
Scientific Paper · 2024-12-01
Comparative Analysis of Artificial Intelligence Models for Real-Time and Future Forecasting of Environmental Conditions: A Wood-Frame Historic Building Case Study
Machine learning enhances real-time and future climate monitoring in historic buildings, aiding preservation efforts amid climate change.
Scientific Paper · 2024-12-04
Probabilistic Weather Forecasting with Machine Learning
GenCast, an ML-based weather model, outperforms traditional forecasts in accuracy and speed, enhancing weather prediction and risk assessment.
What is missing from the metadata?
Missing metadata is a curation task, not evidence of absent research or activity. The Resource Hub does not currently contain a validated incident dataset, expert survey, or measurements of net environmental benefit.