Google's AI search summaries use 10x more energy than just doing a normal Google search By THOM DUNN<https://boingboing.net/author/thomdunn> 5:14 AM FRI JUN 28, 2024 https://boingboing.net/2024/06/28/googles-ai-search-summaries-use-10x-more-energy-than-just-doing-a-normal-google-search.html
It seems like every company in the world right now is eager to push their large-language modeling algorithms as a brilliant innovation in the realm of generative general artificial intelligence. That includes Alphabet, whose primary tool (and originally namesake) was a genuinely revolutionary search engine that became so ubiquitous that the company seemingly had no choice but to completely eviscerate everything good about it in order to keep making an even more absurd profit. But it turns out that these AI scams aren't just bad for user experiences, user trust, and democracy; they're also terrible for the environment. >From >Jacobin<https://jacobin.com/2024/06/ai-data-center-energy-usage-environment>: >Each time you search for something like "how many rocks should I eat" and >Google's AI "snapshot" tells >you<https://arstechnica.com/information-technology/2024/05/googles-ai-overview-can-give-false-misleading-and-dangerous-answers/> > "at least one small rock per day," you're consuming approximately three >watt-hours of electricity, according to Alex de Vries, the founder of >Digiconomist, a research company exploring the unintended consequences of >digital trends. That's ten times the power consumption of a traditional Google search<https://googleblog.blogspot.com/2009/01/powering-google-search.html>, and roughly equivalent to the amount of power used when talking for an hour on a home phone<https://www.daftlogic.com/information-appliance-power-consumption.htm>. (Remember those?) Collectively, De Vries calculates<https://www.cell.com/joule/abstract/S2542-4351(23)00365-3?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2542435123003653%3Fshowall%3Dtrue> that adding AI-generated answers to all Google searches could easily consume as much electricity as the country of Ireland. All just so that Google AI can go tell you about the nutritional value of eating rocks every day. The Jacobin piece goes even more in-depth into the absolutely absurd energy demands of this sort of generative AI technology. Consider: These facilities suck up substantial amounts of water<https://www.nature.com/articles/s41545-021-00101-w> to cool their servers, and are often located<https://gridstrategiesllc.com/wp-content/uploads/2023/12/National-Load-Growth-Report-2023.pdf> in places where land is cheap — like deserts. Only a few operators<https://www.bv.com/perspectives/water-efficiency-opportunities-data-centers/> report their water usage, even though a fifth<https://iopscience.iop.org/article/10.1088/1748-9326/abfba1/pdf> of servers draw water "from moderately to highly stressed watersheds." One paper estimates that globally, the demand for water for data centers could be half that of the United Kingdom<https://arxiv.org/abs/2111.00364> within the next several years. None of this is particularly surprising, given the similar environmental impacts of other recent web3 trends like cryptocurrency and NFTs. But it is frustrating. The last decade has seen consistent record-breaking rises in global temperatures. As such, most societies are trying to figure out ways to transition their energy grids to less-carbon-intensive energy sources—a move which has a legion of financial incentives fighting back against it, even though it's pretty clearly fucking necessary. And now we're using up all the energy (and other resources) that we do have in order to fucking summarize the exact same thing your brain could have figured out with 3 seconds of scanning down a web page of reliable search results. But that's because capitalism breeds innovation, baby!! The Hidden Environmental Impact of AI<https://jacobin.com/2024/06/ai-data-center-energy-usage-environment> [Lois Parshley / Jacobin] _______________________________________________ Link mailing list [email protected] https://mailman.anu.edu.au/mailman/listinfo/link
