In an unprecedented “call for collaboration,” a group of 22 respected AI experts that includes Andrew Ng, Yoshua Bengio, and Demis Hassabis have published a paper exploring how machine learning (ML) could help deal with climate change by reducing greenhouse gases (GHG) and proposing how societies might initiate and adapt to these changes. The paper is a joint effort with the Mercator Research Institute on Global Commons and Climate Change.
Undeniably, extreme weather events such as storms, droughts, fires and flooding are causing more and more injuries, deaths and damage worldwide. The changing global biological system is also threatening natural resources and agriculture on which human rely for existence. The paper points to a 2018 IPCC special report on the impacts of global warming that predicts disastrous consequences if GHG are not curbed within 30 years.
In its examination of the electrical power systems — which produce the most GHG — the paper looks at how traditional fuels could be replaced by low-carbon energies using ML. It delves into generation and demand forecasting, nuclear fission and fusion, and reducing life-cycle fossil fuel emissions. The scope also includes improving electricity access in regions where this is an issue, and methods for low-data ML techniques.
The authors identify researchers, entrepreneurs, corporate leaders and governments as key players whose attention needs to be drawn to climate change; and introduces three tags — High Leverage, High Risk, and Long-term — to define the priority, feasibility, and possibility for the application of ML methods on different challenges. The subsection “Accelerated science for materials” for example is marked with all three tags.
The paper is broken down into 13 topics and identifies specific AI techniques for each at the application level (see the table below).
The authors propose that both mitigation of GHG emissions through changes to electricity systems, transportation, buildings, industry, and land use; and adaptation to risks through climate modeling, risk prediction, and planning for resilience and disaster management present a wide range of areas where ML researchers could apply their efforts. The paper has over 40 pages of references, identifying existing studies which could be good starting points for further research.
The authors stress that ML could be an effective tool towards an eco-friendly future, but that different ML tools should be used in different societal settings depending on the structures of specific countries. They propose a four-step roadmap for researchers and others who want to make good use of this paper to combat climate change — Learn, Collaborate, Listen and Deploy.
The paper Tackling Climate Change with Machine Learning is on arXiv. An associated website provides additional resources.
Author: Reina Qi Wan | Editor: Michael Sarazen