400 litres of water for a movie
“If we are lucky and humanity starts to take the climate crisis seriously, we will probably end up spending even more time interacting with the world remotely through digital platforms”: Vili Lehdonvirta (keynote speaker on October 22 nd to an INCA event at the University of Bologna to talk about “Cloud Empires”) seems to be winking at the idea that remote interaction would have a better impact on the climate crisis we are experiencing. Is it true? The answer is articulated. Attention must be high.
We must start with a statement that is as banal as it is fundamental: invisible (or, better said, hidden) does not mean that it does not exist. The world of data centres that support digital platforms behind the scenes, and even more so the supercomputers that support artificial intelligence, are far from “neutral” in terms of their environmental footprint. Back in 2015, researchers published a paper in the journal Sustainability looking at the impact of data centres on water consumption for cooling. Results are impressive. 400 litres of water for a streaming movie. 20 litres for a TikTok-style social video. A 5-minute shower uses 80 litres. These are the proportions. And with AI, things seem to be getting worse: 2 litres every mail –size text produced by ChatGPT.
Indeed, recently Alex De Vries estimated that AI overviews, which use generative Artificial Intelligence to respond to users, consume ten times the energy of a traditional Google search. As a matter of fact, studies point to an exponential increase in energy and water consumption due to AI: the more complex they become, the greater their energy requirements. This consumption threatens to undermine progress in the use of renewable energy sources and further increase the use of fossil fuels or nuclear energy (as it is the case of the Three Mile Island nuclear powerplant in Pennsylvania, which should open back in 2028 for answering Microsoft requests). In addition to climate issues, social issues have also been raised: in some urban projects, AI is at the heart of decision-making processes that profoundly redefine governance, while algorithms replace social processes. Finally, many analyses have highlighted ethical aspects, particularly related to data mining and the use of information collected and used by AIs.
On the other hand, artificial intelligences play – or could play – an important role in various fields, supporting processes that address the causes of global warming. Climatologists hope to be able to study atmospheric phenomena more accurately, while engineers want to use AI to make energy use more efficient, to make renewable energy sustainable, to improve the use and recycling of raw materials, or to manage traffic in urban areas.
All in all, if AI can offer new opportunities, but at the same time its consumption can become unsustainable, the question to be addressed is about its forms of governance: when does the energy consumption of its use produce a social benefit? Who controls its algorithms and data? After all, as Sasha Luccioni told The Guardian, “if you want to save the planet with AI, you first have to consider its environmental impact. It makes no sense to burn forests to feed systems that track deforestation”. As for platforms, the question for AI is to examine who is pulling the strings: to break out of the circle of exclusive direction by large corporations and push towards models that are oriented towards social and collective interests.
Mattia Frapporti, University of Bologna
Marco Palma, University of Applied Sciences and Arts of Southern Switzerland