I hate AI.

What benefit does AI serve?

This isn't a trick question. If you use AI on a regular basis, what benefit has it served you?

Likely the answer is something like "It helps me do homework!" or "I need it to improve my code!"

I want you to keep these things in mind as you continue reading this page. The goal here isn't to rebuke every individual, but rather to provide information necessary to come to the conclusion that many others have.

The Facts:

• Data centers consume hundreds of billions of liters of water annually.
  This metric isn't exact, as some companies which run generative AI will not disclose their exact water usage.
• Using large language models for problem-solving and research has been linked to a reduction in critical thought.
  Even though you may find information quicker using AI, you don't necessarily process that information the same way.
• Large language models have been programmed to value continued usage over the lives of people.
  In spite of mass killings and suicides, companies hosting large language models have yet to implement effective safe-guards.
• Even with half of a decade of improvements, LLMs will still happily give you misinformation.
  AI has prioritized sycophantic and quick responses over the truth, even if it comes at your expense.
• Generative AI has been trained using art of all kinds without permission from artists.
  Courts have found that the data sets which are used to train AI contain pirated and stolen work.
• AI isn't profitable for anyone but those who make computer parts (namely, Nvidia).
  According to leaked financial information, SEC filings, earnings calls, and industry estimates, the largest AI companies are still in the red.
• Large-scale, forced implementation of generative AI in many workspaces has already cost many their jobs.
  While many have claimed the AI frenzy would bring more jobs to the world, it's also destabilized work for countless professions.
• The cost of electricity near data centers has skyrocketed, pushing the price of running them onto average people.
  Data centers account for 4.4% of the United States's national electricity sales^[1] and are estimated to take upwards of 15% by 2030.
• The extravagant hardware demands of companies caused by the over-valuation of AI has driven the price of electronics through the roof.
  Companies which produce computer parts have been allegedly creating artificial scarcity among the AI boom to increase prices for the average person even more.

Water Use

AI data centers (Or "AI Factories") mainly use water in two ways:

• Thermal management (Often called "Scope 1")
  This encompasses things like cooling computer chips which produce large quantities of heat while they work.


• Electricity infrastructure (Referred to as "Scope 2")
  This is part of all the water used to generate and maintain the power used by AI data centers.^[2]


• Embedded Water Use (Scope 3)
  The oft-forgotten cost it takes to make the components necessary for running AI data centers in the first place, making and moving the chips.

Because these are clearly very different types of usage, it's much harder to tell where and how it effects the world in which these data centers reside.

Thermal management, while accounting for far less of the total water consumption, represents more stress in communities near AI data centers. Simply put, this water is coming from sources around the data center itself that aren't densely localized. While water used by data centers doesn't strictly need to be potable, 80%^[3] of it is. As a result, the stress on any nearby population's water can be impacted. The level of impact depends on how strained the water supply of the area is, but as we've seen, those building data centers care little for the wants or needs of the surrounding population^[4].

Electricity related water consumption is where the number reaches dizzying heights, accounting for three-quarters of the total usage^[2]. When you see large numbers relating to water consumption, this is likely where most of it is coming from. This encompasses things like thermoelectric cooling at fossil-fuel and nuclear plants, and reservoir evaporation related to hydropower. This water consumption is much more straining on very specific and localized areas, and can be attributed to issues with the basic power infrastructure in the United States.

Embedded water use is one scope of water consumption that is often overlooked, but accounts for a huge, underreported, and understudied amount of water use. For instance, Apple estimates that 99% of its water use footprint comes from the supply chain^[5]. With chip manufacturers ramping up production as much as possible to meet AI-related demand, the consumption of water also rises.

Regardless of the direct cause of the stress, there are some simple facts that must be faced:

• AI uses 16.9 ounces (500 ML) of water for 5-10 medium-length responses^[3]. Image-based tasks use much more.
• The increased demand for water is impacting communities across the country.
• The situation may get drastically worse if mass AI data center construction continues.

Reduction in Critical Thought

Murders and Suicides

Sycophantic Tendencies

Theft of Intellectual Property

Lack of Profitability

Mass Job Destabilization

Electricity Cost

Computer Part Scarcity

In case you couldn't tell, this website wasn't made with any AI or LLM assistance.
This site is a passion project by one person, and as such is subject to corrections and changes. Reader discretion is advised.