AI Complicates Sustainability Profiles of Tech Companies — ESG Insight

livemint.com:

• Training AI-powered models requires high levels of energy and water consumption, but they can also be used to help offset greenhouse-gas emissions and control water consumption by making systems more efficient • Energy management and water-and-wastewater management aren’t considered financially material for the technology and communications sector, although it contains most AI-focused companies • Among the 10 largest companies in the Wisdom Tree Artificial Intelligence and Innovation Index, Microsoft and Alphabet lead in energy management, while Nvidia, Meta Platforms and Microsoft are among the leaders for water-and-wastewater management, Dow Jones sustainability data shows The increasing development and implementation of artificial intelligence by businesses has both positive and negative sustainability implications for companies, specifically those in the technology and communications sector. Training AI-powered models, such as the one behind ChatGPT, requires significant levels of energy and water consumption, as data centers use large amounts of water for cooling.

For instance, training an AI-powered model such as GPT-3 can generate up to 552 metric tons of carbon emissions from the energy it uses, equivalent to three times a jet plane’s carbon emissions for a round trip from San Francisco to New York, according to researchers at Cornell University. When it comes to water usage, Alphabet’s Google used up to an estimated million liters of water to train its large-language model, known as LaMDA, according to research by Shaolei Ren, a professor of electrical and computer engineering at the University of California Riverside, The Wall Street Journal reported.