Greenhouse gas emissions linked to digitalisation are expected to grow rapidly. The French report AI for humanity recently presented by mathematician Cédric Villani offers important insights. In a well-written section on AI and ecological economy, the Villani task force notes: “Digital energy consumption increases by 8.5% per year and its contribution to world electricity consumption (which is growing by 2% per year) could reach 20% (moderate scenario) or even 50% (pessimistic scenario) by 2030, and therefore be multiplied 10-fold in 20 years’ time.”
The group comments on other environmental aspects: “The production of digital hardware uses large quantities of rare precious metals which are only partly recyclable, and the available reserves are limited (15 years in the case of Indium, for example, the consumption of which has multiplied 7-fold in 10 years).” Generation of dangerous waste, unsustainable mining of rare earth metals and high water consumption of data centres can be added to this picture.
Other studies have also found rapid increases of energy use and climate impact from digitalisation. Storage and processing of big data require large amounts of electricity, as well as the production of computers, screens and smartphones. Energy use linked to block-chain technology grows rapidly. In one recent study, researchers Lotfi Belkhir and Ahmed Elmeligi found that by 2040, emissions from the use of ICT could correspond to more than 14% of today´s total emissions (published in Journal of Cleaner Production (Volume 177, 10 March 2018).
Of course, digitalisation also has benefits for the environment. This is often highlighted in policy documents, for example in the European Commission´s communications on the digital single market. Intelligent transport systems in smart cities is one example, more effective energy production and distribution another.
Still, increasing energy consumption is an issue that needs to be addressed, as well as other negative environmental effects. However, policy responses to this challenge are to a large extent lacking. Yes, there are initiatives to power data centres with renewable energy and to reduce their power consumption. But this is not enough. For example, the drastic increase of energy use for bitcoin production should have been identified earlier as an environmental issue and alternative, more efficient algorithms actively promoted.
It is high time for the ICT sector to feature more prominent in climate strategies, and for ecological sustainability to be a key issue in digital strategies. As the European Trade Union Confederation, ETUC, notes in a resolution, “the deployment of digital technologies should be accompanied by a set of regulations and standards, which will help to ensure the – social, economic and environmental – sustainability of ICT value chains. The EU must also ensure that its action on digitalisation fits with the targets of its climate, energy and environment policies.”
One particular aspect is the current rapid development of machine learning. AI strategies are rapidly being adopted by a number of governments. The European Commission aims to agree a coordinated plan on AI with Member States by the end of this year. There are positive elements in the Commission communication on AI, for example on algorithmic awareness building. But it is crucial that environmental concerns, such as energy consumption for data processing, are included among the issues to be addressed. Currently, they are not. A key issue is how AI systems are trained. There is for example an urgent need for research and development on how to apply machine learning in a way that does not reproduce earlier ecological mistakes.
The Villani report highlights both risks and positive effects of AI on sustainability: “Although AI is a potential threat to the environment, it is also a potential solution. Indeed, there are many opportunities to use AI in the field of ecology: AI can help us understand the dynamics and the evolution of whole ecosystems by focusing on their biological complexity; it will allow us to manage our resources more efficiently (particularly in terms of energy), preserve our environment and encourage biodiversity.”
Cédric Villani offers a number of recommendations, inter alia:
- Establishing a Meeting-Point for the Ecological Transition and AI
- Establishing a Platform for Measuring the Environmental Impact of Intelligent Digital Solutions
- Designing AI that Uses Less Energy (data centres, cloud providers, and alternatives to today´s energy intense graphics processing units, GPU´s)
Integration of environmental concerns in all sectors is a key principle in the European Union Treaties. Strategies on digitalisation and AI should reflect that. It will make EU policy stronger, both when it comes to sustainability and to long-term competitiveness. As part of such a broader ecological approach, the European Commission should listen to Cédric Villani and his colleagues and integrate the environment in the AI strategy to be agreed this year. That would also facilitate international cooperation on this truly global issue.