The Environmental Impact of Artificial Intelligence: The Tech-Fossil Fuel Connection

The Environmental Impact of Artificial Intelligence: The Tech-Fossil Fuel Connection

Photo Credit: Adib Hussain on Unsplash (edited by Ashley Moniz)

Photo Credit: Adib Hussain on Unsplash.

Ali MesbahianAli Mesbahian is an IPilogue Writer and a 2L JD Candidate at Osgoode Hall Law School.

 

The ever-growing reliance on artificial intelligence (AI) in our everyday life and industry is an undisputable condition of our time. Whether we speak of gaming, speech and facial recognition, smartphones, medical research, agriculture, trading and investment, cybersecurity, or resource extraction (and of course, much more)—few, if any, sectors fall outside the purview of AI. As with any emerging technology, debates as to whether AI is a “net good” are abundant. This brief article focuses on this issue with respect to AI’s environmental impact.

For starters, it is important to acknowledge AI’s potential in combatting climate change. Among other things, AI can lead to better climate predictions, create virtual simulations that demonstrate what a given area would look like after the impacts of climate change, and help track the source of carbon emissions for regulation purposes.

On the other hand, AI requires infrastructure that consumes a great deal of energy. A 2019 research study conducted at the University of Massachusetts Amherst shed light on the enormous scale of this consumption: the energy required to train a single natural language processing (NLP) model leaves a carbon footprint of roughly 300,000 kg—the equivalent of 125 round-trip flights between New York and Beijing (for a fascinating map on the human and environmental costs of AI, see here). Of more concern, “the computations required for deep learning research have been doubling every few months, resulting in an estimated 300,000x increase from 2012 to 2018.” But what contributes to this increasing energy-intensive dynamic? Consider the following two points .

First, some researchers and academics have raised concerns about the AI community’s hyper-focus on their models’ accuracy and “state of the art results”—which come at the expense of cost and energy-efficiency considerations. Accordingly, calls are being made to research “Green AI”  that not only incorporates the energy consumption levels of a given AI model in its evaluative criteria, but also factors in the renewability of energy-sources and the extent to which a given model’s research results can be reproduced for future research.

Second, a 2020 Greenpeace report, illustrates the close connection between tech and fossil-fuel industries. For instance, while Microsoft has vowed to become “carbon negative” by 2030 in order to counteract its contribution to environmental damage, it also offers AI capabilities to oil and gas companies such as ExxonMobil “in all phases of oil production.” Microsoft is not alone in signing these kinds of lucrative contracts; it’s joined by companies such as Amazon and Google. This casts huge doubt on the achievability and commitment of tech firms’ own climate goals. As was the case when the Coalition for Tech Workers joined the global climate strike, it is important for both civil society and insiders in the tech industry to pressure corporate executives to stop assisting the extractive activities of the fossil fuel industry and be more aggressive in reducing their own carbon footprints.

It is important to mention that the success of initiatives aiming to reduce the negative impacts of AI depend on the regulatory climate, both domestically and internationally. In this regard, I end with a hopeful starting point: Germany’s supreme constitutional court’s historic ruling in April 2021 that rendered the government’s climate goal to achieve carbon neutrality by 2050 as unconstitutional. The Court found that the government’s policy simply does not go far enough in protecting future generations from the catastrophes of climate change. As a result, the German government is in the process of bringing forward a more ambitious climate plan—which undoubtedly bears regulatory implications for the tech industry.

To learn more about the relationship between artificial intelligence and the environment, check out next week’s Bracing for Impact Webinar: AI’s Dirty Footprint, hosted by IP Osgoode and featuring a panel of leaders in the fields of AI and sustainability.