Dr Sushant Singh
Visiting Senior Research Fellow, IISD
International Scholar in Environmental Sustainability, and
Leading Artificial Intelligence (AI) Expert

The very famous definition of Sustainable Development is "The development that meets the needs of the present without compromising the ability of future generations to meet their own needs." The Environmental Sustainability (ES) is one of the most important pillars of Sustainable Development. The 2030 Agenda for Sustainable Development lists seventeen Sustainable Development Goals (SDGs) that comprise 169 targets.

Almost three decades after the Earth Summit held in Rio de Janeiro, Brazil, where more than 178 countries came together and adopted the Agenda 21for sustainable development, the world is far behind in achieving the SDGs.Based on the sustainable development report 2021,only 37 countries crossed 75%, 117 countries scored between 50% and 75%, and the rest scored below 50% on the sustainability ranking scale.There are various ways to achieve the SDGs that depend on how these goals are perceived and approached.

The 21st century is also known as the artificial intelligence (AI) era, where various tools and techniques of AI have shown promising results in addressing numerous global challenges. Therefore, exploring how various AI tools and techniques can help achieve these SDGs is necessary.

Artificial intelligence trains robots to learn from historical information and actions and respond like human beings. Machine Learning (ML), Deep Learning (DL), and Reinforcement Learning (RL) are three essential pillars of AI. Machine learning, an essential pillar of AI, is a computation technique that learns inherent patterns from historical data and uses that training to predict desired outcomes on unseen data.In contrast,DL emulates howthe human brain functions while processing information, understands the hidden complex patterns,and uses these bits of knowledgefor decision making. Moreover, RL learns how intelligent agents take actions in an environment to maximize their cumulative rewards.

Similar to AI, sustainability also has three central pillars, including society, environment, and economy.All the 17 SDGs can be mapped to the three pillars of sustainability. There is a possibility that some SDGs could be shared by two or all three pillars. Scientists have been applying several AI techniques to address numerous environmental challenges for quite some time. For example, they are developing prediction models for natural hazards (floods, landslides, earthquakes, and forest fires). However, the social and economic components have not been explored more. A handful of studies use various AI techniques to develop socio-economic models of environmental problems, such as natural hazards and groundwater arsenic contamination. Considering the advantages of cutting-edge AI technologies, applying them to various SDGs would be imperative. For example, AI could help develop global prediction models of poverty and the communities likely to experience food scarcity. Moreover, AI can help answer why the problems exist, hotspots of all the SDGs, and how the associated challenges can be resolved.

Considering sustainability is a multidimensional process, to assess and develop meaningful prediction models for each SDGs, data must be available from various sources, including but not limited to surveys, policy reports, sensors, and remote sensing images and videos. Therefore, various ML algorithms and methods would be required to process this information appropriately and find hidden patterns. In this regard, various researchers have used AI techniques to address SDGs. For example, various classification algorithms, including Logistic and Multinomial Regression, Support Vector Machines, Classification and Regression Trees, Random forest, K nearest neighbor (K-nn), and Intra or sub-pixel classification have been used to develop various classification models targeting various SDGs, such as Zero Hunger-SDG2, Clean Water and Sanitation-SDG6,Decent Work and Economic Growth-SDG8,Sustainable Cities and Communities-SDG11,Climate Action-SDG13,Life Below Water-SDG14, and Life on Land-SDG15. Additionally, some deep learning algorithms, such as Convolutional Neural Network,are also used. Likewise, many clustering algorithms, including mixture models, K-means, and agglomerative clustering, have been used to address several SDGs, including Zero Hunger-SDG2, Affordable and Clean Energy-SDG7, Industry, Innovation and Infrastructure-SDG9, Sustainable Cities and Communities-SDG11, Climate Action-SDG13, Life Below Water-SDG14, Life on Land-SDG15, and Partnerships for the Goals-SDG17.Furthermore, various regression methods have also been used in assessing and develop prediction models for SDGs.

For example, Linear regression, Boosted Regression Trees, State Space Models, and Neural Networks have been applied to Zero Hunger-SDG2, Good Health and Well-Being-SDG3, Clean Water and Sanitation-SDG6, Affordable and Clean Energy-SDG7, Industry, Innovation and Infrastructure-SDG9, Sustainable Cities and Communities-SDG11, Climate Action-SDG13, Life Below Water-SDG14, and Life on Land-SDG15.

As mentioned above, data from multiple sources requires effectively assessing and developing models for various SDGs. Therefore, to reduce the noise and dimensions from these data sources, various dimension reducing algorithms, including Principal Component Analysis and Functional Data Analysis, have been used, and they are applied to address multiple SDGs, including Zero Hunger-SGD2, Clean Water and Sanitation-SGD6, Affordable and Clean Energy-SDG-7, Industry, Innovation and Infrastructure-SDG9, Sustainable Cities and Communities-SGD11, Climate Action-SDG13, and Life on Land-SGD15.

Therefore, it is an appreciating effort by global scientific communities that they have started marrying the two critical domains, i.e., sustainability and AI. While SDG indexing helps assess how far or how close a country is to achieve its SDGs, it is crucial to understand how strengthening individual SDGs can improve the SDG index, locally and globally. By using Boosted Regression Trees Model, an ensemble of regression trees and boosting, a group of scientists has concluded that three SDGs, including "Good health and well-being", "Quality education" and "Affordable and clean energy" are the utmost collaborative goals.

While it is essential to know how AI could positively help achieve the 17 SDGs, it is also necessary to evaluate whether it could hinder. In this regard, a group of scientists has evaluated, how AI can help accomplish the 169 targets of the 17 SDGs. They found that AI can help achieve 128 targets but may inhibit the rest of them. They estimated that AI could benefit 82% of the societal goals, 93% of the environmental goals, and 70% of the economic goals of the SDGs.However, it is too early to conclude this as it requires rigorous research.

It is a known fact that various countries adopt the SDGs and all have their SDGs defined at a country level. However, to achieve a country's SDGs, their states define various activities to achieve these goals. Therefore, like strategies to achieve SDGs, AI strategies at a country-level is utmost importance. Therefore, the countries that have officially adopted the SDGs should have their AI strategies and policies aligned to achieving the 17 SDGs.

In this regard, Canada's journey of developing National and regional level AI strategies started in 2017. Since then, 32 countries have published their National AI strategies, six countries have already started developing their National AI strategy, and 16 have announced their National AI strategies. The progression of the adoption of the National AI strategy by many countries is highly appreciable.

With AI strategies to target SDGs, it is imperative to allocate funds to various SDGs at regional and global levels. Here, again AI is playing a pivotal role in predicting which SDGs are frequently funded. It is interesting to know that the SDGs, including Good Health and Well-Being-SDG3, Affordable and Clean Energy- SDG7, Industry Innovation, and Infrastructure-SDG9, Zero Hunger-SDG2, and Sustainable Cities and Communities-SDG11 are prominent in receiving grants and aids. Additionally, new methods and algorithms may be the demand of the hour as the ever-growing amount of data is diverse, and various SDGs are also different from each other. However, they do not have clear-cut boundaries, so there may be overlapping the components across many SDGs. Therefore, these unique challenges must be addressed appropriately by developing new computational and analytical methodologies.

Although integrating AI and sustainability may sound very promising, it may bring several challenges while implementing them on the ground. For example, maintaining the data privacy of the collected data, risk of cyberattacks, and lack of readiness of the beneficiaries of AI-based solutions, especially in developing or underdeveloped countries, would be significant challenges.

Since AI is led by several big, private organizations, involving them in social-centric SDGs at a local level may become challengingas these enterprises may push a techno-centric approach, which may not be in the best welfares of the local communities. The beneficiaries may become dependent on these enterprises and may not develop a sense of ownership of the solution, eventually breaking the sustainability chain. Additionally, while digitalization of data hidden in documents and policy reports would be necessary to assess the SDGs better, it may disrupt the local and regional systems and make them highly vulnerable to privacy infringements, stolen data, and malicioushacking.

Marrying the SDGs and AI may not be easy, but it definitely would bring many sustainable outcomes. Like the environmental component of the sustainability pillar, sustainability and AI scientists should focus on the social and economic components. Sustainability scientists and professionals should get trained on various AI tools and techniques, and AI scientists and professions should acquire in-depth knowledge of sustainability. Achieving the SDGs is a daunting task; however, AI could play a critical role in accomplishing this. Moreover, undoubtedly high computation machines would be required to process data from various sources, new technologies and platforms would be must to store them securely, and new tools and technologies will be needed to analyze them.

Above all, a responsible society must be created where people practice responsible AI and environmental sustainability.

____________________________________________

The Author is a Visiting Senior Research Fellow and heads Emerging Technology (AI, ML, IoT, Blockchain) Group of Indian Institute of Sustainable Development (IISD), New Delhi; who is an International Scholar in Environmental Sustainability, and Leading Artificial Intelligence (AI) Expert.