How Much Carbon Emissions Does AI Produce? Researchers Introduce a New Measurement Tool
The rapid advancement of artificial intelligence (AI) across various sectors carries significant environmental consequences due to high energy consumption during model training. In response, a team of researchers consisting of Alexandre Lacoste, Alexandra Luccioni, Victor Schmidt, and Thomas Dandres has introduced a new method for calculating greenhouse gas emissions through the Machine Learning Emissions Calculator.
The study, titled Quantifying the Carbon Emissions of Machine Learning, aims to provide transparency regarding the ecological impact of technology development. Until now, the industry has tended to focus on model accuracy, while the energy usage of Graphics Processing Units (GPUs) or Tensor Processing Units (TPUs) working continuously has often been overlooked.
Based on the research data, the scale of AI’s carbon footprint does not depend solely on model size, but is influenced by four main factors.
The introduction of the Machine Learning Emissions Calculator enables developers to estimate emissions based on location, processor type, and training duration even before a project begins. The tool is designed not to hinder innovation, but rather to encourage the technology community to make more responsible decisions.
Measures such as scheduling training at low-carbon data centres or simplifying experiments can reduce the carbon footprint without compromising the quality of AI models. Emissions transparency is predicted to become a new standard in the technology industry of the future, ensuring that artificial intelligence develops in harmony with environmental preservation efforts.
Source: jawawa.id (https://jawawa.id/newsitem/how-much-carbon-emissions-does-ai-produce-researchers-introduce-a-new-measurement-tool-1786035059)
Originally published by Media Indonesia (https://mediaindonesia.com/teknologi/919183/berapa-emisi-karbon-yang-dihasilkan-ai-ini-alat-ukur-baru-dari-para-peneliti), translated from Indonesian