Communications of the ACM
Researchers Developing Tools to Calculate AI's Carbon Footprint
New tools can help developers understand the environmental impact of their artificial intelligence models.
Credit: EcoMatcher
Researchers in Canada and the U.S. are creating tools to calculate the carbon footprint of artificial intelligence (AI) models to help developers understand their environmental impact.
The goal is to encourage scientists to ameliorate such effects by training algorithms on servers that run on hydroelectricity or solar power, using less training data and pre-trained models.
Sasha Luccioni at Canada's Mila-Quebec AI Institute is co-developing a tool expected to be released in September that produces an estimate of carbon dioxide emissions attributable to the training of an AI model, based on the energy grid the user is connected to, the processors employed, and the model's runtime.
Meanwhile, a U.S.-Canadian team developed a carbon-footprint calculator for machine learning experiments, which they released for free on GitHub in February.
Said Stanford University's Peter Henderson, “Our main goal was basically to raise awareness.”
From The Wall Street Journal
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found