Climate change is one of the greatest threats humans have ever faced, with increasingly severe consequences feared as sea levels rise, ecosystems falter, and natural disasters multiply. Tackling climate change is a huge and complex challenge, where it’s hoped that AI-powered efforts can play an equally huge and beneficial role.
Organizers of NeurIPS 2020 (Conference on Neural Information Processing Systems) see machine learning (ML) as an invaluable tool in the fight against climate change. A wide array of applications and techniques are already being explored, from smart electric grid design to satellite-tracking of greenhouse gas emissions and countless others.
Last Friday, NeurIPS 2020 partnered with Climate Change AI (CCAI) — an organization of researchers, engineers, entrepreneurs, investors, policymakers, companies and NGOs aiming to catalyze impactful work at the intersection of climate change and machine learning — to host the Tackling Climate Change with ML Workshop, which explored how the ML community could collaborate with other fields and practitioners in this fight.
The all-virtual format of NeurIPS 2020, which ran December 6–12, provided a unique opportunity to foster cross-pollination between ML researchers and experts across diverse fields. The Climate Change Workshop featured Rose Mwebaza, director of the UN Climate Technology Centre & Network; Zico Kolter, associate professor in the Computer Science Department at Carnegie Mellon University and chief scientist of AI research for the Bosch Center for Artificial Intelligence; Anima Anandkumar, Bren Professor at Caltech and director of ML Research at NVIDIA, and other speakers, who discussed ML’s contributions to combating climate change.
It’s crucial for the ML community to work with disciplinary experts to identify the urgent problems and figure out where and how to use ML as tools to implement climate change strategies, says Jennifer Chayes, associate provost of the Division of Computing, Data Science, and Society, and dean of the School of Information at UC Berkeley. Chayes was one of the workshop’s keynote speakers.
“I started out as a mathematical physicist, and we would take physics problems from the world and we would abstract them until they were gorgeous models, but that’s not the way we’re going to tackle climate change,” Chayes explains. When researchers deal with climate problems — as is the case with many urgent challenges facing society — they usually don’t have sufficient clean datasets. That’s where ML can really help.
Chayes gave examples of how machine learning can be applied to carbon capture, energy use reduction, material design for renewables and storage, and so on. She also mentioned how ML enables researchers to better understand the dynamics of climate cycling, the impact of climate change, and its profound human interest — for example how the use of ML to determine ecosystem productivity can help identify climate vulnerable or resistant areas.
ML and data science need to play a wide role, Chayes stresses, “by better managing our environment, finding better materials, better controlling our use of energy, pulling carbon out of the atmosphere,” as well as “understanding the economic drivers that move us from one energy source to another,” and “integrating the technical with the social and human dimensions” to propose the best policy interventions. “Indeed, we need an integrated platform for climate — it needs to involve behavioural changes, technical solutions, economic constraints, physical laws, and geopolitical factors.”
Chayes says her team at Berkeley is currently brainstorming how to build such a platform to bring not only heterogeneous data but also different approaches and considerations into play to find end-to-end solutions to pressing climate problems.
Chayes also touched on ways of addressing climate change that are closer to many people’s day-to-day lives, introducing an example of using reinforcement learning (RL) to alleviate traffic congestion. By training agents to maximize the notion of cumulative reward, RL can help drivers optimize for travel routes, save time, and reduce fuel consumption.
Angel Hsu, assistant professor of environmental studies at Yale-NUS College and founder and principal investigator of the Data-Driven Lab, shared a similar use case in the workshop’s The Role of Public Policy panel discussion.
“I was recently living in Singapore, and Singapore I think always tops the charts when it comes to smart city efforts and AI,” Hsu says. “They’ve had the smart nation initiative in place since 2014.”
In the five years she lived there, Hsu says, there wasn’t a single day she was stuck in traffic. “There simply isn’t any traffic, and it’s because they’re using AI to monitor real-time traffic patterns and to be able to control different flows.” Smooth flow means cars and buses are not stuck in stop-and-go traffic, and greatly increases energy efficiency.
Hsu says a big problem now is a lack of evaluation data, from the city perspective for example, that could enable understanding of which climate policies are working and which are not and how cities are performing. “More than 12,000 cities are voluntarily or taking other types of action on climate change, but we don’t have a good sense of whether or not they’re actually achieving those goals, and what is the overall aggregate impact of those efforts,” she explains.
In the same panel discussion, Moustapha Cissé, the Head of the Google AI Center in Accra, Ghana, said he believes governments’ role is to create a framework and conditions that will give private sectors and research institutions the freedom and the incentives to do the work necessary to tackle the most pressing climate-related problems. “I don’t think the burden should be on the government only or on the private sector only,” he says, “but the government has the responsibility of putting the legal framework and infrastructure that allows everyone to contribute, especially the private sector.”
Vinod Khosla, founder of a VC firm focused on energy and technology companies with a particular interest in big problems amenable to technology solutions such as climate change meanwhile identified a dozen problems and sectors he thinks are vital to climate change solutions, such as electric vehicles, meat production, cement industry and grid storage.
The workshop also included several poster sessions and spotlight talks and presented three awards — including the overall Best Paper Award, which went to A Machine Learning Approach to Methane Emissions Mitigation in the Oil and Gas Industry. The authors, from Harrisburg University, University of Illinois at Chicago and Stanford University, propose a machine learning algorithm to predict high-emitting methane leak sites that can be prioritized for follow-up repair.