4 minute read
We live in an increasingly digital world with the growth and use of AI rising exponentially – with this growth comes concerns over its environmental impact. The need for a standardised approach on how to calculate, report, reduce and prevent this impact is vital.
To combat this challenge, international partners have developed a roadmap to ensure the efficient use of resources, enhance clarity, promote consistency in AI environmental sustainability standardisation and facilitate the widespread adoption of best practices.
The roadmap is expected to be used by a wide range of stakeholders, from AI developers, those who build novel AI solutions such as ChatGPT, and AI user, those who apply such solutions in their work and products, to AI consumers/customers, those who use AI-altered products.
There are currently several gaps relating to AI sustainability. A first step of future standardisation efforts will be to identify and establish the following:
The key factors to consider regarding the standardisation:
Valerie Livina, Principal Scientist, NPL and contributor to the roadmap, said: “With novel AI solutions becoming increasingly demanding for energy supply, it is very important to raise awareness of their environmental impact. The current corporate hardware requirements of AI-dedicated clusters are of the scale 200,000 of GPU units, and growing, while most recent models of GPUs may require up to 1.2kW power per unit. This makes AI-focussed hardware so energy-demanding that large corporations already use their own power stations. This roadmap has been developed in preparation for the EU initiative on Frugal AI, and it is well-aligned with the CEN/CENELEC technical report on environmentally sustainable AI that NPL led.”
View the roadmap here
04 Mar 2025