Why this is a growing concern

Overview

While many believe that artificial intelligence (AI) can help us reduce our environmental impact, a recent study from the Yale School of Environment paints a different picture. This shows that AI actually has a significant energy footprint and can harm the environment in various ways, such as using a lot of water and electricity.

The study warns that as AI becomes more widespread, its environmental impact will only increase. Jeremy Tamanini, founder of Dual Citizen, a platform that helps governments, companies and investors improve sustainability through the use of AI, says that if we don’t regulate AI’s carbon footprint, it will crowd out our efforts to achieve sustainable development goals to reduce can seriously affect.

“AI systems will become increasingly embedded in the global economy in the next decade. They should be treated like any other factor that increases energy and resource consumption. If not, these systems will erode our global progress around SDG and emission reduction targets,” he said.

The growing concern

Although AI offers incredible potential, its rapid growth poses environmental challenges, especially in terms of water use. A study by Yale found that ChatGPT-3 alone can consume up to half a liter of water for just 10 to 50 reactions. This high water demand is due to the need to cool the equipment that powers AI systems.

Tech giants like Google, Microsoft and Apple use millions of liters of fresh water in their data centers to ensure the smooth functioning of their AI platforms. In 2022, Google alone will have consumed nearly 20 billion liters of fresh water for cooling. As these companies expand their data center operations globally, the pressure on fresh water resources, especially in countries like India, is expected to increase significantly.

Tamanini emphasized that as AI becomes more integrated into daily life, it should be considered a factor that increases energy and resource consumption. Failure to address these issues could hinder global progress on sustainable development and emission reduction goals.

In contrast, the UN’s 2023 World Water Development Report warns of a looming global water shortage, with billions of people already facing water scarcity. This highlights the urgent need to address the water consumption of AI and other technologies to ensure a sustainable future.

Lack of data

Besides needing a lot of water to operate, AI also creates carbon emissions through the use of non-renewable energy, according to a report from Yale. However, it is currently not possible to find out exactly how much carbon is emitted with each AI command, platform or series.

Dr. S Faizi, an ecologist, United Nations environment negotiator and winner of the 2024 Planet Earth Awards, emphasized the urgent need for public attention to this issue.

“The amount of water that data centers consume is alarming and needs immediate public attention. When you factor in the water and carbon costs of manufacturing the computers and systems used in these centers, along with the long-term electronic waste they generate, the environmental impact is staggering,” he said.

Did you know that there are approximately 9,000 to 11,000 cloud data centers worldwide, with more being built? These centers use a large amount of electricity. According to the International Energy Agency, these data centers could use as much as 1,000 terawatts of electricity by 2026, which is about the same as Japan uses now.

Another report, by the International Electrotechnical Commission, says that by 2027 the AI ​​industry could use as many resources and as much energy as a country the size of the Netherlands!

Despite all this talk, we don’t actually have much specific data on how much AI is affecting the environment when it comes to water, carbon emissions, or electricity. Because there are no rules or standards, most tech companies don’t share exact information about how much energy their AI uses. This makes it difficult to come up with solid plans to manage AI’s energy consumption.

“We don’t know yet as there are no comprehensive studies or analyzes to refer to. But my intuition would be that the efficiency gains and environmental costs basically cancel each other out now. But this could change dramatically in either direction in the future. The challenge is to ensure that efficiency gains improve rapidly while the associated environmental footprint decreases just as rapidly,” said Tamanini.

The dawn of sustainable AI

In 2024, there is a growing focus on making AI more sustainable. Efforts are underway to understand and reduce its energy impact.

Tamanini emphasized the importance of breaking down AI-related emissions into different categories. This transparency can help us better understand AI’s environmental impact and find ways to manage it.

The IEC and ISO collaborate to release the world’s first report on sustainability standards for AI. This report will cover various environmental aspects of AI, such as energy and water consumption, carbon footprint, waste, life cycle and supply chain. The goal is to make these aspects of AI visible to users so they can make informed decisions.

Dr. However, Faizi believes that current efforts may not be enough to prevent environmental damage from AI. He sees this as part of a larger tendency to overstep nature’s limits, which could lead to the extinction of industrial civilization.

“I don’t think there will be a return to AI technology, no return to the capitalist mode of extraction, production and consumption that has worsened to become induced consumption, therefore the environmental crisis we are in will only deepen in the coming years. The efforts to avert it are too modest for the challenge before us.”

As for making the technology more eco-friendly, Tamanini went on to say, “Systems creators can follow best practices to make AI more sustainable. One example is the Google “4M approach. But where AI stands in the sustainable future of technology will depends on regulators and the extent to which system creators are required to report and limit their AI environmental footprint.”

Closure

To conclude with a more recognizable example, India recently in a cabinet meeting gave a plan exceeding Rs 10,000 crore for the next five years. This plan involves the government setting aside funds to help private companies that want to establish AI computing capacity in the country.

Shifting gears, the World Bank highlighted that India faces significant water stress, although it is home to 18% of the world’s population, but has access to only 4% of the world’s water resources.