Leader in Technology and Analytics @ McKinsey | Mathematician, Statistician and Psychometrician | Digital and Analytics Leader | Author and Teacher | Coder, Engineer, Architect
This piece of research looks at the impact of AI on global water withdrawal, caused by the evaporation of water used to cool AI-related data center activity. Two interesting conclusions: 1. It estimates that 700,000 liters of water was evaporated in training GPT-3. 2. It extrapolates that global AI demand in 2027 may lead to 4.4-6.6 billion cubic meters of water withdrawal, which is half the current water withdrawal of the United Kingdom. Link to full paper --> https://lnkd.in/gWCMUfF6 #ai #machinelearning #analytics #datascience #sustainability #peopleanalytics #python #rstats
Compare this to the human brain, which is far more impressive than any supercomputer or AI. The human body consumes about 100W and 2.7-3.7L of water a day.
So it's going to start raining more?
Damn, there definitely needs to be a better environmental approach to both powering AI and retaining water
Now I understand why this May was the rainiest May ever in NL.
Wow very interesting! This is another data layer that can be added to the model https://pozibl.com/work/data-center-site-selection/ Would love to get your view on this Keith McNulty
Can they use recycled water?
Senior AI Engineer
1moNow imagine you could use a good part of that warm water as either heating (think home hearing, defrosting sidewalks, warm pools, ...), or simply for power generation. One key point missing is people tend to focus on one thing (eg computing power), rather than on what could be recovered (and reused).