AI Business is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 3099067.

IT & Data Center

Asking better climate questions with artificial intelligence

by Max Smolaks
Article Image

by Simone Warren, Verne Global
17 March 2020

Artificial intelligence (AI) represents a paradigm shift in our ability to address the most critical issues facing society today. None more so than the questions of how to solve the world’s climate crisis.

If we are going to solve the problems being born from climate change, we cannot rely on brute force. Rather we need to use the tools available to us – AI and machine learning included – to leverage data in a way that has never been done before.

AI enables us to find patterns in areas our eyes and brains literally cannot see. Deep learning, for example, uses numerical techniques based on the architecture of the brain to both learn skills and find information. It’s a process of training all the layers of a neural network to find the correct techniques that yield the correct behavior. It takes days and weeks of compute power processing through trillions and trillions of attempts to get the numbers lined up correctly. These types of AI methods will be able to help us choose the best input conditions, boundary conditions and even the appropriate tools to improve our understanding of where to focus the world's climate energies. 

Last November, Climate Change AI released a 60 page report - Tackling Climate Change with Machine Learning - written by the key thought leaders, academics and researchers in the field. The paper categorized a wide variety of AI solutions into high leverage, long term, and uncertain impact categories, but did also identify a few broad areas of deployment. As The Verge summarized, “Prominent among these are using machine vision to monitor the environment; using data analysis to find inefficiencies in emission-heavy industries; and using AI to model complex systems, like Earth’s own climate, so we can better prepare for future changes.”

The
report, however, did have one major caution - AI is not a magic
bullet. Machine learning is only one piece of the puzzle and all the
applications require cooperation between many stakeholders. Lynn
Kaack, a postdoctoral researcher at ETH Zürich and co-chair of
Climate Change AI, said, “It is important that machine learning
experts work together with those who deeply understand the problems
of climate change to make sure machine learning is applied where it
can make a difference.” That type of collaboration is critical, not
only to taking meaningful climate action, but also in making more
informed climate policy decisions and enabling governments to be
better prepared. 

We not only need to employ the right models and data, we also need to consider the carbon costs of AI itself. Machine learning requires compute power, and a lot of it. AI modelling requires millions to billions of simulations in order to find the right solution. For a problem as large as climate change, it will easily be on the billions end of the scale. The larger the dataset and longer the training time, the higher the level of computing power needed to train the model, and the larger the carbon footprint. With accuracy heralded as the most important metric in terms of training AI models, the emissions of the field are growing rapidly.

Research
conducted by the University of Massachusetts found that the training
process of a single natural language processing AI model can emit the
equivalent amount of carbon dioxide as three hundred round-trip
flights between New York and San Francisco. With this in mind, it
becomes clear that when using machine learning to tackle climate
change, the efficiency of models should be a key consideration.

Alongside
the efficiency of machine learning models, the sustainability of the
data centers powering the AI compute is also a prime concern. Many
data centers talk about being green, boasting of their ‘green
certification,’ but these certificates aren’t a guarantee that
they are powered by green energy. The reality is that most are
located in areas where the primary source of power is based on fossil
fuels.

In
terms of data center sustainability, location is a key starting
point. More than 80% of compute doesn’t need to be near the
end-user, and in those cases, being strategically located where
abundant, predominantly green energy is available has a large impact
on carbon emissions. Geographical climate is also an important factor
of data center sustainability. Cooling
data center IT equipment can constitute over 40% of its total energy
consumption, per Cooling
Energy Consumption Investigation research (Zhang et al). Therefore,
the cooler the climate, the less energy required to cool the data
center. Given the massive compute power needed by AI to tackle the
questions posed by climate change, it is imperative that the
supercomputers running these simulations are running on renewable
energy, in energy efficient locales.

The world’s exponentially increasing demand for technology to solve the most complex climate action challenges will require a corresponding increase in the need for energy sources to power that technology. With this in mind, before we even look into whether AI is the right tool to help us tackle climate change or not, perhaps the time has come regardless to move the most power intensive applications to the locations where carbon impact is minimal.


Simone Warren is VP of Global Customer Strategy at Verne Global, a hyperscale data center operator based in Iceland

Practitioner Portal - for AI practitioners

Story

AI and self-service business intelligence – competing or complementing concepts?

7/8/2020

One term – data analytics – having two meanings – AI and SSBI – this is the classic setup for misunderstandings, failing project pitches, and failed projects. But what exactly are the differences between AI and SSBI? And are they complementing or competing concepts?

Story

Open source platform aims to speed up autonomous car development

7/6/2020

Project ASLAN promises easy to install, fully documented and stable self-driving software for specific low-speed urban autonomous applications

Practitioner Portal

EBooks

More EBooks

Upcoming Webinars

Experts in AI

Partner Perspectives

content from our sponsors

Research Reports

9/30/2019
More Research Reports

Infographics

AI tops the list of most impactful emerging technologies

Infographics archive

Newsletter Sign Up


Sign Up