If Artificial Intelligence (AI) were running the US, it’s unlikely it would abandon the Paris Climate Agreement — rather it would continue working on the climate change challenges facing our planet.
President Trump’s decision to pull the US out of the Paris Agreement was expected. Trump has repeatedly called global warming a “hoax,” and had already asked Congress to cut or eliminate a range of programs designed to help solve climate change issues, from reducing gas emissions to predicting extreme events.
The fact is, we are seeing record increases in extreme weather, deforestation and greenhouse gas emissions. Last year, there were 15 extreme weather events across the US — almost three times the annual average for 1980-2016 (5.5 events per year) — causing loss of life, displacement, and billions of dollars of damage. What can be done?
To reduce economic losses and protect our ecosystem, an increasing number of scientists are employing AI in climate science research. The state-of-the-art technology performs well in handling large datasets from climate archives to help humans understand the future impact of climate change and how to prepare for extreme events.
Mr. Prabhat, a climate scientist from Lawrence Berkeley National Lab, is exploring the use of deep learning to study extreme weather. To study the climate system, scientists usually build a large scale climate simulation to detect and identify extreme weather events. Deep learning can improve accuracy by classifying well-known types of extreme weather events based on large amounts of labeled data.
Last December, Prabhat took a further step. Labeled data is lacking for many localized climate patterns of interest such as tropical cyclones (hurricanes/typhoons), Prabhat and four other scientists designed a new architecture to improve the detection of a wide range of extreme weather events. Their research helps us envision how extreme weather may vary in the future, especially under various carbon emission scenarios.
Deforestation is another research area that suits deep learning’s capabilities. Greg Asner, an ecologist at the Carnegie Institution for Science and Stanford University, used aerial photographs and GPU-accelerated deep learning to identify different tree types in the Peruvian Amazon, an almost 100,000 square-kilometre rainforest in South America.
Asner’s team mapped the distribution of 36 types of trees, helping the government create specific strategies to preserve the biodiversity of the Peruvian Amazon. Asner has applied a similar process in California, Hawaii, Borneo and Ecuador. His plan is to use his technology to study different equatorial regions and build a global map of biodiversity.
AI has also even proven capable of reducing its own carbon footprint. Last year, DeepMind, the London-based AI company owned by Google, developed an algorithm to reduce the amount of energy consumed by data centres (facilities housing large computer systems).
Greenhouse gas emissions are linked with data center efficiency because much of the electricity used by data centers is coal-generated. DeepMind managed to lower the energy required for cooling data centers by up to 40%, or about 15% of Power Usage Effectiveness (PUE). They created a framework to understand data center dynamics — a collective dataset of temperatures, power, pump speeds, setpoints — and then trained the network.
While AI’s application in climate change research is still at an early stage, the technology is undoubtedly contributing to a better understanding of our planet and helping us make better decisions regarding its future. This will continue with or without President Trump.
Journalist: Tony Peng | Editor: Michael Sarazen
0 comments on “Trump Dismisses Global Warming; AI Seeks Solution”