The Public Interest at the Intersection of AI and Climate Change
There is much debate about AI being a bane or a boon for human society. Tech evangelists feel it will solve all human issues and lead us to utopia whereas naysayers predict it will reduce humans to mere puppets and eventually lead to disaster.
There is much less debate about climate change, as a consensus is building that it is a one-way stream to an unsustainable planet if greenhouse gases continue to build up.
AI and climate change are among the most important megatrends today that will increasingly intersect with uncertain but likely significant consequences for both society and the environment.
There is little doubt that AI uses large amounts of fossil fuel-dependent electricity. AI requires large scale computation, storage and transfer of data, which besides consuming energy generates excessive heat that needs to be evacuated through air conditioning.
This use of energy, through its greenhouse gas emissions, is causing the globe to warm and the climate to deteriorate. Each time you ask an LLM (Large Language Model) a question, it could take up to 0.42 Wh of electricity, approximately. This is like running a 10 W electric bulb for 2.5 minutes (Energy (Wh) = Power (W) x Time (Hours)). Considering the average carbon intensity of electricity as 0.4 kg of CO2/kWh, each query generates approximately 1.68 grams of carbon dioxide (CO2 Emission = Energy used (KWh) x Carbon intensity of electricity). In addition to responding to just queries, there is significant energy used and carbon dioxide emitted for training these AI models, running other sundry infrastructure operations and for safe disposal of the equipment, which when amortized could add an estimated similar amount of carbon dioxide per query taking the average emissions per query to say approximately 3.36 grams of carbon dioxide.
Considering ChatGPT alone handles approximately 2.5 billion queries a day, it has the potential to add roughly 8,400 tons of carbon dioxide into the atmosphere every day. That works out to 3.06 million tons per year. Considering that the number of queries is increasing every day and its model needs to be trained every so often, the carbon dioxide emissions from ChatGPT are likely to keep spiraling upwards. And ChatGPT is not alone. There is Google’s Gemini, Anthropic AI’s Claude, Perplexity, X.ai and several other Generative AI platforms that are constantly pumping carbon dioxide into the atmosphere and causing our planet to warm.
LLMs are just one application of AI. There are many others. It is used for recognizing patterns, detecting anomalies, making predictions, controlling robots, conducting research, and even for managing your city traffic lights, besides others. Considering how the use of AI is proliferating, the amount of emissions and their climate deteriorating potential remain a major concern. The energy and associated emissions rise exponentially as we go up the AI complexity ladder, from simple AI to Machine Learning to Deep Learning and Computer Vision. Each successively more pernicious in its emissions. We have not even considered Artificial General Intelligence (AGI) as yet, which could go unbridled in its always-on computations and continue to consume energy till it finds its perfect answer.
AI can be used to optimize industries and processes that are environmentally destructive. Consider AI being used by the oil and gas industry to optimize oil exploration and extraction. AI could be used by illegal loggers to figure out pockets for felling the best trees in a forest, under minimal surveillance. AI could be used to identify, target and blow up fossil fuel energy infrastructure of nations by hostile countries during wars, thereby rapidly filling the air with greenhouse gases.
Of course, the world will continue to use AI, and its use will only grow, when on the other hand our urgent need is to contain climate change.
The relationship between AI and climate is best explained by the age-old saying: ‘Poison and pill in one’. The solution to a problem sometimes lies in the problem itself.
AI seems unstoppable now, so the challenge is how do we harness it to help climate. There are a handful of ways in which AI is coming to our rescue and helping delay the climate crisis.
AI is being used for energy transition towards renewables. It is being used to develop clean power by selecting optimal sites for solar, wind, geothermal, and ocean-tide-based renewable energy generation. It helps optimize these power plants by aligning the direction of solar cells to that of the sun. By adjusting the windmill gear for prevalent wind speeds or by synchronizing the direction of the flaps being used to harness the force of ocean tides. It helps maintain these power systems by predicting faults and undertaking preventive maintenance. AI is helping manage the complex grid by matching supply, demand and storage of power. The result is more power with lower emissions.
Energy efficiency is another area where AI helps reduce the consumption of power and therefore emissions. It keeps a hawk’s eye on the use of energy within an organization - switching off systems, adjusting the load and often suggesting which equipment is ready for replacement to save energy. It is helping scientists in their research to develop more energy-efficient materials and engineers to design less energy-consuming machines and systems.
AI helps us understand climate science better. It is being used to develop robust climate models that can predict climate risks and help us develop effective strategies for countering such eventualities. The regular scourge of forest fires has inspired the development of AI models to predict the paths wildfires could take. This would help firefighters prioritize evacuation and also develop containment strategies to minimize human loss. Google in partnership with Indian Central Water Commission has developed an AI based flood forecasting model that analyses weather data, river levels, terrain and built infrastructure data to predict where floods may occur, where water may flow or where inundation may happen. During the 2020 monsoons, they sent out 30 million mobile messages to people informing them about the impending threat and thus saving many lives.
AI is facilitating research in various fields that can help combat climate change. New material discoveries that use less energy. Agricultural techniques that can help reduce its carbon footprint. Study of ocean chemistry and plants biology such that they can absorb more carbon dioxide. Our scientists, aeronautical engineers and software developers are using AI to research and develop new flight paths to mitigate the warming potential of contrails during a flight.
Methane is eighty times more potent than carbon dioxide in terms of its global warming potential and is prone to unpredicted leaks from oil and gas fields, landfills, natural geological formations and other sources. AI is helping detect these super-emitter events so that the methane leaks can be sealed in time. AI is mirroring the complex chemical and biological reactions in the stomachs of cattle that end up producing methane and matching them with external food additives that can help inhibit these methane generating microbes.
There is another indirect way in which AI is serving as a boon for climate.
AI is driving Hyperscalers like Google, Amazon, Microsoft and Oracle to rapidly build datacenters, which need significant amounts of energy. AI is flush with capital and having committed to use clean energy, these Hyperscalers are the harbingers for deployment of innovative clean energy technologies such as nuclear fission based Small Modular Reactors (SMRs), Hydrogen Fuel Cell based microgrids, Battery Energy Storage Systems, Enhanced Geothermal Energy Systems and others. Besides, these Hyperscalers are investing large amounts of risk capital, helping set up demonstration projects and giving advance purchase commitments to startups working on cutting edge clean technologies. These initiatives are helping maturation of clean technologies, which will ultimately help us accelerate utility scale energy transition. While AI is responsible for just 1.5 percent of electricity consumption, the impetus it is providing towards clean energy transition will hopefully help bend the global emissions curve, sooner than if left to conventional progress.
As they say: ‘Every rose has its thorn’. Our challenge is to avoid the thorns and instead use AI to reduce the scourge of climate change.
I am reminded of an old Indian story about the origin of its holy river - the Ganges, which today serves as a lifeline for over 400 million people. Legend has it that Ganges, a celestial river, called upon earth to cleanse the souls of King Sagara’s 60,000 sons who had been cursed by sage Kapila for disturbing him in his meditation. The Ganges was a ferocious river. While it cleansed the souls of the cursed princes, its mighty flow created havoc and caused mayhem in the region. Lord Shiva was called to step in. He is said to have wrapped the Ganges in his matted hair and slowed it down, thus avoiding its destructive potential. That’s how we got the best of the holy Ganges. Because Lord Shiva regulated it.
AI needs to be regulated similarly, so that the world gets the best of it.
Just like the Ganges, AI’s power can be harnessed for reining-in climate change.
Policy initiatives to mandate AI infrastructure and service providers to commit to a net zero carbon footprint in their Scope 1, 2 and 3 emissions and include audited performance in their annual reports would be a good starting point. The European Union has made a start in this direction through its EU AI Act. Other nations need to follow. We must keep the Goldilocks principle in mind though; policy should not be too severe as to throw the baby out with the bathwater. Policy should encourage innovation in AI that helps solve the climate problem.
About the Author:
Rajan Mehta is a serial technology entrepreneur who held senior positions at Motorola and Nortel. In 2022 he was a fellow at Harvard’s Advanced Leadership Initiative (ALI) where he took a deep interdisciplinary dive into the science and politics of climate change, culminating in his approachable best-selling book, Backstage Climate: The Science and Politics of Climate Change.