Since the release of ChatGPT in 2022, people have begun using artificial intelligence for almost anything in their day-to-day lives. From budgeting to essay outlines to grocery lists, AI has reached the everyday population. When most people use AI, they type in their question or prompt and watch as the response spills out in seconds. However, what most AI users don’t see is the effects of these endless prompts and questions on the environment.
The biggest culprit when it comes to AI’s damage to the environment tends to be the electronic waste coming from the data centers that house AI. According to the United Nation’s Environment Programme (UNEP), the number of data centers has surged from 500,000 to nearly 8 million since 2012.
“There is still much we don’t know about the environmental impact of AI but some of the data we do have is concerning,” said Golestan (Sally) Radwan, the Chief Digital Officer of the UNEP. “We need to make sure the net effect of AI on the planet is positive before we deploy the technology at
scale.”
UNEP also highlights the energy increase in this new technology.
“A request made through ChatGPT, an AI based virtual assistant, consumes 10 times the electricity of a Google search, according to the International Energy Agency,” UNEP said.
In addition to electrical waste, there is also concern for the large amount of water the data centers that house AI are consuming.
“Globally, AI-related infrastructure may soon consume six times more water than Denmark, a country of 6 million, according to one estimate,” UNEP said. “That is a problem when a quarter of humanity already lacks access to clean water and sanitation.”
The topic of AI’s effects on the environment also brings up the issue of the materials used to develop AI technology. For example, an AI microchip uses rare elements that tend to be mined in unethical and environmentally destructive ways.
UNEP urges governments to introduce legislation to reduce the negative environmental impact of AI. “Governments are racing to develop national AI strategies but rarely do they take the environment and sustainability into account,” UNEP said. “The lack of environmental guardrails is no less dangerous than the lack of other AI related safeguards.”
While there are great concerns on the damage AI can have on the environment, this new technology also brings to light many new opportunities to help and advocate for our earth.
Sage Garretson, STA’s own information technology expert and self-proclaimed eco activist, offers his insight and recognizes the positive and negative effects AI has on the environment.
“People have suggested that AI is a double-edged sword,” Garretson said. “It can cause significant environmental harm through its energy consumption and emissions but also can be part of the solution to climate change, like the Spherical Dyffusion Model from UC San Diego and Allen Institute for AI.
“Take the Spherical Dyffusion Model out of UC San Diego and the Allen Institute for AI, for example–this technology can run 100-year climate simulations at 25 hour intervals, which would take most other models weeks to process,” Garretson said. “All while it significantly lowers the computational load. It works as a probabilistic emulator for existing climate models, offering better accuracy than traditional methods.”
Large companies are following suit as some of the biggest buyers of renewable energy are companies like Google and Microsoft. For example, in 2020 Google announced that by 2030, they plan for all of their data centers to be completely run by renewable energy sources like solar and wind. One competitor of Open AI called DeepMind, which prioritizes research through artificial intelligence, has also been making strides in environmental conservation.
“In fact, in 2016, Google’s DeepMind AI was able to significantly reduce the energy used for cooling in their data centers by up to 40%,” Garretson said. “They achieved this by using machine learning to analyze data from thousands of sensors and optimize the cooling centers for maximum efficiency. This resulted in a 15% reduction in overall energy consumption in those data centers, which is a considerable improvement.”
Garretson highlights that since these companies are private, information regarding their energy consumption and emissions is not public, causing many people to be in the dark about the implications of AI.
“We need transparency from AI companies regarding their energy usage and accountability,” Garretson said. “Experts argue that if companies provide data on energy consumption, people can make informed decisions about their AI usage, instead of psyching yourself out about them.”
However, reducing AI’s negative environment isn’t something only the big corporations need to be cognizant of. AI users can also be aware of their daily AI usage. Garretson warns users not to waste AI use on frivolous tasks.
“Users need to realize that they don’t need to rely on generative AI for non-essential things; people should not use ChatGPT as a calculator,” Garretson said. “Like you really don’t need it . Not only is it bad at arithmetic, it’s literally not made to do math, but it’s also orders of magnitude, more energy and a crazy amount of water for something that does not need water.”
One aspect of AI to be cognizant of, as a user, is which kind of AI models are utilized.
“Some AI models, like Large Language Models, are very energy intensive but may not directly contribute to solving climate change,” Garretson said. “Some models utilize Mixture-of-Experts architecture, so they activate only a small fraction of their parameters at a given time, which significantly reduces the computational cost and makes them more efficient.”
