Amidst a terrible debate about the environmental toll of Artificial Intelligence, Google released a new Study It says that its Gemini AI assistant uses minimal water and energy only for each text prompt. But experts say that the claims of the tech veteran are misleading.
Google estimates that an average Gemini text prompt uses approximately five drops, or 0.26 ml, and more electricity, as more electricity, about 0.24 watt-hours (WH), which produces about 0.03 grams of carbon dioxide emissions.
Google’s estimates are less than the previous research on water and energy-intensive data centers that undergo the generative AI model. This is due to improvement in efficiency that the company has created in the last one year. But Google also abandoned the major data points in its study, which made Mithun’s incomplete understanding of the environmental impact, told by experts. Ruckus,
“They are just hiding important information.”
“They are just hiding important information,” called an Associate Professor of Electrical and Computer Engineering at Sholi Rain, University of California, Rivaraside. “It actually spreads wrong messages in the world.” Rain has studied water consumption And Air pollution AI is associated with, and a paper is one of the authors of Google who mentions in his Gemini study.
A major issue experts reported that Google leaves the use of indirect water in its estimates. Its study included water that used to use the server from overheating to use in data centers cooling systems. Those cooling systems have sparked Worry How the data centers can increase water scarcity in dried-affected areas over years. Now, how much meditation is Power data centers may require To accommodate the new AI model. The demand for rising electricity has started a part of new schemes to manufacture Gas And nuclear power plant, which also Drink water in your own cooling systems And To turn on turbine using steamIn fact, A data center consumes most of the water with its electricity using its electricity. – Which Google ignores this study.
As a result, with Google’s water estimate, “You only see the tip of Iceburg,” Originally, “website founder Alex de Vis-Gao says. Laxist And a PhD candidate at the Vrije University Amsterdam Institute for Environmental Studies, who has studied the energy demand for data centers used for cryptmining and AI.
Google left another important metric when it comes to electricity consumption and pollution. The paper only “share”Market-based“The measurement of carbon emissions, which takes into account a company that makes a company to support renewable energy growth on the power grid.
Another holistic approach will also have to be included “location based“The measurement of carbon emissions, which considers the effect that is near a data center, operates keeping in mind the current mixture of clean and dirty energy of the local power grid. The location-based emissions are usually more than market-based emissions, and provides more insight into a company’s local environmental impact.” Standard set by Greenhouse Gas Protocol,
Google paper cites previous research Rain And Day Vis-sing And argues that it can provide a more accurate representation of environmental impact than other studies based on modeling that leads to a lack of first-sided data. But Rain and D Vris-Gao says that Google is comparing an apple-to-orange. The previous work was based on average rather than the medium that uses Google, and the Rain blames Google for not sharing numbers (Word Count or token for text signals) how it came in the middle. The company writes that it makes its estimates the base on an average prompt, which is on an average signal to prevent outlers using energy more than the outcomes.
“You only see the tip of the iceberg, basically.”
When it comes to calculating the consumption of water, Google says that the discovery of the PROML per text Prompt K. “is the lower quantity order” than the previous estimates “the order of the lower quantity” which reached the 50 ml in the rain research of the rain. This is a misleading comparison, Rain again said, as the paper rain co-writer takes into account the total direct and indirect water consumption of a data center.
Google has so far presented its new paper for colleague reviews, although spokesperson Mara Harris said in an email that it is open to do so in future. The company refused to answer on records in the list of other questions RuckusBut the study and the blog simultaneously say that Google wants to be more transparent about its AI chatbot water consumption, energy use and carbon emissions and offers more standardized parameters to measure environmental impacts. The company claims that it proceeds compared to previous studies, which supports the infrastructure in a data center like factoring and cooling systems in energy used by machines.
VP/GM of AI and Infrastructure for Google Cloud, and Google Dipmind and Google Research Chief Scientist, Jeff Dean, Jeff Dean, Jeff Dean, Jeff Dean, Jeff Dean, Google Dean, Chief Scientist of Google Dean, Google Dean, Chief Scientist of Google Dean, “As long as we say in ahead of our efficiency in the years,” The reforms are committed to continue, “AMIN Wahdhat, and Jeff Dean, AMIN Wahdhat, and Jeff Dean, said AI and VP/GM of the Infrastructure for Google Cloud, and Google Dipmind and Google Deepender, and Jeff Dean. Blog,
Google has claimed that the energy efficiency of a Gemini text prompt between May 2024 and May 2025 has improved significantly, leading to a reduction of 33x per indication power consumption. The company says that the carbon footprint of an average prompt fell from 44x in the same time. Those benefits also explain why Google’s estimates are now less than previous years of study.
However, zoom out, and the real picture is more severe. Efficiency benefits can still be used more pollution and more resources – an unfortunate – an unfortunate The event is known as Jevons ContraditionGoogle’s so-called “ambition-based carbon emissions” increased by 11 percent last year and 51 percent since 2019 as the company continues to carry forward AI aggressively, according to its latest, AI aggressively forward. sustainability report(The report also noted that Google began to exclude some categories of greenhouse gas emissions from its climatic goals this year, which says that “peripheral” or the company’s direct control.)
“If you look at the total number that is posting Google, it is really really bad,” says D Vries-GAO. When it comes to estimates that it has been released on Gemini today, “This is not telling the whole story.”
