Measuring the environmental impacts of artificial intelligence computing and applications: the AI footprint

This report looks at how policy makers can better understand and react to the positive and negative environmental impacts of AI for the good of the planet.

Get the report on the OECD’s iLibrary

Artificial intelligence (AI) systems can help achieve sustainability goals but can also use massive computational resources, raising sustainability concerns. This report looks at how policy makers can better understand AI’s direct and indirect impacts on the environment and react to both the positive and negative for the planet’s good.

More concretely, the report calls on policy makers to establish measurement standards, expand data collection, identify AI-specific impacts, look beyond operational energy use and emissions, and improve transparency and equity. It also details existing measurement tools and critical challenges governments face to quantify the positive and negative environmental impacts of training and deploying AI models and applications.

It concludes by saying countries should unite to create indicators specific to AI computing, or “compute”, share best practices, and support new and innovative AI applications to fight climate change. In doing so, countries can ensure that AI is trained and deployed in the most sustainable way possible while minimising negative environmental impacts.

Presentation and panel discussion at COP27, November 2022

WATCH THE PRESENTATION AND PANEL DISCUSSION

At the OECD’s COP27 Virtual Pavilion

Opening remarks:

  • Jens Lundsgaard, Deputy Director of the Science, Technology and Innovation Directorate, OECD

Presentation of the report:

Confirmed panellists

From the AI Wonk

The OECD Framework for Classifying AI Systems to assess policy challenges and ensure international standards in AI
Intergovernmental

The OECD Framework for Classifying AI Systems to assess policy challenges and ensure international standards in AI

The classification framework can help policy makers understand and map the affects of AI systems on societies and economies.

February 17, 2022 — clock4 min read

Familiar methods can help to ensure trustworthy AI as the algorithm auditing industry grows
Academia

Familiar methods can help to ensure trustworthy AI as the algorithm auditing industry grows

This is a first step towards defining the key components of algorithm auditing.

August 10, 2021 — clock6 min read

Deepmind’s Alphafold: a solution to the 50-year-old challenge of protein folding
Technical community

Deepmind’s Alphafold: a solution to the 50-year-old challenge of protein folding

In the future, AlphaFold should help drive advances in many areas, such as drug design and environmental sustainability.

December 14, 2020 — clock2 min read

Artificial intelligence in health still needs human intelligence
Intergovernmental

Artificial intelligence in health still needs human intelligence

COVID-19 has shown that AI can deliver benefits, but it has also exposed its limits, often related to having the right data.

September 15, 2020 — clock8 min read

Can AI combat cognitive capacity issues and information overload?
Business

Can AI combat cognitive capacity issues and information overload?

Cybersecurity, viral pandemics and climate change make fostering international co-operation and access to knowledge more urgent than ever.

June 15, 2020 — clock3 min read