In your regular daily life you might be using AI and not even know it.
You ask your smart speaker to play music, your phone to predict the next word in your text, or you rely on navigation to avoid traffic.
These amazing quality-of-life conveniences, powered by artificial intelligence (AI), are now such a big part of our everyday lives that it almost seems impossible to remember a time without them.
But did you know every AI convenience requires energy to use? And as you know, that energy must come from somewhere whether fossil fuels or renewable resources.
Now the debate seems to be over whether or not AI is bad for the environment.
I’ve found it’s complex.
On one hand, AI can lead to significant environmental costs due to its high energy consumption and resource use.
On the other hand, AI has the potential to benefit the environment through innovations in energy efficiency, resource management, and climate change mitigation.
So which side wins out? The costs or the benefits? We are going to explore each side of the argument on whether AI is bad for the environment.
The Environmental Cost of AI
Energy Consumption
Before AI can be used it must be trained on data. Training AI models, especially large language models (LLMs) like ChatGPT, requires thousands of GPU hours, consuming a serious amount of electricity. Sources estimate that training GPT-4 required around 1,750 MWh of energy, enough to power 175 American households for an entire year.
Once deployed, AI systems continue to consume energy through inference, which is when the AI model processes user requests to perform tasks. This constant use of AI adds to the daily energy consumption of data centers, especially because they run 24/7.
Deep learning algorithms, which are at the heart of many AI systems, involve complex mathematical operations on large datasets. Each of these operations requires a lot of computational power, leading to high energy use.
A single query to ChatGPT can consume up to ten times more energy than a traditional Google search, and a single AI image generation can equal the energy used to fully charge a smartphone.
And as AI as a whole grows in popularity, the energy demand will only increase.
Carbon Footprint
The energy needed for the daily use of AI adds to humanity’s overall carbon footprint, due to 60% of energy used in the United States coming from fossil fuels.
According to Columbia University, AI data centers already account for about 2.5 to 3.7 percent of global greenhouse gas emissions.
Estimates suggest that training a single AI model like ChatGPT can emit over 626,000 pounds of carbon dioxide, the lifetime emissions of roughly five average cars.
Resource Use
The specialized chips and processors needed for AI, like GPUs or TPUs (Tensor Processing Units), require rare earth metals and other materials.
The lifecycle of AI hardware starts with mining, which has an environmental footprint due to land use, energy consumption, and pollution. Manufacturing the hardware also contributes to greenhouse gas emissions.
Data centers hosting AI require servers that are both energy-intensive and resource-heavy. The energy these servers use generate a ton of heat, and need to be properly cooled in order not to overheat and shut down. This is done using water-based cooling, which can potentially cause water scarcity due to the amount needed.
At the end of their life cycle, if not properly recycled, AI hardware can contribute to electronic waste because they contain hazardous materials like lead, mercury, and other heavy metals that harm the environment. Recycling rates for AI hardware are low due to cost, lack of infrastructure, and complexity.
Potential Environmental Benefits of AI
Energy Efficiency
AI can optimize the operation of smart grids by predicting energy demand and supply in real-time, balancing loads to prevent outages, and integrating renewable energy sources more effectively. This means reduced waste from overproduction and more efficient use of generated energy, potentially lowering greenhouse gas emissions.
AI-driven HVAC systems can learn from the usage patterns in a building to adjust temperature settings automatically, ensuring energy is not wasted on heating or cooling unoccupied spaces. Not only will this lower the carbon footprint of the building, but also save on energy costs. Win-win!
Smart home devices powered by AI can learn from people’s behavior to manage energy consumption more effectively, from turning off lights in unused rooms to optimizing the use of solar panels.
Climate Change Mitigation
AI allows us to model complex climate systems, offering us more accurate predictions about climate change impacts, including sea-level rise, temperature changes, and
extreme weather events.
Since AI algorithms can analyze a ton of data quickly, we can predict natural disasters like hurricanes, floods, or wildfires long before they happen, allowing for earlier warnings and better preparation, which can save lives and reduce property damage.
AI is also being used to research carbon capture and storage (CCS) technologies by studying chemical reactions, optimizing processes to bring costs down, and even discovering new materials that could be better at capturing carbon dioxide from the atmosphere.
Resource Management
AI helps in precision farming by analyzing data from drones, satellites, and sensors to tailor crop care to specific areas of a field, reducing the use of water, fertilizers, and pesticides. This not only increases yield, but also lowers environmental impact by keeping soil healthy and reducing harmful chemicals in our waterways.
AI can predict water usage patterns, detect leaks in water systems, and optimize irrigation schedules. It can help manage reservoir levels or predict when a treatment plant might need maintenance to avoid water wastage. This results in water conservation, which is important for regions facing water scarcity.
AI can improve waste sorting and recycling processes by identifying materials for proper disposal or reuse. AI in smart bins or waste management systems can optimize collection routes, reducing fuel consumption and emissions from garbage trucks.
In some industries, AI can predict product demand to guarantee just-in-time manufacturing, reducing overproduction so they can cut down on waste.
Other Sustainable Practices in the AI Industry
AI can analyze production data to suggest ways to reduce energy consumption, minimize waste, or recycle materials more effectively.
AI can significantly decrease the carbon emissions associated with transportation by perfecting routes, reducing idle times, and improving traffic management.
AI can assist in designing products with a lower environmental footprint, and consider the entire lifecycle from material choice to disposal.
Current Initiatives and Future Directions
Green AI
One goal is to reduce the size of neural networks without badly compromising performance. Methods include pruning, where unnecessary connections in neural networks are removed, or quantization, where model parameters are represented with fewer bits.
When optimal techniques like transfer learning are used, pre-trained models are adapted for new tasks (kind of like recycling), reducing the need for energy-sucking retraining. Another technique called federated learning allows model training across decentralized devices, reducing data transfer and energy use.
New types of hardware are being developed to handle AI computations more efficiently. Chips like Google's TPUs or NVIDIA's latest GPUs are designed specifically for AI tasks, offering better performance using less energy compared to regular chips.
Efforts are already underway to make data centers greener. Major tech companies are committing to power their data centers with renewable energy sources since it’s cheaper.
Regulation and Policy
Newer policies include rules for environmental sustainability to make sure AI systems are developed with an eye toward reducing their ecological impact.
Policies might soon require transparency in how much energy AI systems consume, encouraging more efficient designs with some countries already mandating energy efficiency for tech products including AI hardware.
Global initiatives like the AI for Good Global Summit aim to align AI development with sustainable development goals, with the hope of encouraging policies that address AI's environmental impact.
Innovation in AI for Environmental Good
AI technologies like machine learning will analyze satellite imagery to help
monitor deforestation, track wildlife populations, and detect illegal activities in protected areas.
AI is an important step in the concept of smart cities, where it helps in reducing traffic and emissions through intelligent traffic light systems, and optimizing collection routes and sorting processes for better recycling rates.
AI helps in understanding soil health and climate patterns to suggest the most sustainable farming practices and can predict and detect crop diseases or pests early, reducing the need for as many chemicals.
AI is used to analyze data from sensors in the ocean to monitor pollution, temperature changes, and marine life, to help ocean conservation efforts.
AI optimizes the performance of renewable energy installations, from wind farms to solar panels, increasing their efficiency and reducing reliance on fossil fuels.
AI is being used to find optimal sites for carbon sequestration or to aid in planting forests for carbon capture.
Ethical and Economic Considerations
Ethical Dilemmas
AI's 'black box' nature raises ethical concerns about how decisions are made that impact the environment. There's a call for greater transparency in how AI systems are trained and used, especially when their outcomes affect the Earth we all share.
We also have to think about the long-term impacts of AI on the environment. The immediate benefits of AI might be overshadowed by long-term ecological damage if not managed properly. This means considering future generations who will inherit these systems and their environmental legacies.
It’s possible that future AI systems could even continue messing up the environment if they're biased. For example, if AI is used to optimize urban planning, it might prioritize areas with better data or higher economic value, neglecting marginalized communities or regions where environmental monitoring is less advanced due to economic reasons.
Who should bear the responsibility for mitigating AI's environmental impact? Is it the developers, the companies deploying AI, or the users? There's a moral responsibility to make sure AI contributes positively to environmental sustainability rather than making things worse.
Finally, there's an ethical debate about whether the resources (energy, water, materials) dedicated to AI could be better used elsewhere. Should we continue to invest in AI at its current pace, knowing its environmental costs, or should there be a shift towards more sustainable technologies or practices?
Economic Factors
The development and deployment of AI systems involve huge upfront costs in terms of energy, hardware, and human resources. These are long-term investments, but the environmental cost, like energy consumption and carbon emissions, is happening right now and we can see that.
As it learns, AI can lead to major savings by perfecting certain tasks. These efficiencies can offset initial environmental costs, potentially leading to net positive environmental impacts in the long run.
Companies adopting AI for sustainability could see reduced operational costs. Things like smart building systems can decrease energy bills, or precision agriculture can cut down on fertilizer costs, both would have a positive impact on the environment.
While AI might automate some jobs, it also creates new opportunities in AI development, monitoring, and maintenance, particularly in green tech. The challenge is to ensure these new jobs are accessible and contribute to sustainable economies.
There's an
increasing market demand for sustainable products and services. Businesses that use AI to contribute to lowering their environmental footprint might gain a competitive advantage due to consumer loyalty towards sustainability.
Governments could potentially provide tax breaks or subsidies for companies adopting AI in ways that reduce environmental impact. Regulations could be used to impose fines or restrictions on AI systems with high environmental costs, nudging companies towards more sustainable AI practices.
Some companies have already started investing in research to make AI more environmentally friendly. This includes funding for green AI algorithms, recycling technologies for AI hardware, or renewable energy solutions for data centers.
Investments, regulation, and future developments could lead to breakthroughs that significantly lower or even negate the environmental impact of AI.
Case Studies
Negative Impact Case: Increased Energy Consumption by AI in Data Centers
In 2024, Google reported a significant increase in its greenhouse gas emissions, partly due to the expansion of AI workloads in its data centers. The trend was in and demand for AI in services like Google Search, Google Cloud, and Gemini, led to higher energy consumption.
Google's data centers have seen their energy use rise, with a reported 48% increase in emissions linked directly to AI growth in 2024 because the energy used came from non-renewable resources.
Alongside energy, water consumption for cooling these massive data centers has also increased. Google's data centers in the U.S. alone consumed billions of liters of fresh water for cooling in previous years, and this trend will likely continue with its AI expansion.
They are trying to fix it though, as Google was able to use AI to reduce the energy load of its data centers by optimizing cooling systems, achieving up to a 40% reduction in energy use for cooling.
In this case, AI led to a larger carbon footprint, and excess water usage.
Positive Impact Case: AI-Assisted Mangrove Restoration in Asir, Saudi Arabia
The Asir region in Saudi Arabia was working on a project to restore its dwindling mangrove ecosystems, which are crucial for coastal protection, carbon sequestration, and biodiversity. The project used AI to help the restoration.
Machine learning algorithms were used to analyze satellite imagery for identifying optimal locations for mangrove replanting. These algorithms could predict soil salinity, water depth, and other environmental conditions ideal for mangrove growth.
AI-driven drones monitored the health of mangroves after being planted, detecting early signs of stress or disease, which allowed for people to help the plant before it got worse.
The project led to a big increase in the area of the mangrove, helping local biodiversity, providing a habitat for marine life, and acting as a natural barrier against coastal erosion.
One of the best benefits? Mangroves are excellent carbon sinks, and their restoration means more CO2 is absorbed from the atmosphere.
The use of AI made the whole process more cost-effective and less labor-intensive, allowing for scalable conservation efforts and leading to a successful project.
Conclusion
AI is a double-edged sword for the environment. It uses a lot of energy and resources, which can harm the planet. However, it also has the power to help the environment by making processes more efficient, managing resources better, and fighting climate change.
We all need to think about how we use AI. We can support companies that use AI in environmentally friendly ways, push for greener AI tech, and expect clear information about AI's impact on nature. Governments should also make rules to keep AI green.
What if AI was used only for good, to protect our planet? Imagine AI helping us save energy, predict natural disasters, or clean our air and water. How can we make sure AI is part of a sustainable future for everyone?
References
‘AI’s Growing Carbon Footprint.’
Columbia Climate School, June 9, 2023,
https://news.climate.columbia.edu/2023/06/09/ais-growing-carbon-footprint
‘AI has an environmental problem. Here’s what the world can do about that.’
UN Environment Program, September 21, 2024,
https://www.unep.org/news-and-stories/story/ai-has-environmental-problem-heres-what-world-can-do-about
‘The Green Dilemma: Can AI Fulfil Its Potential Without Harming the Environment?’
Earth.org, July 18, 2023,
https://earth.org/the-green-dilemma-can-ai-fulfil-its-potential-without-harming-the-environment
‘Can We Mitigate AI’s Environmental Impacts?’
Yale School of the Environment, October 10, 2024,
https://environment.yale.edu/news/article/can-we-mitigate-ais-environmental-impacts
