Algorithmic machine learning systems are being lauded as an ‘integral’ part of the new clean energy economy, but critics see them as an energy-intensive distraction
By Arthur Neslen
BRUSSELS – (Thomson Reuters Foundation) – As climate change intensifies the devastation from storms, wildfires and droughts, artificial intelligence (AI) and digital tools are increasingly being seen as a way to predict and limit its impacts.
Governments, tech firms and investors are showing growing interest in machine-based learning systems that use algorithms to identify patterns in data sets and make predictions, recommendations or decisions in real or virtual settings.
In June, the Rise Fund, an impact investing arm of private equity firm TPG, invested $100 million in a data and AI-driven “nowcasting” system devised by Kentucky-based startup Climavision to predict weather patterns with granular accuracy.
And an intergovernmental roadmap on AI’s role in fighting global warming is due to launch at November’s COP26 climate summit in Scotland.
But AI can also be highly energy-intensive and environmentally damaging, say critics who warn that the tech could be a costly distraction from more effective ways of tackling climate change.
How can AI help combat climate change?
AI applications could also help design more energy-efficient buildings, improve power storage and optimise renewable energy deployment by feeding solar and wind power into the electricity grid as needed.
On a smaller scale, it could help households minimize their energy use – automatically switching off lights not in use or sending power from electric vehicles back into the grid to meet anticipated demand.
By 2030, the tech could help cut global greenhouse gas emissions by 4%, according to a recent study by accounting firm PricewaterhouseCoopers for Microsoft, which is developing machine learning products for the climate change market.
Peter Clutton-Brock, co-founder of the Centre for AI and Climate (CAIC), a Britain-based think tank, said the technology was “pushing back boundaries” for climate modelling.
AI can process huge amounts of unstructured data like pictures, graphs and maps, opening “huge possibilities for understanding the dynamics around sea level rise and ice sheets,” he told the Thomson Reuters Foundation.
Who will be able to use it?
The high cost of AI computational resources has pushed research largely into the private sector, where the market is “extremely vibrant,” according to Chris Goode, Climavision’s CEO.
Climavision’s system uses a high-resolution radar network along with data from satellites and high-altitude weather balloons to fill in what the company says are “hundreds of gaps” in existing weather forecasting networks.
Energy and transport businesses, farmers, even the U.S. military will have access to “real-time elements of the atmosphere, what’s happening at this very moment, because it’s updated on a second-by-second mode,” Goode said.
Ari Cohen, the external affairs director for TPG, which last week announced a new $5.4 billion Rise Climate fund to invest in “climate solutions around the world”, said the AI market was likely to grow, as countries and corporations transition toward low-carbon energy.
From electricity grids to smart appliances, “data and AI-driven software will be integral to predicting market behaviour, balancing operations in real time and maximising energy yield,” he said in emailed comments.
The storage and processing of data needed to fully train a large algorithm can consume huge amounts of energy – as much as 626,000 pounds (284,000 kg) of carbon dioxide, according to a study by the Massachusetts Institute of Technology.
That is the equivalent of nearly five times the lifecycle emissions of an American car, researchers said.
Data centres processing and storing data from online activities, such as sending emails and streaming videos, already account for about 1% of global electricity use, according to the International Energy Agency.
And some estimates say computing will account for up to 8% of the world’s total power demand by 2030, raising fears this could lead to the burning of more fossil fuels.
“AI is both an enabler and, potentially, a destroyer of the climate fight,” said Virginia Dignum, a professor in social and ethical AI at Sweden’s Umea University.
The need for rare earth metals to make the hardware has a destructive impact, she said, adding that AI itself is “not a magic wand – and not without errors.”
The potential logging of people’s energy use throws up privacy concerns about what Clutton-Brock at the CAIC called the ability to “backtrace data to individuals”.
Biased predictive outcomes are also possible, depending on assumptions used in data inputs.
Climate change should “primarily (be) about those who are responsible for the lion’s share of emissions making significant changes to our lifestyles,” said Kevin Anderson, a professor of energy and climate change at universities in Britain, Sweden and Norway.
“Relying on technology to solve the problem in the future – and in doing so, salving our consciences in the near- to medium term – is far from sufficient,” added Anderson, who also works with the Tyndall Centre for Climate Change Research, a partnership between five universities.
Ultimately, the uses of AI in fighting climate change could depend on how it is regulated across borders.
Privacy issues are “very specific to the type of technology that we are using at this moment,” said Dignum, who sat on the European Commission’s High-level Expert Group on Artificial Intelligence last year.
“If we add other types of algorithms which are less heavily dependent on personal data, then that balance between energy consumption and privacy becomes less of a dilemma,” she said.
Such algorithms would soon be “as efficient” as types currently in use, she added.
The AI Now Institute, a research centre at New York University, has proposed green AI principles requiring full transparency on a tech company’s carbon, energy and environmental impact and the integration of tech and climate regulation.
The European Union has signalled that it, too, is considering these issues.
For the moment though, the threats posed by climate change ensure that efforts to find a fix using AI will continue, said Clutton-Brock.
“The opportunity is bigger than the risk,” he said.