Powering climate action strategy with AI (and notes on AI’s carbon footprint)

Powering climate action strategy with AI (and notes on AI’s carbon footprint)

Keen to bring you up to date on interesting articles, publications and sites I come across, here’s an interesting one:

Back in November last year Capgemini, an international company headquartered in Paris and involved in consulting, digital transformation, technology and engineering services, released a PDF document by their Capgemini Research Institute entitled “Climate AI – How artificial intelligence can power your climate action strategy”. Available for download here, the document outlines their research on how artificial intelligence (AI) can accelerate our response to climate change.

The research highlights:

  1. AI offers many climate action use cases
  2. AI-enabled use cases are already reducing GHG emissions and can accelerate climate action
  3. Even though climate action is a strategic priority, most organizations are struggling to support climate action with AI capabilities
  4. How organizations can leverage AI’s full climate action potential
Publication cover image
Publication cover

The nicely-detailed and thorough document, at 68 pages long, breaks each of these items down in great detail and examines each thoroughly and is definitely worth your time.

According to their modelling, some of the ways in which AI can, or already is doing so, impact climate change strategy, include the following:

  • Improve Energy Efficiency
  • Optimize Clean Energy Development
  • Avoid Waste
  • Make Transportation More Efficient
  • Tools to Help Understand Carbon Footprint
  • Monitor Environment
  • Create New Low-Carbon Materials

Download the PDF file here.

MIT News logo

As an aside, in case anyone is wondering, yes, there have been concerns raised about the carbon footprint of AI networks themselves. Discussing this, an article by MIT News states the good news that:

MIT researchers have developed a new automated AI system with improved computational efficiency and a much smaller carbon footprint. The researchers’ system trains one large neural network comprising many pretrained subnetworks of different sizes that can be tailored to diverse hardware platforms without retraining.

This comes off the back of a 2019 report by the University of Massachusetts Amherst which estimated that:

…the amount of power required for training and searching a certain neural network architecture involves the emissions of roughly 626,000 pounds of carbon dioxide. That’s equivalent to nearly five times the lifetime emissions of the average U.S. car, including its manufacturing.

Writing in Geographical Magazine about the carbon footprint of AI and cloud computing, the article’s author notes that AI…

…offers high potential solutions to the climate crisis, but evidence suggests that AI systems and cloud computing will need to clean up their own energy bills.

This is where the MIT report mentioned above comes in. The system, which they call a once-for-all network, dramatically reduces the energy required to train each specialised neural network, which can include billions of Internet of Things (IoT) devices. Thus use of this system resulted in an estimate that…

…the process required roughly 1/1,300 the carbon emissions compared to today’s state-of-the-art neural architecture search approaches, while reducing the inference time by 1.5-2.6 times.

“The aim is smaller, greener neural networks,” says Song Han, an assistant professor in the Department of Electrical Engineering and Computer Science. “Searching efficient neural network architectures has until now had a huge carbon footprint. But we reduced that footprint by orders of magnitude with these new methods.”

I expect of LOT of development to take place in the AI field as the field progresses and grows. One can barely move without hearing something about AI in the news and on the web – it has practically become a buzzword of late. I thus suspect that there’ll be plenty more posts on this topic on the site in the future. Plus, as an added bonus, I’ll get to learn a lot more about it!

References and Sources:

Burgess, M., 2018. What is the Internet of Things? WIRED explains. [online] WIRED UK. Available at: https://www.wired.co.uk/article/internet-of-things-what-is-explained-iot

Cai, H., Gan, C., Wang, T., Zhang, Z. and Han, S., 2019. Once-for-all: Train one network and specialize it for efficient deployment. arXiv preprint arXiv:1908.09791. Available at https://arxiv.org/pdf/1908.09791.pdf

Capgemini Research Institute. 2020. Climate AI – How artificial intelligence can power your climate action strategy. [online] Available at: https://www.capgemini.com/wp-content/uploads/2020/11/Climate-AI_Final.pdf

Dykes, J., 2020. The carbon footprint of AI and cloud computing – Geographical Magazine. [online] Geographical.co.uk. Available at: https://geographical.co.uk/nature/energy/item/3876-the-carbon-footprint-of-ai-and-cloud-computing

Marr, B., 2021. How Artificial Intelligence Can Power Climate Change Strategy. [online] Forbes. Available at: https://www.forbes.com/sites/bernardmarr/2021/01/04/how-artificial-intelligence-can-power-climate-change-strategy/

Matheson, R., 2020. Reducing the carbon footprint of artificial intelligence. [online] MIT News | Massachusetts Institute of Technology. Available at: https://news.mit.edu/2020/artificial-intelligence-ai-carbon-footprint-0423

Strubell, E., Ganesh, A. and McCallum, A., 2019. Energy and policy considerations for deep learning in NLP. arXiv preprint arXiv:1906.02243. Available at: https://arxiv.org/pdf/1906.02243.pdf


Keen to be notified of new posts? Sign up for Coding Climate’s newsletter:

Leave a Reply