csworker09
12 hrs ago
Research Papers / Environmental Impact of Artificial Intelligence [2]
No longer a science fiction dream, Artificial Intelligence is here, and its growth is expected to explode over the next couple of years. Artificial Intelligence technology is suddenly integrated into every industry and every software system from healthcare and education to transportation, finance, education, and entertainment. Integration of Artificial Intelligence is expected to save time and resources and increase efficiency and accuracy in all aspects of life. The general presumption from the consumer is that Artificial Intelligence is a good technology that will benefit mankind. The reality is Artificial Intelligence is not an invisible tool in the cloud. There are real life physical consequences for using Artificial Intelligence and building and maintaining the required data centers. Unfortunately, we do not yet know the full extent of the environmental effects. As consumers of Artificial Intelligence, it is our moral and ethical obligation to understand the consequences of using this tool on the world around us. Artificial Intelligence has concerningly negative environmental effects, including staggering use of electricity, water and rare earth minerals that may damage the environment and endanger humans and other animals. There are currently no clear solutions.
The primary environmental problems associated with Artificial Intelligence stem largely from the explosive growth of data centers and the physical infrastructure required for it to exist. "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 United Nations Environment Programme (UNEP). "We need to make sure the net effect of AI on the planet is positive before we deploy the technology at scale." Unfortunately, the large-scale deployment is already happening without the data necessary to determine what is at risk. The number of data centers has soared from 500,000 in 2012 to 8 million today. (AI has an Environmental Problem) Data centers are massive physical building compounds full of super computers that are required for the use of Artificial Intelligence, but most Artificial Intelligence users will never physically interact with one or even know they exist. One of the biggest problems is that we simply don't have enough data on the environmental consequences of Artificial Intelligence to create viable solutions. The limited information that is available suggests that the situation is already dire, while regulatory legislation and policy lag far behind the pace of technological development.
One of the most significant environmental impacts of Artificial Intelligence is its enormous demand for electricity. Data centers consume vast amounts of energy to power servers, train Artificial Intelligence models, run queries, and maintain operations. It is estimated "that data centers, cryptocurrencies, and artificial intelligence (AI) consumed about 460 TWh of electricity worldwide in 2022, almost 2% of total global electricity demand." (Cam, 31) Electricity demand of data centers in 2026 could roughly double compared to 2022 which is "roughly equivalent to adding at least one Sweden or at most one Germany" (Cam, 31) In the United States alone, data center electricity consumption is expected to grow from approximately 4% of US electricity demand to 6% of total US electricity demand by 2024. (Cam, 32)
Artificial Intelligence is even becoming integrated into preexisting software and programs. For example, "earch tools like Google could see a tenfold increase of their electricity demand in the case of fully implementing Artificial Intelligence in it. When comparing the average electricity demand of a typical Google search (0.3 Wh of electricity) to OpenAI's ChatGPT (2.9 Wh per request), and considering 9 billion searches daily, this would require almost 10 TWh of additional electricity in a year." (Cam, 32) The limited preliminary information and numbers we do have are absolutely staggering. Over one third of Ireland's electricity demand is expected to come from data centers by 2026. (Cam, 32) This is simply not sustainable.
Electricity demand in data centers is mainly from two processes. Electricity is used in the actual computing, but it is also used in the physical cooling of the hardware. It is estimated that while computing accounts for 40% of electricity demand of a data center, cooling by itself is about another 40%. (Cam, 31) As we increase integration of Artificial Intelligence into everything around us, energy demands continue to rise and strain the electrical grid. Further, old technology quickly becomes obsolete, and the energy used in its creation is wasted when progress demands the newest cutting-edge technology. (Zewe)
The more we integrate Artificial Intelligence into everyday technology, everything uses far more energy. "Even a 0.42 Wh short query, when scaled to 700M queries/day, aggregates to annual electricity comparable to 35,000 U.S. homes, evaporative freshwater equal to the annual drinking needs of 1.2 million people, and carbon emissions requiring a Chicago-sized forest to offset." (Jegham) The more integrated Artificial Intelligence becomes into everything people do, the more unavoidable its use becomes and the more strain is placed on the world's energy demands contributing to more fossil fuel dependence and greenhouse gas emissions. The increase in energy demand even exasperates problems with green energy technology such as the production of parts for solar energy, death of animals and destruction of ecosystems with wind energy, and the long-term environmental risks associated with nuclear energy such as water consumption and transportation and safe storage of radioactive waste.
Beyond electricity consumption, artificial intelligence also places extraordinary demands on water resources. While across the world, people lack access to clean and safe water sources, and many more populations across the globe are at risk of inadequate water supplies, Artificial Intelligence is projected to require more than six times as much water as the entire country of Denmark. (Li, p.1) Large volumes of water are imperative for cooling hardware to prevent servers from overheating and maintain optimum operating conditions. While some facilities have begun development of closed systems that reuse the water used to cool the servers, many facilities use evaporative cooling systems. These systems require a continuous supply of fresh water that is ultimately evaporated into the atmosphere. The extreme demands on water use are gravely serious given the concerns regarding adequate water supply around the world. Even in the United States, cities like Phoenix, in the Southwest region, have been suffering from prolonged drought and water scarcity to the extent there is genuine worry if the water supply is sufficient to support the growing population. Yet, there are proposals to build more data centers in the Southwest in places like Chandler, a suburb of Phoenix, and Page, a town on the Colorado River that provides water to Phoenix, due to availability of land and favorable economic laws, without any regard for the impact on the water supply. In addition to the human water supply, desert ecosystems are incredibly delicate and local wildlife and plants rely on scarce water supplies for their survival. Many species are already struggling due to water diversion for human needs.
Another critical environmental consequence of artificial intelligence is that it requires mining rare earth minerals for production of hardware. Artificial intelligence systems rely on physical computer chips and processing units that require mining of rare earth elements such as lithium, cobalt, and neodymium. Mining these materials has significant environmental impacts of its own including fossil fuel emissions from the equipment used to mine the materials, dust from blasting, massive amount of chemical waste from processing the ore, risk of chemical and radioactive spillage and leaching into the environment poisoning soil, water, animals and humans.
To further exasperate these problems, mining activities occur mostly in regions lacking adequate regulation which puts local communities and the environment as a whole at risk. Mining towns are known for polluted water sources, poisoned farmland, and a myriad of health problems, including respiratory diseases and health ailments resulting from chronic radioactive exposure. Various mining techniques typically destroy the physical environment where the mine is located and can wipe out entire wildlife populations as their habitat is destroyed and their food and water supplies are poisoned or completely disrupted. While we may be able to measure the energy and water use of data centers, the environmental consequences of mining are perhaps the most difficult to measure and mitigate.
To make matters worse, chips are quickly outdated as technology develops at warp speed, and in order to remain competitive, companies must create new chips quickly to supply to their customers or risk becoming obsolete as they are outdone by their competitors. "Market research firm TechInsights estimates that the three major producers (NVIDIA, AMD, and Intel) shipped 3.85 million GPUs to data centers in 2023, up from about 2.67 million in 2022. That number is expected to have increased by an even greater percentage in 2024." (Zewe) The continual production of new components and there being no use for the old ones, results in never ending mining activity and increasingly built up toxic electronic waste.
The world is already fighting a losing battle against fossil fuel carbon emissions, water scarcity, pollution, and habitat destruction. Artificial Intelligence accelerates the crisis to a shocking degree. Even "green technology" has been proven inefficient, ineffective and/or to have environmental impacts of its own on the ecosystem and wildlife. While combating carbon emissions, the focus should be on consuming less, not more than needed. Unfortunately, Artificial Intelligence does just the opposite; it necessarily increases the use of resources. Artificial Intelligence is being automatically integrated into daily use where it creates no real advantage. It is quickly becoming a matter of convenience and habit rather than a tool that is selectively and responsibly utilized where absolutely necessary to achieve an advantage.
Ironically, despite the consumer demand for Artificial Intelligence services and the anticipated economic benefits of data centers, communities do not want data centers in their own towns. While building of data centers may bring temporary jobs and increase of economic activity, once the data centers are built, they do not require many jobs to operate and provide very few to no benefits back to the community. Residents are generally concerned about pollution, water use, energy strain, lack of corporate transparency and government regulation, and an overall negative impact on the esthetics of the area. As a result, development of data centers are being proposed in remote and delicate locations like deserts and wetlands that further threaten destruction of ecosystems.
Addressing these challenges will require both immediate and long-term solutions. In the short term, more comprehensive research is needed to accurately measure Artificial Intelligence's environmental impact. Regulatory bodies should immediately require technology companies and data centers to disclose the energy, water, and material costs required by their products and services in an obvious way so consumers can make informed choices and act in a responsible manner. Increased transparency and reliable data would also enable lawmakers to craft policy and legislative measures to minimize the environmental harm already occurring.
In addition, another immediate and continuing solution is for companies to invest in more efficient algorithms and better technology, such as closed cooling systems, that could significantly reduce Artificial Intelligence's impact on the environment.
Public education campaigns may also be useful and effective in discouraging unnecessary or excessive Artificial Intelligence use and encouraging consumers to consider the full environmental impacts when utilizing the technology. Universities and employers should include information in their syllabi and employee handbooks regarding the environmental consequences of Artificial Intelligence. Public libraries and government agencies can provide brochures and other public information packets. Social media, radio and television advertising can help educate the general public and inspire them to seek additional information and work towards creating solutions.
Long-term solutions will require technological innovation, coordinated policy and regulation efforts, public transparency, and corporate accountability. UNEP specifically makes the following five recommendations: 1) establish standardized procedures for measuring the environmental impact of Artificial Intelligence; 2) develop regulations that require Artificial Intelligence companies to disclose the direct environmental consequences; 3) Make Artificial Intelligence algorithms more efficient 4) Encourage data centers that use renewable energy and offsetting carbon emissions; 5) integrate Artificial Intelligence -related policies into broader environmental regulations. Most solutions currently being proposed by experts fall within these categories.
Governments can also offer incentives for companies that power data centers with renewable energy sources rather than fossil fuels. While renewable energy is not without environmental costs, it is still the best option of what is available at this time. Artificial Intelligence itself can be utilized in various ways to at least attempt to mitigate the environmental damage. Yuan Yao at Yale School of the Environment set forth a few ways Artificial Intelligence itself can "enhance energy efficiency and reduce energy usage, and it assists in environmental monitoring and management, such as tracking air emissions. Moreover, Artificial Intelligence supports process and supply chain optimization to minimize environmental impacts." More data is needed to understand if utilizing Artificial Intelligence for environmental conservation is a net positive.
In conclusion, while artificial intelligence offers remarkable benefits and the potential to transform mankind for the better, its environmental consequences cannot and should not be ignored. The massive consumption of electricity and water, combined with the destructive mining of rare earth minerals, poses serious risks to ecosystems and human well-being. Without regulation, transparency, accountability, and sustainable innovation, severe damage to the environment is inevitable. The reality is Artificial Intelligence's growth will likely come with an unacceptable environmental cost. Recognizing and addressing these impacts now ensures a sustainable path forward with minimal environmental consequences.
Works Cited
No longer a science fiction dream, Artificial Intelligence is here, and its growth is expected to explode over the next couple of years. Artificial Intelligence technology is suddenly integrated into every industry and every software system from healthcare and education to transportation, finance, education, and entertainment. Integration of Artificial Intelligence is expected to save time and resources and increase efficiency and accuracy in all aspects of life. The general presumption from the consumer is that Artificial Intelligence is a good technology that will benefit mankind. The reality is Artificial Intelligence is not an invisible tool in the cloud. There are real life physical consequences for using Artificial Intelligence and building and maintaining the required data centers. Unfortunately, we do not yet know the full extent of the environmental effects. As consumers of Artificial Intelligence, it is our moral and ethical obligation to understand the consequences of using this tool on the world around us. Artificial Intelligence has concerningly negative environmental effects, including staggering use of electricity, water and rare earth minerals that may damage the environment and endanger humans and other animals. There are currently no clear solutions.
The primary environmental problems associated with Artificial Intelligence stem largely from the explosive growth of data centers and the physical infrastructure required for it to exist. "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 United Nations Environment Programme (UNEP). "We need to make sure the net effect of AI on the planet is positive before we deploy the technology at scale." Unfortunately, the large-scale deployment is already happening without the data necessary to determine what is at risk. The number of data centers has soared from 500,000 in 2012 to 8 million today. (AI has an Environmental Problem) Data centers are massive physical building compounds full of super computers that are required for the use of Artificial Intelligence, but most Artificial Intelligence users will never physically interact with one or even know they exist. One of the biggest problems is that we simply don't have enough data on the environmental consequences of Artificial Intelligence to create viable solutions. The limited information that is available suggests that the situation is already dire, while regulatory legislation and policy lag far behind the pace of technological development.
One of the most significant environmental impacts of Artificial Intelligence is its enormous demand for electricity. Data centers consume vast amounts of energy to power servers, train Artificial Intelligence models, run queries, and maintain operations. It is estimated "that data centers, cryptocurrencies, and artificial intelligence (AI) consumed about 460 TWh of electricity worldwide in 2022, almost 2% of total global electricity demand." (Cam, 31) Electricity demand of data centers in 2026 could roughly double compared to 2022 which is "roughly equivalent to adding at least one Sweden or at most one Germany" (Cam, 31) In the United States alone, data center electricity consumption is expected to grow from approximately 4% of US electricity demand to 6% of total US electricity demand by 2024. (Cam, 32)
Artificial Intelligence is even becoming integrated into preexisting software and programs. For example, "earch tools like Google could see a tenfold increase of their electricity demand in the case of fully implementing Artificial Intelligence in it. When comparing the average electricity demand of a typical Google search (0.3 Wh of electricity) to OpenAI's ChatGPT (2.9 Wh per request), and considering 9 billion searches daily, this would require almost 10 TWh of additional electricity in a year." (Cam, 32) The limited preliminary information and numbers we do have are absolutely staggering. Over one third of Ireland's electricity demand is expected to come from data centers by 2026. (Cam, 32) This is simply not sustainable.
Electricity demand in data centers is mainly from two processes. Electricity is used in the actual computing, but it is also used in the physical cooling of the hardware. It is estimated that while computing accounts for 40% of electricity demand of a data center, cooling by itself is about another 40%. (Cam, 31) As we increase integration of Artificial Intelligence into everything around us, energy demands continue to rise and strain the electrical grid. Further, old technology quickly becomes obsolete, and the energy used in its creation is wasted when progress demands the newest cutting-edge technology. (Zewe)
The more we integrate Artificial Intelligence into everyday technology, everything uses far more energy. "Even a 0.42 Wh short query, when scaled to 700M queries/day, aggregates to annual electricity comparable to 35,000 U.S. homes, evaporative freshwater equal to the annual drinking needs of 1.2 million people, and carbon emissions requiring a Chicago-sized forest to offset." (Jegham) The more integrated Artificial Intelligence becomes into everything people do, the more unavoidable its use becomes and the more strain is placed on the world's energy demands contributing to more fossil fuel dependence and greenhouse gas emissions. The increase in energy demand even exasperates problems with green energy technology such as the production of parts for solar energy, death of animals and destruction of ecosystems with wind energy, and the long-term environmental risks associated with nuclear energy such as water consumption and transportation and safe storage of radioactive waste.
Beyond electricity consumption, artificial intelligence also places extraordinary demands on water resources. While across the world, people lack access to clean and safe water sources, and many more populations across the globe are at risk of inadequate water supplies, Artificial Intelligence is projected to require more than six times as much water as the entire country of Denmark. (Li, p.1) Large volumes of water are imperative for cooling hardware to prevent servers from overheating and maintain optimum operating conditions. While some facilities have begun development of closed systems that reuse the water used to cool the servers, many facilities use evaporative cooling systems. These systems require a continuous supply of fresh water that is ultimately evaporated into the atmosphere. The extreme demands on water use are gravely serious given the concerns regarding adequate water supply around the world. Even in the United States, cities like Phoenix, in the Southwest region, have been suffering from prolonged drought and water scarcity to the extent there is genuine worry if the water supply is sufficient to support the growing population. Yet, there are proposals to build more data centers in the Southwest in places like Chandler, a suburb of Phoenix, and Page, a town on the Colorado River that provides water to Phoenix, due to availability of land and favorable economic laws, without any regard for the impact on the water supply. In addition to the human water supply, desert ecosystems are incredibly delicate and local wildlife and plants rely on scarce water supplies for their survival. Many species are already struggling due to water diversion for human needs.
Another critical environmental consequence of artificial intelligence is that it requires mining rare earth minerals for production of hardware. Artificial intelligence systems rely on physical computer chips and processing units that require mining of rare earth elements such as lithium, cobalt, and neodymium. Mining these materials has significant environmental impacts of its own including fossil fuel emissions from the equipment used to mine the materials, dust from blasting, massive amount of chemical waste from processing the ore, risk of chemical and radioactive spillage and leaching into the environment poisoning soil, water, animals and humans.
To further exasperate these problems, mining activities occur mostly in regions lacking adequate regulation which puts local communities and the environment as a whole at risk. Mining towns are known for polluted water sources, poisoned farmland, and a myriad of health problems, including respiratory diseases and health ailments resulting from chronic radioactive exposure. Various mining techniques typically destroy the physical environment where the mine is located and can wipe out entire wildlife populations as their habitat is destroyed and their food and water supplies are poisoned or completely disrupted. While we may be able to measure the energy and water use of data centers, the environmental consequences of mining are perhaps the most difficult to measure and mitigate.
To make matters worse, chips are quickly outdated as technology develops at warp speed, and in order to remain competitive, companies must create new chips quickly to supply to their customers or risk becoming obsolete as they are outdone by their competitors. "Market research firm TechInsights estimates that the three major producers (NVIDIA, AMD, and Intel) shipped 3.85 million GPUs to data centers in 2023, up from about 2.67 million in 2022. That number is expected to have increased by an even greater percentage in 2024." (Zewe) The continual production of new components and there being no use for the old ones, results in never ending mining activity and increasingly built up toxic electronic waste.
The world is already fighting a losing battle against fossil fuel carbon emissions, water scarcity, pollution, and habitat destruction. Artificial Intelligence accelerates the crisis to a shocking degree. Even "green technology" has been proven inefficient, ineffective and/or to have environmental impacts of its own on the ecosystem and wildlife. While combating carbon emissions, the focus should be on consuming less, not more than needed. Unfortunately, Artificial Intelligence does just the opposite; it necessarily increases the use of resources. Artificial Intelligence is being automatically integrated into daily use where it creates no real advantage. It is quickly becoming a matter of convenience and habit rather than a tool that is selectively and responsibly utilized where absolutely necessary to achieve an advantage.
Ironically, despite the consumer demand for Artificial Intelligence services and the anticipated economic benefits of data centers, communities do not want data centers in their own towns. While building of data centers may bring temporary jobs and increase of economic activity, once the data centers are built, they do not require many jobs to operate and provide very few to no benefits back to the community. Residents are generally concerned about pollution, water use, energy strain, lack of corporate transparency and government regulation, and an overall negative impact on the esthetics of the area. As a result, development of data centers are being proposed in remote and delicate locations like deserts and wetlands that further threaten destruction of ecosystems.
Addressing these challenges will require both immediate and long-term solutions. In the short term, more comprehensive research is needed to accurately measure Artificial Intelligence's environmental impact. Regulatory bodies should immediately require technology companies and data centers to disclose the energy, water, and material costs required by their products and services in an obvious way so consumers can make informed choices and act in a responsible manner. Increased transparency and reliable data would also enable lawmakers to craft policy and legislative measures to minimize the environmental harm already occurring.
In addition, another immediate and continuing solution is for companies to invest in more efficient algorithms and better technology, such as closed cooling systems, that could significantly reduce Artificial Intelligence's impact on the environment.
Public education campaigns may also be useful and effective in discouraging unnecessary or excessive Artificial Intelligence use and encouraging consumers to consider the full environmental impacts when utilizing the technology. Universities and employers should include information in their syllabi and employee handbooks regarding the environmental consequences of Artificial Intelligence. Public libraries and government agencies can provide brochures and other public information packets. Social media, radio and television advertising can help educate the general public and inspire them to seek additional information and work towards creating solutions.
Long-term solutions will require technological innovation, coordinated policy and regulation efforts, public transparency, and corporate accountability. UNEP specifically makes the following five recommendations: 1) establish standardized procedures for measuring the environmental impact of Artificial Intelligence; 2) develop regulations that require Artificial Intelligence companies to disclose the direct environmental consequences; 3) Make Artificial Intelligence algorithms more efficient 4) Encourage data centers that use renewable energy and offsetting carbon emissions; 5) integrate Artificial Intelligence -related policies into broader environmental regulations. Most solutions currently being proposed by experts fall within these categories.
Governments can also offer incentives for companies that power data centers with renewable energy sources rather than fossil fuels. While renewable energy is not without environmental costs, it is still the best option of what is available at this time. Artificial Intelligence itself can be utilized in various ways to at least attempt to mitigate the environmental damage. Yuan Yao at Yale School of the Environment set forth a few ways Artificial Intelligence itself can "enhance energy efficiency and reduce energy usage, and it assists in environmental monitoring and management, such as tracking air emissions. Moreover, Artificial Intelligence supports process and supply chain optimization to minimize environmental impacts." More data is needed to understand if utilizing Artificial Intelligence for environmental conservation is a net positive.
In conclusion, while artificial intelligence offers remarkable benefits and the potential to transform mankind for the better, its environmental consequences cannot and should not be ignored. The massive consumption of electricity and water, combined with the destructive mining of rare earth minerals, poses serious risks to ecosystems and human well-being. Without regulation, transparency, accountability, and sustainable innovation, severe damage to the environment is inevitable. The reality is Artificial Intelligence's growth will likely come with an unacceptable environmental cost. Recognizing and addressing these impacts now ensures a sustainable path forward with minimal environmental consequences.
Works Cited
