AI Technologies Like ChatGPT 'Drink' Massive Amounts Of Water To Thrive: Study

ChatGPT and related artificial intelligence technologies thrive on large amounts of data to become smarter. But did you know that such AI algorithms also consume tremendous volumes of water as well?

According to researchers from the University of Colorado Riverside and the University of Texas Arlington, a lot water goes into keeping the data centres for such AI tools cool. In a paper titled "Making AI Less Thirsty" that is yet to be peer reviewed, scientists noted that just to train GPT-3 alone required 700,301 litres of water. Crazy, right?

Unsplash

GPT-3 takes as much as water a nuclear reactor

This is the same amount of water, as the researchers found out, that would be required to cool a nuclear reactor. The paper claims that the water required by Microsoft to cool its data centres in the US to train GPT-3 was enough to "70 BMW cars or 320 Tesla electric vehicles."

If Microsoft's data centres in Asia had been used instead, "these numbers would have been tripled." Think that's nuts? Wait till you hear this - apparently, "ChatGPT needs to 'drink' [equivalent of] a 500 ml bottle of water for a simple conversation of roughly 20-50 questions and answers."

Unsplash

Also read: Researchers Create 'DetectGPT' To Determine Which Papers Are Written By ChatGPT

"While a 500ml bottle of water might not seem too much, the total combined water footprint for inference is still extremely large, considering ChatGPT¡¯s billions of users," the authors noted.

In essence, ChatGPT is gulping down a lot of water to keep our curiosity alive. Researchers have a suggestion for companies like Google and OpenAI - they think such companies should take "social responsibility and lead by example by addressing their own water footprint." Accountability from tech companies? That's quite the demand.

Also read: ChatGPT Maker Releases New Tool That Detects AI-Generated Text To Limit Abuse

Unsplash

Regardless, ChatGPT remains a formidable tool to perform mundane tasks like writing e-mails and understanding new concepts in simple words - of course, it also has its limitations, like the times it sometimes shows false information.

What do you think about AI's thirst? Let us know in the comments below. For more in the world of technology and science, keep reading Indiatimes.com.