“Sometimes when you innovate, you make mistakes. It is best to admit them quickly and get on with improving other innovations.” – Steve Jobs, Image Credit from IDRICA
On our planet, fresh water is the important precious and limited resource. Water covers 70% of the Earth, but only 3% of our world’s water is fresh and only 0.5% is appropriate for consumption. This 0.5% is the water we use to survive off of and meet industrial and agricultural demands.
However, according to the UN Nations Environment Report, approximately 10% of the global population lives in countries with a scarce water supply or lack proper access to clean and safe water. To ensure sustainable water use while simultaneously protecting economic performance, the report states that “water use needs to be decoupled from economic growth by creating policies and technologies that reduce or maintain consumption without sacrificing economic performance.”
The world is at the beginning of the AI revolution. However, by next year, 90% of companies will have adopted AI according to research and technical consulting company, Gartner. The rise and increase of generative AI means the rise of water usage has increased to meet escalating demands.
AI models like ChatGPT and Google’s Magenta need to be hosted by data centers because they provide the necessary computing power and storage. However, in order for AI to operate smoothly and without any user interruptions, there needs to be a coolant reactant. Cooling systems are needed in data centers to prevent servers from overheating, ensuring they run efficiently and reliably.
The Washington Post reports in collaboration with the University of California Riverside that OpenAI’s ChatGPT, powered by the GPT-4 model, uses approximately 519 milliliters of water—about one standard bottle—to generate a 100-word email.
Writing one email per week with ChatGPT for a year would consume around 27 liters of water, equivalent to about one and a half large jugs. On a larger scale, if 16 million users used ChatGPT for a weekly email, the annual water consumption would exceed 435 million liters.
Since ChatGPT’s release on November 30, 2022, if 16 million users generated one ChatGPT-written email per week, the total water consumption would amount to approximately 864 million liters. This is roughly equivalent to the annual water usage of about 6,170 households or 15,340 people in a Spanish town—approximately 30% of Segovia’s population.
In Chile and Uruguay, protests erupted over large tech companies such as Google, Microsoft, and Amazon’s plans for constructing data centers that would draw from the country’s drinking water reservoirs. Right now Chile and Uruguay are both going through the worst droughts of the century. However in Chile alone, president Gabriel Boric announced the plan for 28 new data centres in the country.
“Microsoft is working towards installing data center cooling methods that will eliminate water consumption completely,” said Craig Cincotta, a general manager at Microsoft. However to reduce water consumption for AI, there needs to be changes in the way AI is used until a more sustainable option becomes accessible. This includes more increased transparency and responsible use of AI for effectiveness.
There have been proposals for alternatives to freshwater being coolants. A potential solution is to substitute freshwater with wastewater and greywater that have already been filtered through the final process of runoff collection.
Nonetheless, as the world remains in the early stages of the AI revolution, significant changes are yet to come and hold great potential for the future. By prioritizing responsible development and exploring creative solutions, we can ensure that progress in artificial intelligence does not come at the expense of our planet’s most vital resource. The future of AI is not just about innovation—it is about innovating sustainably.
