The training of a single artificial intelligence (AI) program can require as much as 284 tons of carbon dioxide equivalent, or about five times the lifetime emissions of an average car, according to researchers at the University of Massachusetts Amherst.
A common approach to training AI involves feeding the program billions of written articles so it learns to understand the meaning of words and how sentences are constructed. The researchers measured the environmental impact of this approach by training four different AIs for one day each, and sampling the energy consumption of each.
The team then calculated the total power required to train each AI, and estimated the resulting carbon footprint based on average carbon emissions of power production in the U.S.
From New Scientist
View Full Article
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA