Sure, AI is super helpful. But it’s draining Nature’s resources to get you that answer.

AI makes it so easy to ask anything. It’s a powerful, ubiquitous tool that helps us be more efficient, and in some cases, completes tasks better than we can. We get it: having answers at your fingertips within seconds is a heady feeling. Your blinking cursor is eager – practically begging – to help you.

AI is undeniably useful. But what is the price of this power and efficiency?

On the surface, AI feels harmless. After all, searching on the internet doesn’t require a credit card or any significant amount of skill. It’s all in the conceptual, nebulous cloud, right? But the truth is: everything has a carbon footprint…and AI’s footprint is gargantuan.

A carbon footprint is the total amount of greenhouse gas emissions an activity releases into the environment. These gases – mainly carbon dioxide – trap heat in the Earth’s atmosphere which fuels climate change over time. Anything that requires power – from driving a car to the foods we eat to the photos and emails we send – has a carbon footprint.

According to one estimate, global AI-related infrastructure may soon require over six times the water requirements of Denmark and its 6-million inhabitants.  This is problematic, especially considering that nearly 2 billion people do not have access to clean water.  The technology that powers AI, therefore, adds stress to an already worrying global water crisis.

The Vampiric Energy Suck of Data Centers 

But water consumption is only one piece of AI’s environmental toll. The amount of electricity required to build, train, and fuel the computational power of generative AI models with billions of parameters creates pressure on existing electrical grids and contributes to the overall carbon emissions of the technology. Training an AI model takes a massive amount of energy, as the program continually runs for weeks and months at a time until it’s built. Even after being trained, the AI program is still using energy to improve performance and source answers for billions of user queries.

Noman Bashir, a fellow with the MIT Climate and Sustainability Consortium, states: “The demand for new data centers cannot be met in a sustainable way. The pace at which companies are building new data centers means the bulk of the electricity to power them must come from fossil fuel-based power plants.”

Source: These Four Charts Sum Up the State of AI and Energy, MIT Technology Review

Polluting Hardware Production

There’s also an environmental implication to producing the electronics and microchips required to power AI.  Obtaining raw materials like lithium, cobalt, and other rare earth elements can involve dirty mining procedures and toxic chemicals.  This extraction process can lead to habitat destruction, water pollution, and soil degradation.

And when those electronics are eventually discarded, that waste is difficult to recycle and likely to contain hazardous substances, like mercury and lead, that can lead to soil and water contamination. More stringent waste laws and ethical disposal practices are necessary to reduce the negative environmental effects of e-waste.

That being said: AI is – ironically – well equipped to help solve environmental issues.

AI has the capacity to synthesize data faster to help us tackle environmental and humanitarian emergencies across the globe.  One of the biggest benefits of AI technology is that it can identify anomalies and similarities in data patterns, and use that data along with historical learnings to predict future outcomes. When used to monitor our planet, AI can help governments, businesses, and individuals make more informed climate decisions. 

UNEP, for example, uses AI to detect and measure methane emissions from oil and gas installations.  Scientists use AI to forecast solar and wind energy outputs, and to optimize temperature control in buildings. AI even helps forest rangers detect the early markers of wildfires with the help of solar-powered gas sensors.  These advancements allow humans to react more immediately to changing and extreme weather conditions, and to synthesize data more effectively to combat climate change.

Such technological advances give us hope that AI could be a key tool in solving the climate crisis, restoring nature and biodiversity, and mitigating our waste and pollution problem.

Creating emissions guardrails and AI best practices are essential for our future.

It seems unlikely that we will go back to a world without AI. This is why it’s essential for governments, businesses, and communities to strive towards more ethical, environmentally-friendly AI practices. Greater environmental transparency and accountability is essential to ensuring that AI is used responsibly.

The UNEP recommends five ways the world could lessen AI’s toll on the planet:

1. 1. Countries can create standardized procedures to globally measure AI’s environmental impact.
   2. Governments can establish regulations that require companies to disclose the environmental impact of their AI-based products and services.
   3. Tech companies can reduce AI energy demands by making algorithms more efficient, and recycling water and electronic components where possible.
   4. Countries can incentivize companies to make their data centers more environmentally friendly through renewable energy and offsetting carbon emissions.
   5. Countries can modify their current environmental regulations to include AI-related policies.

In the meantime, we should be mindful of how we use AI, and understand its environmental consequences. By using AI with purpose and thoughtfulness, we can harness AI’s power for good, and drive a more sustainable future forward. 

AI’s energy requirements exacerbate existing demands for Earth’s natural resources. 

Earth Overshoot Day is an estimation of the day in which humans exceed their annual ecological budget. Each year, this date arrives earlier and earlier.  It’s not too late to #MoveTheDate.  Learn more about Earth Overshoot Day and the ecological deficit of the United States.