How sustainable are AI policies? Evaluating the environmental impact of different national AI strategies.
There is a broader consensus among major stakeholders that AI based systems aid in fighting climate change, reduce carbon emissions, optimise energy production and resource utilisation. However, while AI may impact environmental outcomes positively, it can also affect environmental sustainability adversely. It could aid in increasing carbon emissions further as it requires more energy to train Machine Learning, Deep Learning models and neural networks. It is within this background that this project puts the question of the Sustainability of AI at its core and proposes to analyse how AI national strategies, policies and regulations in five major AI technology developing and adopting countries (USA, UK, Germany, China and India) deal with the question of AI sustainability in their policy formulation and implementation.
Main Results
The key findings substantiate the hypothesis that the selected national AI policies (China, Germany, India, UK and USA) do not have adequate emphasis on the ecological impact of AI systems. The multilateral and supranational bodies like UN, OECD, EU and Global Partnership on AI (GPAI) have published studies which provide policy direction to the national policy making bodies.
The research also elucidates vivid contradictions between the two seemigly similar policy goals: the national AI policies and the climate policy of the selected countries. While national AI policies emphasise the use of AI in finding solutions for climate challenges the climate policies do not recognise the emissions and ecoclogical damage caused by the AI systems.
Based on this research several policy recommendations are made:
- Policy convergence among different national policy making bodies that deal with both AI Policy and Climate Policy is essential to deal with the ecological challenges posed by AI systems.
- Open standards and transperancy in reporting metrics of emission caused by AI systems have to be promoted.
- Adopting open source AI will enable reusability of models and reduces the need for retraining the data which in turn decreases the intense use of resources including high energy utilization.
Main Outcomes
- In June 2023, I was invited to a roundtable discussion to talk about sustainable AI by ITforChange, an India based digital policy focused think tank. https://itforchange.net/sites/default/files/2374/Agenda.pdf
- I am currently working on a research paper on Regenerative AI Framewrok with inputs form over 70 experts led by Hippo AI Foundation. This is expected to be published in the mid of 2024. Some of my reasearch at CAIS will be part of this paper.
Main Research Topics
- Sustainable AI
- Open AI and Data governance
- Political economy of digital transformation
- Data induced inequalities
Curriculum Vitae
2019 – Present: Independent digital policy researcher.
2012 – 2017: Senior Advisor, Market Development, IBM India.
2007 – 2011: Assistant Manager, Business Research, ITC Infotech India Ltd.
2006 – 2007: Analyst, Oracle India.
2005 – 2006: Analyst, Deloitte India.
Lectures and Publications
Mukku, S. (2022) Future of FinTech in Asia: Technology governance in a transitioning political economy. (Forthcoming)