When contemplating where our computational technology is going, the adage “Just because you can, doesn’t mean you should,” comes to mind. There are many types of dangers relating to the increasing digitization of our world, that activists, journalists and scholars are drawing our attention to. The environmental risk to our planet is one of those. Evidence of the damage that global warming is causing is all around us. July of this year was the hottest month on record.
Cloud computing, which takes place in data centres all over the world, uses a tremendous amount of electricity. Much of this electricity comes from the burning of fossil fuels. Machine learning, conducted in those data centres, requires very large quantities of data. Processing that data requires a great deal of power, meaning they have a large carbon footprint. A team at the University of Massachusetts, Amherst, described in a recent paper, just how much carbon dioxide we’re talking about. In order to create “virtual assistants” like Alexa, that can understand natural language, as much as 626,155 lb of carbon dioxide can be emitted. That’s about equivalent to the carbon dioxide generated by flying 125 times roundtrip between New York and Beijing.
The growing desire for more data in all areas of our lives, is moving us towards digitizing everything. With the help of machine learning, patterns can be detected which help governments and corporations make decisions. Worldwide data centres currently use about the same amount of power per year as does South Africa – 200 terawatt hours. A Huawei researcher predicts that number will likely increase 4 or 5-fold by 2030. At that point, data centre power consumption will be on par with Japan; currently the fourth largest consumer of energy worldwide.
The good news is that data centres have been turning to renewable energy sources. Improved efficiencies have also been achieved to help lower energy consumption. Tech workers are even helping to push the goal towards a “greener” future. Amazon employees, for instance, participated in the September 20, 2019 global walkout. Their demands of Amazon include a commitment for zero emissions by 2030 and to no longer sell cloud services to fossil fuel companies.
But let’s get back to the ever-increasing demand for more data. We are fed a constant diet of the bright future the advancement of Artificial Intelligence (AI) will bring. Let’s consider the downside. Building more computers in itself requires more energy, as will the increased power demands of more and larger data centres. This is counter-productive to our goal of not heating our planet any further. In addition, let’s not forget that AI is allowing governments, corporations, advertisers, and even cyber criminals to learn more about us than we would like and thus to gain more control over us. Should we not be chasing ecological and social goals first, before anything else? Should we pursue the advancement of AI, if it is against the common good? This takes us back to my original adage, “Just because you can, doesn’t mean you should.”
Meeting the world’s carbon emission goals is not an option. It is a necessity if humankind is to continue on this planet much longer. A plan for reining-in the digitization of our world needs to be a component of achieving these goals.
-blog inspired by a submission in the International Edition of The Guardian, by columnist Ben Tarnoff, entitled “To decarbonize we must decomputerize: why we need a Luddite revolution”–