Best to be clear right up front: AI makes me ragey. Like, really really ragey. Now, if it’s AI for good (think: to advance medical science), I have no problem with it. Where the rage bubbles up is when AI is rammed down our throats and sold as the ultimate productivity hack when it is absolutely unreliable, invades our privacy, and renders us unable to think critically. I am experiencing this first-hand at my day job and can say unequivocally that my job satisfaction has plummeted because of it.
There are a lot of reasons to hate AI—enough to build a whole mosaic series around <insert thoughtful raised eyebrow…>—but let’s start with water. AI data centres use enormous amounts of water to cool their servers, and if you go further upstream (heh), there is also water used in producing the staggering amounts of energy these data centres consume, and go one step further and there’s all the water needed to produce the servers and their components themselves.
You’ll no doubt have seen reference to how much water is consumed for every AI query. The most-cited statistic is 500mL for every 10-50 GPT-3 queries or the same amount to draft a 100-word email (do we seriously not have it in us anymore to pull together 100 words?). The majority of this water is potable water, and a lot of it gets evaporated when it cools the servers, thus leaving the local water system and meaning companies have to take yet more water to replace what was lost. To top it all off, when the water evaporates it can also concentrate pollutants, making any remaining water returned to the local system highly dangerous.
With the “imminent risk of a global water crisis” looming, it’s safe to say we should be thinking very carefully about how we use the water we have. And yet, there’s a global rush to attract AI infrastructure, with data centres proposed for areas already experiencing high water stress, as is the case in Alberta, setting the stage for major conflict between local communities and data centres. Spoiler alert: the data centres will win because they’ve called dibs on the water, so tough luck if you’re a farmer in the area who needs to irrigate your crops.
The idea for “Thirst Trap” came after reading about a proposed data centre less than an hour from where I live that will consume 39.75 litres of water per second to cool its servers. The article mentioned this was the equivalent of 500 Olympic-sized swimming pools per year, but being someone who isn’t around Olympic-sized swimming pools that often, I had a tough time imagining the scale of it. But 39.75 litres was something I could make more concrete. And so, I started collecting 500mL water bottles, plucking them from recycling boxes around the neighbourhood while out walking our dogs. But 39.75 litres meant I would need 80 bottles. I got to about 25 before I realized I needed to call for reinforcements (aka put out a call on social media). Well, turns out I should have just gone directly to my aunt of Garbage Fairies KW fame. This trio of women have been cleaning up local spaces and turning “trash into cash” to donate to a wonderful local non-profit for several years now. Last year alone they raised $13,000! Anyway, my aunt saw my call, mobilized her fellow Garbage Fairies, and in less than a day had dropped off two huge garbage bags FULL of plastic bottles.
The mosaic itself is fairly simple. A central raised “data centre” sits in the middle, covered in old circuit board and the mouths of 80 plastic bottles. (I rounded up. Technically it should have only been 79.5, but hey, this is art.) Picture these 80 bottles being emptied every second of every day. All around is an expanse of brick and terracotta, which looks particularly arid in contrast to the lush green and blue of the data centre. The name, “Thirst Trap”, does double duty, alluding to AI’s intense water consumption as well as those social media selfies designed to tease and lure you, which parallels what the AI industry is doing by dangling shiny fun features designed to get you hooked.
Will there be more AI-themed mosaics in my future? Probably. Will any of my work be done using AI, whether it’s project planning, writing, graphic design, or anything else? A resounding absolutely not.