In November 2022, Open AI released ChatGPT’s demo, quickly gaining global popularity as users showcased its capabilities. The success of Open AI’s language model intensified the global technological competition, presenting a significant challenge. The rising demand for digital services parallels the accelerating technological progress. Amidst the pervasive impact of climate change, particularly affecting developing economies like Pakistan, the escalating carbon footprint of Artificial Intelligence raises concerns. While AI can mitigate climate effects through smart grid design and low-emission infrastructure, its substantial energy consumption contributes to environmental issues. It is crucial for the countries to balance the sustainable AI usage to mitigate carbon emissions and achieve global environmental goals.
We are living in the fourth industrial revolution, an era in which technology is advancing at an extraordinary rate, where the digital world is immense within the physical and biological world, and science fiction decides the future of warfare. With increase in technological advancements the demand for generative AI products is also increasing and adds about $280 billion of new software revenue. According to the recent available data, Chat GPT has around 1.7 billion visits monthly. According to recent reports, Open AI is currently valued at $80 billion and the company has raised a total of $11.3 billion in funding over seven rounds so far. Since 2012, the number of users around the world has more than doubled, while global internet traffic has expanded 25-fold. Notably, since 2012, the comprehensive training sessions for AI have consistently escalated in computational intensity, with a doubling frequency every 3.4 months, on average.
The increasing pace of generative AI is also increasing the demand of energy and according to the International Energy Agency, data centres currently account for about 1-1.5 percent of global electricity use. The larger the model is, the more the energy is consumed. According to the Climate Reality Project, 97% of climate scientists concur that human activity is the main driver of climate change. Various researchers at the University of Massachusetts Amherst analysed that the carbon footprint of training a single big language model is equal to around 300,000 kg of carbon dioxide emissions. This is of the order of 125 round-trip flights between New York and Beijing. And the cost of carbon required to train the large machine models is an important part of the problem.
Generative AI models needs to be updated continuously. For instance, Chat GPT was trained on data up to 2021, so it does not know anything happened since then. In order to update the models of Generative AI, the demand of energy will increase and as the main source of energy around the world is fossil fuels, so more fossil fuels will be consumed and ultimately the carbon emission level will increase and it will be difficult to maintain the global net zero level.
AI has a significant impact on the energy sector. More energy is used in training the AI models as well as in data centres. For example, when we compare the energy consumption of a random household and the energy required to train large language models like MegatronLM, an AI model developed by Nvidia. The MegatronLM’s training involved the use of 512 V100 GPUs over nine days, with each GPU consuming approximately 250 watts. This totals to 128,000 watts or 128 kilowatts (kW) for the entire duration of the training. Whereas, a random household consumes an average of 12,000 kilowatt hours (kWh) annually. So, it becomes evident that the energy used in training the AI model surpasses the annual energy consumption of more than two households. This comparison highlights the energy requirements involved in training of advanced AI models.
Moreover, data centres are using a huge amount of energy. For instance, the Weather Company, an entity under IBM, engages in the processing of approximately 400 terabytes of data on a daily basis, a computational endeavour directed toward enhancing the predictive capabilities of its weather forecasting models globally. In a parallel vein, Facebook undertakes the generation of a voluminous 4 petabytes (equivalent to 4,000 terabytes) of data each day. According to the latest estimates, 328.77 million terabytes of data are created each day.
It is essential to implement sustainable practices in AI development and usage, such as optimizing energy efficiency in model training and data centres. Collaborative initiatives between governments, industries, and research institutions can drive the development of eco-friendly AI technologies. Investing in renewable energy sources for AI operations and promoting transparency in carbon footprint reporting are crucial steps.
Hence, it is important to note that the most extensive data centres necessitate power capacities exceeding 100 megawatts, a magnitude sufficient to meet the energy needs of approximately 80,000 households in the USA, according to findings by the Energy Innovation, an organization focused on energy and climate analysis. From an environmental perspective, the collective energy consumption attributed to data centres and AI operations emerges as a considerable concern, invoking apprehensions regarding its ecological ramifications.
The escalating global need for energy arises from the ongoing efforts of nations to foster economic development and maintain societal stability amid the backdrop of climate change. According to projections from the International Atomic Energy Agency (IAEA), there is an anticipated near-doubling of electricity consumption by the year 2050. Paradoxically, despite this trajectory, a significant majority— exceeding 60%— of the world’s electricity generation continues to rely on unabated fossil fuels, which thereby remain a substantial contributor to the prevailing challenge of climate change.
Gerry McGovern, the author of World Wide Waste, writes that we are going through the energy consumption crisis. AI is an energy intensive tool and the greater the demand for AI, the more the use of power. He further stated that it’s not simply the electrical energy to train an AI but also about building the supercomputers and collecting and storing the data.”
With the increasing consumption of energy, the temperature of the planet is becoming warmer and climate change impacts are also worsening day by day. AI has the ability to analyze vast amounts of data so it can be used to mitigate climate change. The World Economic Forum defines AI as “pertains to computer systems endowed with the capability to perceive their surroundings, engage in cognitive processes, acquire knowledge, and execute actions in alignment with both their sensory input and programmed objectives.”
For instance, in India, the integration of AI has significantly increased groundnut yields by 30 percent per hectare for farmers. This improvement is attributed to AI’s provision of crucial information regarding land preparation, optimal fertilizer application, and the selection of appropriate sowing dates.
Conversely, in Norway, AI has played a pivotal role in establishing a versatile and self-regulating electric grid, enhancing its capacity to assimilate a greater proportion of renewable energy sources. Additionally, AI applications have demonstrated notable success in achieving accuracy rates ranging from 89 to 99 percent in the identification of tropical cyclones, weather fronts, and atmospheric rivers. The latter, known for inducing heavy precipitation, poses a challenge for human identification, further underscoring the value of AI assistance. By enhancing the precision of weather forecasts, these AI-driven programs contribute to the safeguarding of individuals, exemplifying the multifaceted benefits AI brings to diverse sectors.
It is essential to implement sustainable practices in AI development and usage, such as optimizing energy efficiency in model training and data centres. Collaborative initiatives between governments, industries, and research institutions can drive the development of eco-friendly AI technologies. Investing in renewable energy sources for AI operations and promoting transparency in carbon footprint reporting are crucial steps. Additionally, ongoing research and innovation should focus on creating AI solutions that actively contribute to environmental conservation and climate change mitigation. Striking a balance between technological advancements and ecological responsibility will pave the way for a sustainable future.
