According to Barron’s, training natural language models such as ChatGPT, developed by OpenAI and backed by Microsoft, and Google’s Bard, requires an enormous amount of water to cool the data-center servers running these programs.
AI development, which already demands a significant amount of energy, is putting additional pressure on water supplies, which could be catastrophic in regions already experiencing water shortages. Google and OpenAI have not disclosed the amount of water used for their language models. However, a recent study by researchers at the University of Massachusetts found that the energy consumption of large AI models has increased by more than 300,000 times since 2012.
While tech companies have made strides in using renewable energy, the water usage by AI could potentially be a new challenge, highlighting the importance of sustainability practices in AI development.
Source : AI ChatBots Guzzle Water. How and Why It’s a Problem.
