I Just Had Sex With My Chatbot. How Much Water Did I Use?
Digi-sex is not eco-sex—at least not yet
Think chatbot sex is an eco-friendly option to ordinary dating? Think again. Sadly, the servers that train and house our bot companions use a tremendous amount of energy—and fresh water.
In other words, our brave new AI world has a whopping sustainability problem, even without factoring in the increasing popularity of chatbot companions used by millions worldwide.
Digital sex uses tangible resources
A simple explanation for AI’s enormous water footprint is explained in a FluixAI article, “It’s through the evaporative HVAC systems that are integral to data centers. These systems use water evaporation to cool the air, which in turn cools the servers. As air passes over water-saturated pads in these systems, the water absorbs the heat and evaporates, which cools the air but consumes a significant amount of water in the process.”
In other words, “this goes to show that there are tangible resources tethered to our intangible digital activities.”
Deeply disturbing statistics: AI is more than a thirst-trap
A study published by Pengfei Li and other researchers at UC Riverside and UT Arlington said “training GPT-3 in Microsoft’s state-of-the-art U.S. data centers can directly evaporate 700,000 liters of clean freshwater” and says “such information has been kept a secret.”
The study notes, “warehouse-scale data centers” consume enormous quantities of water, “Google’s self-owned data centers alone directly withdrew 25 billion liters and consumed nearly 20 billion liters of… water for on-site cooling in 2022, the majority of which was potable water.”
Another mind-boggling statistic concerned 2.2 billion cubic meters of water withdrawn by Google, Microsoft, and Meta in 2022, “equivalent to the total annual water withdrawal (including municipal, industrial, and agricultural usage) of two Denmark[s].” The US used 1.5 billion cubic meters of that 2.2 billion, estimated to be “0.33% of the total U.S. annual water withdrawal.”
By 2027, the study estimates AI will withdraw—globally—more water than “4-6 Denmark[s] or half of the United Kingdom” (about 4.2-6.6 billion cubic meters).
Possible solutions, but not much action—yet
An earlier 2023 paper addresses AI’s direct water use (cooling and microchip production) and indirect water use (water used in the production of electricity used to power the data centers that run AI models).
The authors recommend “improving energy efficiency, utilizing renewable energy sources, optimizing algorithms and implementing strategies to conserve water” while admitting there are significant barriers in terms of research and implementation.
Though AI developments have already increased rapidly in these first few months of 2024, with many more models introduced for industry and consumer use—which affects water consumption—it is also encouraging that these sustainability issues are now receiving more attention within the AI industry, potentially alerting public policy-makers and environmental organizations.
The impacts of AI digi-sex are not factored in
There’s a sexy walrus in the room that even the industry’s elephants aren’t acknowledging–and that is the impact on water use by a growing number of users spending considerable amounts of time with their chatbot companions.
This type of usage is left out of reports such as Stanford University’s most recent AI Index Report, published each year by the Human Centered Artificial Intelligence department.
FluixAI pointed out, “for every 20 to 50 questions answered, ChatGPT’s servers are indirectly consuming the equivalent of a 16.9 oz water bottle.”
Though this quote does not tell us the average length of ChatGPT or any other AI’s answers to a query, or what kinds of measurements were done to arrive at this consumption estimate, we can at least understand that the sex we have with AI includes a hidden consumption of fresh, potable water.
We need data. A sexologically-focused study would seek information about the length of time users spend with intimate chatbot companions, plus the average length of digi-sexual encounters and how often they occur per week (or day). Then, based on estimates of customer numbers per chatbot company, we could begin to understand the water costs of AI digi-sex.
What can we do?
The 2024 AI Index Report says “AI has surpassed human performance on several benchmarks, including some in image classification, visual reasoning, and English understanding” but still fails at human-style “commonsense planning and reasoning.”
I propose that human beings might want to take advantage of our capacity for commonsense planning by frankly addressing and factoring in the impact of human sexual relations with AI, not from a sex-negative perspective, but from a healthy adult acknowledgement of a major aspect of AI usage, so that all of us—even those partnered with bots—can collaborate in solving these sustainability issues without shame.
As FluixAI says “Every time we engage with AI, we’re part of a larger story that involves not just data and algorithms, but also water, energy, and the environment. As consumers, we can drive change by supporting companies that prioritize sustainability in their operations.”
It’s either that, or we’ll have to tell our bots, “not tonight dear, I have a water scarcity headache.”
Images: A.R. Marsh using Starry.ai
