AI benefits won't outweigh environmental cost, insist opponents

Google is among the big-tech firms to trumpet the possibility of artificial intelligence (AI) helping to solve the climate crisis. But green groups dismiss this idea as nonsense and say that, by contrast, AI prevalence will vastly increase energy use from data centers and spread climate denial falsehoods.

Recent research authored by Alex de Vries, founder of digital trends research and consulting company Digiconomist, suggests that within three years AI servers could consume as much energy annually as that produced by the whole of Sweden or the Netherlands.

He told CGTN that data centers currently account for one percent of global electricity consumption but that the figure could quickly quadruple. 

According to De Vries: "Combined trends in AI related electricity consumption, digital currency mining, and also likely growth in regular data center hardware energy consumption will lead to data centers consuming 3 to 4 percent of our total global electricity consumption."

AI and crypto industries are served by data centers - huge, often remote, warehouses full of very large computers that run complex code continually. 

However, despite calling for transparent regulation of AI, the United Nations has joined big tech firms in touting the potential benefits of AI. The technology's advocates see the use of AI as helping to ameliorate global heating via tools tracking and extreme weather events, while identifying pollution leaks.

Last October, UN Secretary-General António Guterres said: AI "could supercharge climate action and efforts to achieve the 17 Sustainable Development Goals (SDGs) by 2030." 

A 2023 Google report stated that AI could cut global emissions by as much as 10 percent within seven years, with chief sustainability officer Kate Brandt saying "AI has a really major role in addressing climate change." 

But De Vries disagrees. "It's impossible to unite AI and sustainability simply due to - from an AI perspective - (saying) bigger is always better and that these models perform better the bigger you make them. It just means more resources and more energy required to run them."

He insists that from an environmental point of view the world needs "to put in fewer resources," adding that increasing the efficiency of individual services might also push up demand for them. "We have seen this happen pretty often in the past (where we) end up needing more resources than we were starting out with, so it can actually be a double negative," he said.

READ MORE

Where is Flight MH-370?

Meet the 'Chinosaurs'

Why young Chinese are returning to the country

A recent report published by the Climate Action against Disinformation coalition expressed similar views. It quoted OpenAI's CEO Sam Altman reportedly conceding that AI will use "vastly more" energy than people expect. The report authors also asserted that statistical models insisting that it will cut energy use are biased.

According to the report: "Such statistics are only estimates, because AI companies continue to withhold most of the data. Transparent reporting would allow researchers to know if the use of AI systems offset any potential savings. 

"For example, if the AI industry improves data center energy efficiency by 10 percent but also doubles the number of data centers, it would lead to an 80 percent increase in global carbon emissions."

De Vries says he recognizes the potential advantages of AI but maintains that policymakers must look deeper into the issue. He argued "there might be some benefit, but looking at the whole picture….I don't think that the benefits will outweigh the extra cost of it." 

Subscribe to Storyboard: A weekly newsletter bringing you the best of CGTN every Friday