Artificial Intelligence for Sustainable Development: A Structured Critical Review of Applications, Benefits, Risks and Governance Priorities
This structured critical review examines the dual role of artificial intelligence in advancing global sustainability through diverse applications while simultaneously addressing its ecological footprint and ethical risks, ultimately arguing that achieving sustainable development requires balancing technological capabilities with effective governance and responsible management of societal and environmental consequences.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
The world faces a dual reality: the rapid acceleration of artificial intelligence, a technology capable of learning and solving problems without human instruction, and the urgent need to protect the planet from climate change, resource depletion, and social inequality. To navigate this, global leaders have agreed on a shared roadmap called the Sustainable Development Goals, a set of seventeen targets designed to end poverty, protect the environment, and ensure prosperity for everyone. At the heart of this challenge is a simple but difficult question: can the very technology driving our modern economy also be the tool that saves it? The concept of "Green AI" has emerged to describe efforts to make these intelligent systems themselves more efficient, while "sustainable development" remains the broader vision of meeting today's needs without compromising the ability of future generations to meet theirs. The tension between using powerful computers to solve environmental crises and the massive energy those computers consume creates a complex puzzle that researchers are only just beginning to untangle.
In a comprehensive review of existing scientific literature, researchers Stelios Stefanos Drousiotis, Giorgos Nicou, and Marios I. Katsioloudes set out to examine exactly how artificial intelligence is being used to support these global goals. They did not run new computer simulations or collect fresh data; instead, they gathered and analyzed 138 different studies and reports published since 2018. Their goal was to look past the hype and determine where artificial intelligence is genuinely helping the planet, where it is creating new problems, and what the real-world evidence actually says. The team focused on six key areas where technology is applied: managing energy grids, farming, monitoring the environment, building smart cities, recycling materials, and running factories. By sorting through hundreds of pages of research, they aimed to separate genuine environmental progress from mere technological improvements that look good on a screen but do not necessarily help the earth.
The review found that artificial intelligence is indeed a powerful tool for efficiency. In the energy sector, smart algorithms are helping to balance electricity supply and demand in real time, which allows for more renewable energy like wind and solar to be used without wasting power. In agriculture, these systems help farmers use water and fertilizer more precisely, reducing runoff and chemical use while predicting crop diseases before they spread. In cities, AI is optimizing traffic lights to reduce congestion and emissions, and in factories, it is helping to predict when machines will break, preventing waste and saving energy. The technology is also proving vital for environmental monitoring, using satellite images to track deforestation, predict floods, and even identify illegal poaching activities. These applications directly support the global goals for clean energy, zero hunger, and climate action.
However, the researchers uncovered a significant contradiction they call the "AI Green Paradox." While artificial intelligence can help solve environmental problems, the process of building and running these systems creates its own heavy environmental burden. Training the most advanced AI models requires enormous amounts of electricity, often generated by fossil fuels, and the massive data centers that house them consume vast quantities of water for cooling. Furthermore, the rapid pace of hardware updates generates significant electronic waste. The authors argue that if the energy and resources used to create an AI system are greater than the environmental savings it eventually produces, the net result is negative. This is not just a theoretical risk; the review highlights that many studies claim success based on improved computer performance or better predictions, but fail to prove that these improvements actually lead to a measurable reduction in carbon emissions or resource use in the real world.
Beyond the physical footprint, the paper identifies deep risks related to fairness and transparency. Many AI systems are trained on historical data that may contain hidden biases, which can cause the technology to reinforce existing social inequalities rather than solve them. For instance, an AI designed to allocate resources might inadvertently favor wealthy regions over poorer ones if the data it learns from reflects past inequalities. Additionally, the complex nature of these systems often makes them "black boxes," meaning even their creators cannot always explain how they reached a specific decision. This lack of clarity makes it difficult to hold anyone accountable when things go wrong. The review also points out a growing divide in access to this technology; the high cost of developing and implementing AI means that wealthy nations and large corporations are likely to capture most of the benefits, leaving developing countries and smaller businesses behind.
To address these challenges, the researchers propose a framework for prioritizing risks, suggesting that the most critical issues are those involving biased data and unequal access, as these threaten the very foundation of a fair and sustainable society. They argue that the path forward requires more than just building smarter machines; it demands a shift toward "Green AI," which focuses on creating energy-efficient algorithms and managing hardware lifecycles responsibly. This approach must be supported by international cooperation to share knowledge and technology, as well as new policies that ensure transparency and accountability. The authors conclude that artificial intelligence has the potential to be a transformative force for good, but only if its development is guided by a clear understanding of its costs. Without careful management of its environmental impact and a commitment to equitable access, the very tools we hope will save the planet could end up deepening the crises they were meant to solve.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.