Climate change has become a substantial threat to the earth. It affects the most disadvantaged populations and vulnerable environments unevenly. One of the major drivers of climate change is carbon dioxide (CO2), leading heat to trap in the atmosphere, which contributes to the greenhouse effect. In addition to CO2 emissions arising from environmental sources, there are also other main drivers caused by human activities such as heating, production operations, and transportation (Dodge et al., 2022). It can be claimed that AI, which touches each and every aspect of human life and alters it irreversibly, has a complex and dual impact on this subject. As it happens in almost every new technology, it comes with its advantages and disadvantages.
AI’s Carbon Footprint
When using AI systems, we usually concentrate on outcomes without considering numerous expenses of training, ignoring their effect on our planet largely unexplored.
Large language models (LLMs), a subset of AI, are developed to estimate the rest of a given text by analyzing an extensive amount of data derived from many different sources like the internet, books, articles, blogs, and social media. To achieve this, a comprehensive training procedure enabling models to comprehend particular elements and associations in the data must be carried out. These learned associations are kept as parameters.
Each new AI model encompasses more parameters than the preceding model, allowing them to produce more advanced and accurate outcomes. For example, the GPT-2 model introduced in 2019 had 1.5 billion parameters, whereas GPT-3 had 175 billion. Even though the number of parameters for GPT-4 is unknown, Google’s LLM PaLM has 540 billion (Cho, 2023).
These models, which have a huge number of parameters and data sets, necessitate long-term training because of their complex structure. This training process largely demands electricity from nonrenewable energy sources such as coal and natural gas. As AI becomes more prevalent in our daily lives, the need for it grows, as does the environmental impact of its development (Wu et al., 2022).
So, the power consumption of AI might be the most prominent environmental impact of it. The global energy consumption of AI was predicted to be 4.5 gigawatts in 2023, making up % of energy consumption of data centers. Moreover, it is estimated that, by 2028, this number will be three times bigger, generating 20% of the data center’s energy use (Alves, 2024).
AI’s Contribution to Sustainability
The transformative power of AI creates a huge potential for a number of areas like climate change mitigation and sustainability. One prevalent study on this subject is Google’s DeepMind AI, developed by its research team, which decreased the energy needed by data center cooling systems by 40%. This rate represents a 15% reduction in total energy expenditures. The system improves the effectiveness of cooling systems by analyzing data from sensors in information systems (Evans & Gao, 2016).
Similarly, AI is capable of predicting the availability of renewable energy which allows greater incorporation of both solar and wind energy into the power system. This enables improving total energy consumption. Additionally, AI models have a huge breakthrough both in reducing the total amount of carbon consumed in data centers and emissions generated by operations. These improvements demonstrate how AI contributes to usage of renewable energy while decreasing its environmental impact (Wu et al., 2022).
Furthermore, AI contributes to foster sustainability by assessing the quantity of carbon in building materials over their whole life cycle. Machine learning detects materials with high carbon content, improving material selections. Plus, it supports ecologically friendly and sustainable construction by establishing a global standard for carbon assessment procedures. As a result, it contributes to the sector’s lower carbon impact (El Hafdaoui et al., 2023).
Takeaways
In sum, the impact of AI on sustainability is like two sides of the same coin. Therefore, it cannot be said that the impact is one-dimensional; it is rather controversial. On the one hand, long-term AI model training increases carbon emissions, but it also has great potential for providing varied predictions to minimize carbon emissions. By prioritizing responsible use and innovative solutions, AI can evolve into a powerful tool for combating climate change and promoting a more sustainable future.
References
Alves, B. (2024, November 22). Topic: Environmental impact of AI. Statista. https://www.statista.com/topics/12959/environmental-impact-of-ai/#topicOverview
Cho, R. (2023, June 9). Ai’s growing carbon footprint. State of the Planet. https://news.climate.columbia.edu/2023/06/09/ais-growing-carbon-footprint/?utm_source=chatgpt.com
Dodge, J., Prewitt, T., Tachet des Combes, R., Odmark, E., Schwartz, R., Strubell, E., Luccioni, A. S., Smith, N. A., DeCario, N., & Buchanan, W. (2022). Measuring the carbon intensity of AI in cloud instances. 2022 ACM Conference on Fairness, Accountability, and Transparency, 1877–1894. https://doi.org/10.1145/3531146.3533234
El Hafdaoui, H., Khallaayoun, A., Bouarfa, I. et al. Machine learning for embodied carbon life cycle assessment of buildings. J. Umm Al-Qura Univ. Eng.Archit. 14, 188–200 (2023). https://doi.org/10.1007/s43995-023-00028-y
Evans, R., & Gao, J. (2016, July 20). DeepMind AI reduces energy used for cooling Google data centers by 40%. Google. https://blog.google/outreach-initiatives/environment/deepmind-ai-reduces-energy-used-for/
Wu, C. J., Raghavendra, R., Gupta, U., Acun, B., Ardalani, N., Maeng, K., … & Hazelwood, K. (2022). Sustainable ai: Environmental implications, challenges and opportunities. Proceedings of Machine Learning and Systems, 4, 795-813.
