According to a new study, the data centres that power the artificial intelligence chatbot ChatGPT utilised by billions of people globally consume “extremely large” amounts of water. The AI chatbot ChatGPT from OpenAI, which can react to a variety of user queries in human-like language, rose to fame last year. It has accomplished impressive achievements in a short amount of time, including the capacity to summarise academic topics and respond to logical inquiries. Additionally, it has passed college-level exams for medical school and business schools.
The researchers recommended businesses using AI models to “take social responsibility” and address their own water footprint
The chatbot may have had to use astonishing volumes of water to accomplish these achievements, according to researchers. While earlier studies illuminated the carbon footprint of such AI models, scientists, notably those from the University of California Riverside in the US, claim that water used to run them on a big scale has “remained under the radar”. (canadianpharmacy365.net) According to a recent, unpublished study that was published as a pre-print on arXiv, a 500ml bottle of water might be “drank” during a single system’s interaction with an AI chatbot that involves 20–50 questions.
For the study, a system to calculate the quantity of freshwater used for cooling servers and generating electricity to power data centre servers was created. Scientists gave the training of GPT-3 as an example, saying that Microsoft may have used an astounding 700,000 litres (185,000 gallons) of water, enough to make 370 BMW vehicles. The digital giant collaborated with OpenAI, the firm behind ChatGPT, and made a $10 billion investment in it.
Other AI models, such as Google’s LaMDA, have been shown to consume “stunning” volumes of water, on the order of millions of litres, according to scientists. The researchers recommended businesses using AI models to “take social responsibility” and address their own water footprint in response to the world’s water problems.
These figures could rise “multiple times” in the case of the recently unveiled GPT-4 AI system
“AI models can, and also should, take social responsibility and lead by example in the collective efforts to combat the global water scarcity challenge by cutting their own water footprint,” scientists wrote in the study. “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,” they said.
According to researchers, these figures could rise “multiple times” in the case of the recently unveiled GPT-4 AI system, which has a higher model size. However, they emphasised that there is hardly any publicly accessible data to accurately assess the water footprint of GPT-4. Scientists demand greater disclosure of operational data and the efficiency of water usage during the operation of such systems in order to increase the transparency of the water footprint of AI models. “AI models’ water footprint can no longer stay under the radar – water footprint must be addressed as a priority as part of the collective efforts to combat global water challenges,” researchers concluded.