Written question to the EC on the carbon footprint of artificial intelligence

Artificial intelligence (AI) has significant potential when it comes to reducing greenhouse gas emissions and helping society to combat and adapt to climate change.

However, at the same time, according to a recent Cornell University study(1), machine learning improvements ‘depend on the availability of exceptionally large computational resources that necessitate similarly substantial energy consumption’.

In the light of the ICT sector’s significant environmental footprint, as acknowledged by the Commission in its communication of 19 February 2020 on shaping Europe’s digital future(2):

  1. How does the Commission intend to address the matter as part of its upcoming Coordinated Plan on Artificial Intelligence and its legislative initiative on AI?
  2. Is it considering including any best practices or tools for measuring the electricity consumption or carbon emissions of machine learning models, or ways of making this information transparent?
  3. Will it consider these best practices or tools when evaluating proposals for funding programmes under the 2021-2017 multiannual financial framework?

The question was co-signed by MEPs Stelios Kympouropoulos (PPE), Anna-Michelle Asimakopoulou (PPE) and Loucas Fourlas (PPE).

The question here and the reply from Commissioner Breton here.