AI - a tool for sustainability with a resource cost
AI's sustainability remains a nuanced question, balancing its potential benefits against its environmental costs.
AI has become a significant area of focus across industries, promising to provide innovative solutions to complex challenges, including those posed by climate change. Its potential is evident. Taking the Pharmaceutical Sector as an example, use cases span a wide array of support activities, from drug discovery to distribution and supply chain management, as discussed at last year's PING Conference, 'AI in Pharma - Threat or Opportunity?'
How is AI a beneficial tool for sustainability?
AI holds immense promise for advancing environmental goals, primarily through its potential to optimise energy systems, enhance efficiency, and improve climate modelling.
Optimising Energy Systems - AI excels at analysing vast datasets to optimise energy generation, distribution, and consumption. One use case for AI is to integrate renewable energy sources into power grids, so reducing dependency on fossil fuels. For example, predictive algorithms could enable smarter load balancing and energy storage, allowing the effective harnessing and utilisation of renewable energy.
Enhancing Efficiency Across Sectors - AI-driven technologies can significantly reduce energy waste in key sectors such as transportation, agriculture, manufacturing, and building management. For instance, AI-powered smart manufacturing processes have the potential to cut energy consumption, waste, and carbon emissions by 30-50%, and could cut carbon emissions by approximately 60% within the transportation sector (Artificial intelligence-based solutions for climate change: a review, Chen et al - Environ. Chem. Lett. 21 (2023)).
Improving Climate Modelling and Prediction - Accurate climate models are essential to understand and mitigate climate change. AI algorithms can enhance the precision of these models, enabling better predictions of weather patterns, sea-level rise, and other climate variables. This in turn can allow for informed decision-making and the development of effective climate adaptation strategies.
What are the challenges against AI's sustainability?
Despite its benefits, AI requires substantial amounts of energy as well as resource for its hardware.
Energy Footprint - The computational demands of training AI models, particularly deep learning models, require significant computing resource, leading to greater energy consumption. Data centres hosting these processes substantially contribute to electricity consumption, and the demand for AI is only increasing. Globally, data centres account for 1% of energy consumption which doesn't sound like much. However, for more advanced economies this percentage is far greater. For example, data centres in the Republic of Ireland account for 20% the nation's energy consumption in 2023 (Republic of Ireland Central Statistics Office). If these centres rely on fossil fuel-based energy to meet demand, their carbon footprint undermines the potential for sustainability of AI.
Natural Resources - The hardware needed to support AI systems relies on extensive mining and/or refinement of raw materials like lithium, cobalt, copper and silicon. These materials are critical for manufacturing semiconductors, batteries, and other AI-enabling technologies. However, mining and processing these resources can result in deforestation, habitat destruction, and water pollution. The extraction and purification processes also generate substantial greenhouse gas emissions, compounding their environmental impact. In 2022, the global production of lithium was 130,000 tonnes compared to 37,000 tonnes a decade prior. For each tonne of Lithium produced, the CO2 emissions produced range from between 11 and 37 tonnes (MIT Climate Portal). Although not specific to AI, its increased demand has and will continue to contribute towards these figures.
Addressing these issues and moving towards sustainable AI
Suggested changes to improve the sustainability of AI include:
- Developing energy-efficient algorithms and hardware to reduce the environmental impact of AI infrastructure
- Mandating transparency about the energy consumption and carbon footprint of AI systems, encouraging greener practices and accountability
- Creating certification programmes to identify and promote energy-efficient AI solutions (green AI), aiding consumers and investors in making sustainable choices
- Encouraging data sharing among researchers and organisations to reduce redundant training of AI models, lowering overall energy consumption in AI development.
The above points offer some solutions, but to facilitate successful and widespread adoption, government and stakeholder backing is crucial.
In February 2024, the UK Government published voluntary guidance for regulators to support the implementation of its pro-innovation regulatory principles, which did not cite sustainability within its five key principles. However, the UK Sustainability Reporting Standard will require companies caught by the Standards to report on certain sustainability-related information. Clearly there is a balance to be struck between the UK Government's pro-innovation agenda and sustainability regulation, which needs to make its way into formal regulation.
In the EU, progress towards regulating the use of AI is already underway. The Artificial Intelligence Act (Regulation (EU) 2024/1689) (EU AI Act) will incrementally enter into force over the next few years and contains environmentally minded provisions. For example, Recital 27 sets out that AI systems should be developed in a "sustainable and environmentally friendly manner as well as in a way to benefit all human beings, while monitoring and assessing the long-term impacts on the individual, society and democracy". It will be interesting to see how this will be implemented on a member state basis.
The UK Government is likely to be carefully monitoring the progress of the EU AI Act to let their own requirements around sustainability crystallise. With this in mind, businesses can get ahead by implementing their own AI policies that incorporate measures aimed at addressing sustainability. Our specialist lawyers at VWV can assist with drafting such policies and so please reach out to us if you require any assistance.