Each ChatGPT search is estimated to consume 5x more energy than a Google search. Beyond individual use, the continual demand for newer iterations means companies are constantly training new models - training that requires huge amounts of energy. Specifically, the amount of electricity consumed results in both substantial CO2 emissions and water consumption. 1, 3
"Scientists have estimated that the power requirements of data centers in North America increased from 2,688 megawatts at the end of 2022 to 5,341 megawatts at the end of 2023, partly driven by the demands of generative AI. Globally, the electricity consumption of data centers rose to 460 terawatts in 2022. This would have made data centers the 11th largest electricity consumer in the world, between the nations of Saudi Arabia (371 terawatts) and France (463 terawatts), according to the Organization for Economic Co-operation and Development." 1 This trend is projected to continue, leading to data centers being 5th on the global list (between Japan and Russia) by the end of 2025. 1, 2
While the increased use of electricity is often highlighted, the amount of water consumed to cool data centers and computing equipment also has significant environmental impacts. It's estimated that 2L of water are required for each kilowatt of energy the data centers consume. To put this in direct perspective: ChatGPT needs a 500-ml bottle of water to answer 20-50 questions (efficiency is influenced by things like outside temperature, etc). 4,5
Data centers use purified water free of bacteria - meaning direct competition for water resources our communities rely on. Protests have begun in countries like Chile and Uruguay over planned data centers that would tap drinking water reservoirs. In Dalles, Oregon, the city government filed a lawsuit to keep water usage by Google's three data centers a secret, and when records were made public it became clear the data centers use more than a quarter of the city's water supply. 4
Read more:
Unlicensed Materials & Copyright
AI is notorious for being trained on information that it doesn't legally have access to. From book collections to unlicensed works to art, multiple lawsuits are underway to determine what is "fair use" and what is a copyright violation.
What does this mean?
Data usage - using copyrighted data could violate copyright, and infringement - direct or unintentional - can be costly. A business using AI trained on unlicensed works could be on the hook for willful infringement, which includes damages of up to $150,000 for each instance. (Appel, Neelbauer, Schweidel, 2023)
Attribution - even if training data is used legally, properly citing your sources is always important! There are citation guidelines out for how to attribute information written or created by AI.
AI & Copyright
Copyright protects only original human-authored and created works, including those made with AI assistance, but does not extend to works generated by AI.
Part 1 of Artificial Intelligence Report released by the U.S. Copyright Office in July, 2024.
Keep in mind that what you share with AI becomes a part of the training model. Anything you share in a question with ChatGPT becomes publically available.
What information should I avoid sharing with AI?
