%20--%3e%3csvg%20version='1.1'%20id='Layer_1'%20xmlns='http://www.w3.org/2000/svg'%20xmlns:xlink='http://www.w3.org/1999/xlink'%20x='0px'%20y='0px'%20viewBox='0%200%20100.6%2073.7'%20style='enable-background:new%200%200%20100.6%2073.7;'%20xml:space='preserve'%3e%3cstyle%20type='text/css'%3e%20.st0{fill:%23AF001E;}%20.st1{fill:%23FFFFFF;}%20%3c/style%3e%3cg%3e%3crect%20class='st0'%20width='100.6'%20height='73.7'/%3e%3cpath%20class='st1'%20d='M25.2,36.3l-0.5,2.3c-0.2,0.7-0.2,1-0.2,1.2c0,0.5,0.2,1,0.5,1.6L31.5,53c0.5,0.9,1,1.3,2.3,1.3h5.9%20c0.8,0,1.7-0.5,0.8-1.8l-7.9-13.1c2.7-1.7,5.6-4.1,5.6-9.3c0-6.8-3.8-10.6-12.2-10.6H13.9c-1.1,0-1.4,0.4-1.4,1.5v32%20c0,1.1,0.3,1.5,1.4,1.5h5c1.1,0,1.4-0.4,1.4-1.5V26h5.2c3.3,0,4.8,1.2,4.8,4.2c0,2.8-1.2,3.6-2.5,4.2C26.3,35,25.7,34.3,25.2,36.3z%20'/%3e%3cpath%20class='st1'%20d='M45.9,48.8L45.3,53c-0.1,0.8,0.3,1.1,1,1.2c1.6,0.2,5.5,0.5,9.4,0.5c8.6,0,14.9-2.8,14.9-10.9%20c0-4.7-2.1-8.3-9.2-10.1l-4-1c-2.7-0.7-4-1.6-4-3.6c0-2.8,2.6-3.4,5.1-3.4c3.4,0,6.8,0.3,8.1,0.4c1.4,0.1,1.8-0.1,2-1.2l0.7-4.1%20c0.1-0.8-0.3-1.1-1-1.2c-1.6-0.2-5-0.5-9.4-0.5c-9,0-13.3,3.9-13.3,10.1c0,5.6,3,8.9,8.3,10.3l4,1c3.3,0.9,4.7,1.8,4.7,3.9%20c0,3-2.1,3.6-7.1,3.6c-2.3,0-6.2-0.3-7.6-0.4C46.5,47.4,46.1,47.6,45.9,48.8z'/%3e%3cpath%20class='st1'%20d='M79.4,52.9c0,1.1,0.3,1.5,1.4,1.5h5c1.1,0,1.4-0.4,1.4-1.5v-32c0-1.1-0.3-1.5-1.4-1.5h-5%20c-1.1,0-1.4,0.4-1.4,1.5V52.9z'/%3e%3c/g%3e%3c/svg%3e)
Un breve colloquio con un'intelligenza artificiale come ChatGPT o Bard comporta il consumo di mezzo litro d'acqua dolce. Un fabbisogno idrico che potrebbe presto avere un impatto significativo sulle riserve di acqua, soprattutto nell'arido sud degli Stati Uniti, come si legge in un articolo pubblicato lunedì dal portale zurighese Infosperber.
Le piattaforme di intelligenza artificiale imparano a dare risposte precise alle domande utilizzando un gran numero di esempi. Durante l'addestramento, i computer necessitano di molta energia e di raffreddamento. Basti pensare che nell'addestramento di GPT-3 presso i centri dati statunitensi di Microsoft sono stati impiegati circa 700'000 litri di acqua dolce e pulita, ha spiegato l'ingegnere informatico Shaolei Ren in un'intervista alla rivista tecnologica The Markup, ripresa in Svizzera da Infosperber. In un centro dati asiatico il consumo sarebbe di tre volte superiore.
Secondo i calcoli di Ren, una conversazione con ChatGPT di 30-50 domande consuma cinque decilitri. Se si pensa al numero di utenti, le quantità in gioco diventano enormi. Per gli stati americani aridi con molti centri dati, come la California, gli scenari futuri si fanno quindi preoccupanti.
Info Notte del 28.04.2023: ChatGPT di nuovo disponibile in Italia
RSI Info 28.04.2023, 23:22
%20--%3e%3csvg%20version='1.1'%20id='Layer_1'%20xmlns='http://www.w3.org/2000/svg'%20xmlns:xlink='http://www.w3.org/1999/xlink'%20x='0px'%20y='0px'%20viewBox='0%200%20138.2%2048'%20style='enable-background:new%200%200%20138.2%2048;'%20xml:space='preserve'%3e%3cstyle%20type='text/css'%3e%20.st0{fill-rule:evenodd;clip-rule:evenodd;fill:%23AF001E;}%20.st1{fill-rule:evenodd;clip-rule:evenodd;fill:%23FFFFFF;}%20%3c/style%3e%3cpath%20class='st0'%20d='M0,48h138V0H0V48z'/%3e%3cpath%20class='st1'%20d='M9.3,31.7l-0.4,2.7C8.8,35,9,35.2,9.5,35.3c2,0.2,4,0.3,6.1,0.3c5.6,0,9.7-1.8,9.7-7.1c0-3.1-1.3-5.4-6-6.6%20l-2.6-0.7c-1.8-0.5-2.6-1-2.6-2.3c0-1.9,1.7-2.2,3.3-2.2c2.2,0,4.4,0.2,5.3,0.2c0.9,0.1,1.2,0,1.3-0.8l0.4-2.6c0-0.1,0-0.2,0-0.3%20c0-0.1-0.1-0.2-0.1-0.3s-0.1-0.1-0.2-0.2c-0.1,0-0.2-0.1-0.3-0.1c-2-0.2-4.1-0.3-6.1-0.3C11.8,12.4,9,14.9,9,19c0,3.6,2,5.8,5.4,6.7%20l2.6,0.7c2.1,0.6,3,1.2,3,2.5c0,2-1.3,2.3-4.6,2.3c-1.5,0-4-0.2-4.9-0.3C9.6,30.9,9.4,31,9.3,31.7'/%3e%3cpath%20class='st1'%20d='M36.4,23.6l-0.3,1.5C36,25.4,36,25.6,36,25.9c0,0.4,0.1,0.8,0.4,1.1l4.2,7.6c0.1,0.3,0.3,0.5,0.6,0.7%20c0.3,0.2,0.6,0.2,0.9,0.2h3.8c0.5,0,1.1-0.3,0.5-1.2l-5.1-8.6c1.7-1.1,3.6-2.6,3.6-6.1c0-4.4-2.5-6.9-7.9-6.9h-7.8%20c-0.7,0-0.9,0.2-0.9,1v20.8c0,0.7,0.2,1,0.9,1h3.2c0.7,0,0.9-0.2,0.9-1V16.8h3.3c2.1,0,3.1,0.8,3.1,2.8c0.1,0.6,0,1.2-0.3,1.6%20c-0.3,0.5-0.7,0.9-1.3,1.1C37.1,22.7,36.8,22.3,36.4,23.6z'/%3e%3cpath%20class='st1'%20d='M66.1,34V22.7c0-0.7-0.2-1-0.9-1h-5.8c-0.7,0-0.9,0.2-0.9,1v2.5c0,0.7,0.2,0.9,0.9,0.9h1.7v5%20c-0.6,0.1-1.1,0.1-1.7,0.1c-4.8,0-6-3.1-6-7.4c0-5.4,2-6.8,6.5-6.8c1.3,0,3.1,0.1,4.1,0.2c1,0.1,1.2,0,1.3-0.8l0.4-2.5%20c0.1-0.7,0-1-0.8-1.1c-1.8-0.2-3.6-0.3-5.4-0.2c-8.2,0-11.6,4.3-11.6,11.3s2.7,11.9,11.3,11.9c2,0.1,4,0,6-0.4%20C65.9,35.2,66.2,34.9,66.1,34'/%3e%3cpath%20class='st1'%20d='M74.9,31.7l-0.4,2.7c-0.1,0.5,0.2,0.7,0.6,0.8c2,0.2,4,0.3,6.1,0.3c5.6,0,9.7-1.8,9.7-7.1c0-3.1-1.3-5.4-6-6.6%20l-2.6-0.7c-1.8-0.5-2.6-1-2.6-2.3c0-1.9,1.7-2.2,3.3-2.2c2.2,0,4.4,0.2,5.3,0.2c0.9,0.1,1.2,0,1.3-0.8l0.4-2.6c0-0.1,0-0.2,0-0.3%20c0-0.1-0.1-0.2-0.1-0.3s-0.1-0.1-0.2-0.2c-0.1,0-0.2-0.1-0.3-0.1c-2-0.2-4.1-0.3-6.1-0.3c-5.8,0-8.6,2.6-8.6,6.6%20c0,3.6,2,5.8,5.4,6.7l2.6,0.7c2.1,0.6,3,1.2,3,2.6c0,2-1.3,2.3-4.6,2.3c-1.5,0-4-0.2-4.9-0.3C75.3,30.8,75,30.9,74.9,31.7'/%3e%3cpath%20class='st1'%20d='M93.1,31.7l-0.4,2.7c-0.1,0.5,0.2,0.7,0.6,0.8c2,0.2,4,0.3,6.1,0.3c5.6,0,9.7-1.8,9.7-7.1c0-3.1-1.3-5.4-6-6.6%20l-2.6-0.7c-1.8-0.5-2.6-1-2.6-2.3c0-1.9,1.7-2.2,3.3-2.2c2.2,0,4.4,0.2,5.3,0.2c0.9,0.1,1.2,0,1.3-0.8l0.4-2.6c0-0.1,0-0.2,0-0.3%20c0-0.1-0.1-0.2-0.1-0.3c-0.1-0.1-0.1-0.1-0.2-0.2c-0.1,0-0.2-0.1-0.3-0.1c-2-0.2-4.1-0.3-6.1-0.3c-5.8,0-8.6,2.6-8.6,6.6%20c0,3.6,2,5.8,5.4,6.7l2.6,0.7c2.1,0.6,3,1.2,3,2.6c0,2-1.3,2.3-4.6,2.3c-1.5,0-4-0.2-4.9-0.3S93.1,30.9,93.1,31.7'/%3e%3cpath%20class='st1'%20d='M120.2,23.6l-0.3,1.5c-0.1,0.3-0.1,0.5-0.1,0.8c0,0.4,0.1,0.8,0.3,1.1l4.2,7.6c0.1,0.3,0.3,0.5,0.6,0.7%20c0.3,0.2,0.6,0.2,0.9,0.2h3.8c0.5,0,1.1-0.3,0.5-1.2l-5.1-8.6c1.7-1.1,3.7-2.7,3.7-6.1c0-4.4-2.5-6.9-7.9-6.9H113%20c-0.7,0-0.9,0.2-0.9,1v20.8c0,0.7,0.2,1,0.9,1h3.2c0.7,0,0.9-0.2,0.9-1V16.8h3.3c2.1,0,3.1,0.8,3.1,2.8c0.1,0.6,0,1.2-0.3,1.6%20c-0.3,0.5-0.7,0.9-1.3,1.1C120.9,22.7,120.6,22.3,120.2,23.6'/%3e%3c/svg%3e)