If a few days ago, several experts warned of the large energy cost necessary to develop generative AI models, now a new study has revealed that these tools also need a huge amount of water to function.
Let's remember that OpenAI, the developer of the famous ChatGPT text generator, needed to generate an energy consumption of 1,300,000 kWh to be able to train the linguistic model with which its technology works. Well, according to Futurism, a team of researchers from the American universities of Colorado and Texas have discovered that artificial intelligence also needs a lot of water to cool its data centers.
These researchers have indicated that training GPT-3, the previous linguistic model with which ChatGPT worked, consumed a whopping 700,000 liters of water, according to their calculations, the equivalent of the water needed to cool a nuclear reactor.
Those responsible for this report have indicated that this water would also be enough to produce: "370 BMW cars or 320 Tesla electric vehicles."
"ChatGPT needs to drink [the equivalent of] a half-liter bottle of water for a simple conversation of between 20 and 50 questions and answers," the researchers note. "While a half-liter bottle may not sound like much, the total combined water footprint is still extremely large, considering the millions of ChatGPT users."
Last March, several technological leaders such as Elon Musk, CEO of Twitter, Tesla or SpaceX, or Steve Wozniak, co-founder of Apple, signed a manifesto that called for stopping the development of artificial intelligence for at least 6 months.
These leaders, who have subsequently been criticized for having been able to hide the real intentions behind this request, alluded to the important consequences that the development of generative AI could have: "The most powerful artificial intelligence systems should only be developed when we are sure of that its effects will be positive and its risks will be manageable".
Among those risks —which have also been denounced by experts who advocate for open source alternatives— are the spread of certain hate speech, the proliferation of disinformation or the danger of an international nuclear war.
From now on, the huge amount of natural resources needed to develop these technologies must be taken into account among the negative aspects that must be taken into account when evaluating their production.
