Everyone is hearing about Artificial Intelligence (AI) and how it is revolutionizing many science fields or how it may come for our jobs in the near future. Well, another issue that needs to be discussed is freshwater use. GPT-3 (the predecessor to ChatCPT) required 700,000 liters (185,000 gallons) of water to properly train before use. That’s enough to fill a nuclear reactor cooling tower, according to researchers at the University of California Riverside and University of Texas Arlington. In a world with diminishing freshwater, that’s a problem.
But it’s not only the initial training, it’s in the use. For an average user interaction with 25-50 questions, the water use is 500 milliliters – about the size of a bottle of water. The culprit is the sheer size of the computing system – 285,000 processor cores (a typical home computer has four cores) and 10,000 graphic cards. All those processors and cards consume a tremendous amount of electricity to operate and require even more power to remove all the heat produced using air conditioning chillers and cooling towers. Data centers typically operate in the 50-80°F range, to keep all the computers cool and operating properly.
So the next time you ask a search engine to find a recipe or to write your next research paper, look at the glass of water next to you and think about the impact. – Steve Terry, DTC HVAC & Refrigeration Instructor
Mechanical PE Exam Prep That Works
Pay Up Front and Save $100!
Test Drive Your Course for Only $50!
Pay-As-You-Go and Start for Only $50!
Don’t miss out on the latest posts (and more)! Follow DTC on:
