Oxford Contributes to National Report on Sustainable AI

Artificial Intelligence (AI) is transforming industries, from healthcare to climate science, but its rapid growth comes with significant environmental challenges. AI systems require vast amounts of energy and water, and data centres hosting AI models contribute to carbon emissions and the depletion of critical materials.

The report, Engineering Responsible AI: Foundations for Environmentally Sustainable AI, was developed by the Royal Academy of Engineering, Institution of Engineering and Technology, and BCS, the Chartered Institute of IT, under the National Engineering Policy Centre (NEPC).

“In an era in which AI is rapidly becoming ubiquitous, it’s crucial to examine its impact on energy consumption.”

The Engineering Responsible AI report calls on government to promote, prioritise and invest in sustainable AI and proposes five foundational steps that can be taken by governments now to advance efficiency and frugality:

1. Expanding environmental reporting mandates
2. Providing information on environmental impacts of AI systems across the value chain, including AI compute, IT infrastructure, data and algorithms, interaction and use.
3. Setting environmental sustainability requirements for data centres​
4. Reconsidering data collection, transmission, storage, and management practices
5. Leading the way with government investment

Academics at The Department of Engineering Science played a key role in shaping these recommendations, with contributions from Professor Stephen Roberts, Professor of Machine Learning, and Dr Eve Schooler, Visiting Professor, Royal Academy of Engineering Visiting Professor of Sustainable Computing.

Professor Roberts’ research focuses on developing more computationally efficient AI models, ensuring that machine learning systems can be both powerful and resource conscious. His research lies in machine learning approaches to data analysis with particular interests in the development of machine learning theory for problems in time series analysis and decision theory. He says of the report:

“In an era in which AI is rapidly becoming ubiquitous, it’s crucial to examine its impact on energy consumption. The extensive training and widespread deployment of large language models, coupled with their insatiable data appetite, are already impacting global energy consumption and driving demand for increasingly large data centres. To mitigate this trend, we must establish alternative frameworks that enable us to harness the undoubted capabilities of AI whilst minimising its impact.

This report serves as a significant step forward, considering the potential of smaller, capable, AI methods, the opportunities for hardware innovations to reduce energy usage, and the need to provide salient information to AI adopters, enabling them to consider the full costs of AI in their decision making.”

Dr Schooler, who is currently based at Oxford’s Department of Engineering Science, brings expertise in sustainable computing and data infrastructure, offering key insights on reducing the environmental impact of AI systems at scale. She explains,

“Although there is such palpable excitement and expectation around the promise of AI to accelerate innovation, there exists the very real concern about its growing environmental impacts. This report offers a series of concrete steps towards an actionable strategy within the UK to tackle AI sustainability, to ensure we can meet our ambitious long-term environmental goals, particularly around net-zero emissions and decarbonisation.”

Professor Tom Rodden CBE FREng FRS FBCS, Pro-Vice-Chancellor of Research & Knowledge Exchange and Professor of Computing, University of Nottingham and Chair of the working group says:

“In recent years advances in AI systems and services has largely been driven by a race for size scale, demanding increasing amounts of computational power. As a result, AI systems and services are growing in size at a rate unparalleled by other high energy systems – and generally without much regard for resource efficiency.  This is a dangerous trend, and we face a real risk that our development, deployment and use of AI could do irreparable damage to the environment.”

“To build systems and services that effectively use resource, we first need to effectively monitor their environmental cost. Once we have access to trustworthy data pertaining to their environmental impacts, and a sense for where these services and systems are needed, we can begin to effectively target efficiency in development, deployment, and use - and plan a sustainable AI future for the UK.”

This report adds to the growing body of work at Oxford focused on AI’s impact on society and the environment, reinforcing the importance of interdisciplinary collaboration between engineering, policy, and industry.

Read the full report here.