Sustainable AI Framework: Keys, Models and Challenges

Artificial intelligence is one of the most transformative technologies of the 21st century. Samarawickrama (2022) even compares it to electricity in terms of its potential impact. Its rapid progress, however, raises ethical and governance challenges that cannot be ignored. How can we ensure that AI is not only efficient, but also sustainable and ethical? In this article, we will explore the framework proposed by this author as a guide to implementing AI in a sustainable way, as well as the challenges that organisations face in this process.

Image by @OpenClipart-Vectors.

The “Kite” Model, a Holistic Approach to AI Governance

One of the pillars of the framework proposed by Samarawickrama is the KITE model, an abstraction that seeks to reduce complexity in AI governance, focusing on four key dimensions: AI and sustainability on the longer axis and society and organisation on the shorter axis. The interconnectedness of these dimensions is proposed to enable organisations to align their AI strategies with broader social justice and sustainable development objectives.

This model suggests, for example, that diversity, equity and inclusion (DEI) should be an integral part of any AI initiative in order to mitigate biases that perpetuate social inequalities, usually caused by a lack of diversity in teams developing algorithms. In other words, to not only ask how to implement AI, but also why and for whom.

The Wind Turbine Model: Practical Governance for AI

Complementing the kite model is the wind turbine model, which focuses on the practical aspects and proposes the visualisation of governance as a turbine, with each component representing a critical aspect of the process. The turbine blades symbolise the values and policies that guide ethical decisions, while the generator represents the technical capabilities that drive innovation.

This model is particularly useful for business leaders, as it allows them to visualise how the different parts of their organisation can work together to achieve the right balance between technical efficiency and social responsibility. It also emphasises the importance of collaboration between diverse actors, from volunteers to technology partners, to ensure that AI serves humanity and not the other way around.

Image by @PeterDargatz.

Challenges in Implementing Sustainable AI

Despite some progress, the implementation of sustainable AI is not without its challenges.

–          Technical and ethical complexity: while artificial intelligence has the capacity to make autonomous decisions, this raises fundamental questions about human dignity, human rights and autonomy such as how to ensure that racial or gender bias is not perpetuated or how to ensure that users’ privacy is not violated?

–          Lack of specific regulation: While ethical frameworks and regulatory guidelines exist, in many countries there are still no concrete laws regulating AI use, in part also due to the ambiguity of the term, which leaves a loophole that can be exploited (Amnesty International, 2024).

–          Environmental sustainability: data centre energy consumption and the associated carbon footprint become urgent concerns. Here, the wind turbine model offers a partial solution by suggesting a more distributed approach, such as fog computing (Baccarelli et al., 2017), which makes use of distributed peripheral devices connected to the cloud to reduce data transfer to the cloud and thus reduce energy consumption.

Towards Ethical and Sustainable AI

AI has the potential to transform not only the economy, but also society as a whole. However, to fully exploit this potential, it is crucial that organisations adopt governance frameworks that integrate ethical, social and environmental considerations. The models proposed by Samarawickrama offer a roadmap for achieving this, but ultimately the key lies in collaboration between different actors and in ensuring the diversity of those involved in the implementation and development of solutions, to guarantee that all possible points of view are considered. Only by working together can we succeed in making AI not only a powerful tool, but also a force for the common good.

References

The Urgent but Difficult Task of Regulating Artificial Intelligence  (2024, Jan 16). Amnesty International https://www.amnesty.org/en/latest/campaigns/2024/01/the-urgent-but-difficult-task-of-regulating-artificial-intelligence/

Samarawickrama, M. (2022). AI Governance and Ethics Framework for Sustainable AI and Sustainability. Retrieved from https://bit.ly/AIESG

Vinuesa, R., Azizpour, H., Leite, I. et al. The role of artificial intelligence in achieving the Sustainable Development Goals. Nat Commun 11, 233 (2020). https://doi.org/10.1038/s41467-019-14108-y