ChatGPT Consumes 500ml Of Water To Answer 20 Question.
In the world of artificial intelligence (AI), ChatGPT and Google’s Bard are making headlines for their impressive capabilities.
However, a recent study titled “Making AI Less Thirsty” sheds light on a lesser-known aspect of these AI models: their water consumption. This article delves into the research and its alarming findings, exploring the environmental impact of AI training and the steps we can take to mitigate it.
The Study and Its Findings
The study, conducted by researchers from the University of Colorado Riverside and the University of Texas Arlington, investigates the environmental impact of AI training, which demands massive amounts of electricity and water. Water consumption is especially crucial as it cools the data centers housing these AI models, preventing overheating and ensuring smooth operation.
One of the study’s most startling revelations is that Microsoft, in partnership with OpenAI, consumed a staggering 185,000 gallons of water solely for training GPT-3. This is equivalent to the water required to cool a nuclear reactor, or the amount needed to produce 370 BMW cars and 320 Tesla electric vehicles. The research also highlights that if Microsoft had trained GPT-3 in its larger Asian data centers, the water consumption would have tripled.
ChatGPT’s Water Consumption
In addition to GPT-3’s training, the study uncovers that ChatGPT “drinks” a 500ml bottle of water for a simple conversation involving 20–50 questions and answers. While this may not seem significant, the combined water footprint of ChatGPT becomes substantial when considering its billions of users worldwide.
Where Does the Water Go?
The water used in AI training is sourced from rivers, lakes, and other freshwater sources. The study focuses on water consumption, which occurs when water evaporates within data centers. Fortunately, this water is not lost forever; it is released into the air via cooling towers and eventually returns to the earth as rain. Nevertheless, the water used in AI training must be exceptionally clean and fresh to prevent corrosion and bacteria, adding to the overall environmental impact.