{"id":5679,"date":"2024-03-27T10:37:49","date_gmt":"2024-03-27T03:37:49","guid":{"rendered":"https:\/\/wam.vn\/en\/?p=5679"},"modified":"2024-03-27T10:37:49","modified_gmt":"2024-03-27T03:37:49","slug":"how-much-energy-ai-really-needs-and-why-thats-not-its-main-problem","status":"publish","type":"post","link":"https:\/\/wam.vn\/en\/how-much-energy-ai-really-needs-and-why-thats-not-its-main-problem\/","title":{"rendered":"How much energy AI really needs. And why that&#8217;s not its main problem"},"content":{"rendered":"<p><em><strong>TLDR<\/strong>: <\/em>AI&#8217;s energy demand is massive and costly, but the real issue is the future monopolization by wealthy entities, exacerbating socioeconomic disparities. Popular keywords: AI, energy consumption, socioeconomic disparities, Elon Musk, OpenAI lawsuit.<\/p>\n<p>This article is a summary of a You Tube video &#8220;How much energy AI really needs. And why that&#8217;s not its main problem&#8221; by Sabine Hossenfelder<br \/>\n<iframe title=\"YouTube video player\" src=\"https:\/\/www.youtube.com\/embed\/0ZraZPFVr-U?si=U2lQgI7GjfygEsPD\" width=\"560\" height=\"315\" frameborder=\"0\" allowfullscreen=\"allowfullscreen\"><\/iframe><\/p>\n<h3><\/h3>\n<h3>10 Key Takeaways:<\/h3>\n<ol>\n<li><strong>High Energy Consumption<\/strong>: AI, particularly during its training phase, consumes substantial amounts of energy. For example, training GPT-3 required at least 1300 megawatt hours, enough to power around 130 US homes for a year.<\/li>\n<li><strong>Costly Training<\/strong>: The financial costs of training large AI models like GPT-4 are immense, with estimates suggesting it might have cost around 100 million dollars or more.<\/li>\n<li><strong>Elon Musk&#8217;s Lawsuit Against OpenAI<\/strong>: The suit highlights the financial stakes involved in AI development, stemming from disagreements over OpenAI&#8217;s transition from a non-profit to a for-profit entity.<\/li>\n<li><strong>Operational Energy Use<\/strong>: Operational use of AI, such as processing queries and generating images, also requires significant energy, with image generation being particularly intensive.<\/li>\n<li><strong>Environmental Impact<\/strong>: The energy use of AI operations contributes to carbon dioxide emissions, with a single image generation task consuming as much energy as charging a smartphone.<\/li>\n<li><strong>Data Centers&#8217; Energy Consumption<\/strong>: AI and cryptocurrency mining are increasing the energy demand of data centers, which already account for 1-2% of global electricity use, with expectations to double by 2026.<\/li>\n<li><strong>Efforts to Improve Efficiency<\/strong>: There are ongoing efforts to make AI more energy-efficient through dedicated hardware and innovative use of AI itself, such as DeepMind&#8217;s project to cool Google\u2019s data centers more efficiently.<\/li>\n<li><strong>Cost and Accessibility Issues<\/strong>: The high cost of developing and maintaining large AI systems suggests a future where only a few global entities own major AIs, leading to subscription-based access for most users.<\/li>\n<li><strong>Socioeconomic Disparities<\/strong>: The expense of using high-powered AIs for tasks like finding a cure for cancer or creating influential content could exacerbate wealth disparities, privileging those who can afford the computational time.<\/li>\n<li><strong>Educational Resources<\/strong>: The video promotes educational resources like Brilliant.org for those interested in learning more about neural networks and other scientific topics, highlighting the importance of accessible education in understanding AI.<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>AI&#8217;s energy demand is massive and costly, but the real issue is the future monopolization by wealthy entities, exacerbating socioeconomic disparities. Popular keywords: AI, energy consumption, socioeconomic disparities, Elon Musk, OpenAI lawsuit.<\/p>\n","protected":false},"author":3,"featured_media":5680,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[40],"tags":[],"class_list":["post-5679","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-40","description-off"],"_links":{"self":[{"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/posts\/5679","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/comments?post=5679"}],"version-history":[{"count":1,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/posts\/5679\/revisions"}],"predecessor-version":[{"id":5681,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/posts\/5679\/revisions\/5681"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/media\/5680"}],"wp:attachment":[{"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/media?parent=5679"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/categories?post=5679"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/wam.vn\/en\/wp-json\/wp\/v2\/tags?post=5679"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}