From bitcoins to NFTs to artificial intelligence, the freshest and shiniest forms of digital innovation capture the hearts of the masses with haste.
Launched in 2022, OpenAI’s ChatGPT was expected to completely redefine the future of the workplace and beyond. Although the jury is still out on how much day-to-day life will change from these generative chatbots, one aspect will be irrevocably impacted by this tech surge: the climate.
How do chatbots run?
Chatbots run from information collected in large language models (LLM), “neural networks, which are machine learning models that take an input and perform mathematical calculations to produce an output.” These inputs include billions of pieces of text from the Internet and, increasingly, non-public data.
In order to maintain this high level of computation, powerful computers using a lot of energy are required. The collection of computers, housed in data centers, have sparked a lot of critique regarding the energy and resources needed to keep them running.
The demand for electricity is in part due to AI
According to the International Energy Agency, the global demand for electricity is expected to double between 2022 and 2026, with AI attributed for a large part of that surge. The energy harm does just stop at the data centers — as demand and investment into generative AI increases, the increase in computation only requires more energy.
“The entire supply chain is a part of the problem,” said Tamara Kneese, project director of Data and Society’s Algorithmic Impact Methods Lab. “To produce the kinds of chips that you need for working with AI to really allow for the powerful level of computation, you have to manufacture even more intense, more powerful chips. So companies, in order to keep up with the demand posed by generative AI and other high-energy workloads, they are building more carbon intensive data centers.”
Data centers are used for AI
These data centers take up land and disrupt the surrounding communities, with reports of nausea, hearing loss, and migraines from excessive noise pollution harming nearby residents.
Additionally, data centers require gallons of water for cooling — experts estimate that “training” a chatbot like OpenAI’s ChatGPT-3 could require 185,000 gallons of water and the equivalent of a 16-ounce gulp of water for every 10-50 responses given once it’s in use. Research predicts data center water usage could demand 450 million gallons of water per day by 2030. Some data centers are located in already drought-stricken areas like Goodyear, Arizona or Santiago, Chile.
Big Tech companies like Microsoft and Google, major investors into the future of artificial intelligence, have diverted plans toward carbon neutrality and climate repair directly because of data center usage. In 2021, Google committed to being carbon zero by 2030, but increased carbon emissions by 13% in 2023. Similarly, Microsoft’s carbon emissions by 30% between 2020 and 2023 due to energy needed for data center expansion.
“A lot of big tech companies were transitioning to only using renewable energy, but because they’re expanding at such a fast pace and there isn’t enough renewable energy infrastructure, they’re not transitioning to using renewable energy,” said Kneese. “We’re seeing coal plants not being closed when they were supposed to because of the demand for energy.”
More on the ecological harm of data centers
The ecological harm of data centers is not something that is being localized to far-off corners of the world — places like Northern Virginia, home to the highest concentration of data centers in the world, are seeing water usage jump more than 60% over 4 years from data center demands.
As tech giants are looking to expand data centers rapidly, critics question the necessity of such ambitious projects in the face of unprecedented climate crises.
“If we have a limited amount of renewable energy or there’s only so much electricity that we have available, does it make sense to funnel it into things like generative AI, rather than, say, air conditioning for people who are experiencing the crisis of climate change already?” Kneese said.
