Generative AI and carbon emissions in the Oxford context

AI brings new opportunities to the ways in which we study and work, but it also brings challenges. As AI continues to grow, there is an increasing focus on its environmental implications. This article from the University‘s AI Competency Centre explores these issues and how Oxford is responding to them. 

Understanding the real impact of AI 

One of the challenges of considering AI’s energy impact is the difficulty of getting a real understanding of its energy usage and environmental implications from tech companies. 

It’s challenging to get a reliable estimate of the energy consumption of AI because a lot depends on the application and model being used, and available approximations vary considerably. Whilst the sector is expanding rapidly, the technology is also changing over time, bringing improvements in the energy consumption of AI models, hardware and data centres. 

As AI adoption continues to increase, the University is working to understand usage and impact, to help explore how we can mitigate the carbon emissions that are an inevitable consequence of any new technology being widely adopted. The AI Competency Centre is at the forefront of this work, as well as supporting you during your time here. There are opportunities for free training and guidance on using AI effectively and responsibly, including group training sessions on Getting Started with AI which your department or college can book for you. On-demand resources are also available on Canvas.   

Meanwhile, some Oxford researchers are using GenAI to enable energy-saving innovations, as a recent ChatGPT use case from Regent Lee, Professor of Interdisciplinary Innovation in Oxford’s Nuffield Department of Surgical Sciences demonstrates. 

How the University is monitoring, using and mitigating AI related emissions  

Work is also underway in other parts of the University to consider how AI could help to reduce Oxford’s carbon footprint– for example, by exploring opportunities to use technology in trialling sensors that help us to manage buildings as efficiently as possible. This technology could help the University to use space more efficiently, in terms of cooling, heating, and lighting. 

Elsewhere, the University’s Gardens, Libraries and Museums (GLAM) division has been working on a project to understand carbon emissions across its digital activities, including the impact of AI.  

AI’s energy usage comes in two parts: the initial energy required for training new models, which is usually a large amount over a short period of time; and the energy cost per individual query, which is relatively low per prompt but can add up over time. Although the University-supported GenAI platforms accessed via Single Sign-On (SSO), such as ChatGPT Edu, ensure that your data is not used to train these models, the University is engaging with tech companies such as OpenAI, Google, and Microsoft to push for more sustainable practices. Oxford’s collaborations mean it can directly question providers on how they are delivering on their net zero targets. 

Several partners have published statements on their green technology ambitions: for example, Microsoft’s commitment to becoming carbon negative by 2030. OpenAI has published a commitment to minimising its environmental impact and reducing energy consumption at all stages of AI development, stating that each generation of their models uses less energy to train and run. They also have a new data centre in Norway that will be powered entirely by renewable energy. 

Including AI in Oxford’s Environmental Sustainability Strategy 

The University sets out its commitment to sustainability in the Environmental Sustainability Strategy, which includes targets for achieving net-zero carbon and biodiversity net gain by 2035. This strategy is reviewed in five-year cycles. Given the rapid global adoption of AI over the last few years, it is becoming increasingly important for Oxford to understand how these tools are being used for work and study, to get a clear picture of the environmental implications. 

The University is carrying out a comprehensive review of its carbon accounting. In line with best practice, this will include indirect greenhouse gas emissions related to products the University uses, such as the emissions from third-party data centres that AI platforms rely on. This review will give us a better picture of Oxford’s full environmental impact and will help us to track the changing impact of AI use over time and identify purposeful steps to address associated emissions. 

Understanding how AI will affect not only the University’s carbon footprint, but also everyday teaching and learning, research, and administration, is an ongoing process. The University is actively monitoring and adapting to this fast-changing field, building a clearer picture of how AI is being used and of how this usage aligns with our sustainability targets, in order to inform and shape our approach.