{"id":106591,"date":"2025-08-30T11:07:10","date_gmt":"2025-08-30T11:07:10","guid":{"rendered":"https:\/\/www.newsbeep.com\/au\/106591\/"},"modified":"2025-08-30T11:07:10","modified_gmt":"2025-08-30T11:07:10","slug":"ai-learns-to-create-images-with-minimal-energy-use","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/au\/106591\/","title":{"rendered":"AI learns to create images with minimal energy use"},"content":{"rendered":"<p>By Alimat Aliyeva<\/p>\n<p>A team of researchers led by Aydogan Ozcan at the University of<br \/>\nCalifornia, Los Angeles (UCLA), has developed a groundbreaking<br \/>\ntechnology that could transform the way generative artificial<br \/>\nintelligence (AI) systems operate \u2014 by using light instead of<br \/>\nelectricity. Their findings were recently published in the<br \/>\nprestigious journal Nature, Azernews reports.<\/p>\n<p>Modern generative models \u2014 especially diffusion-based ones like<br \/>\nDALL\u2022E 2 or Stable Diffusion \u2014 require enormous amounts of energy.<br \/>\nTraining these models on large datasets can consume tens of<br \/>\nthousands of kilowatt-hours. To put that into perspective, the<br \/>\naverage household uses around 3,000 to 10,000 kilowatt-hours of<br \/>\nelectricity per year. So training just one AI model can require as<br \/>\nmuch energy as several homes consume annually.<\/p>\n<p>Even generating a single image isn\u2019t energy-free. On traditional<br \/>\nGPUs, a single full diffusion process can consume tens of<br \/>\nwatt-hours. Multiply that by 1,000 steps per image, and you&#8217;re<br \/>\nlooking at hundreds of watt-hours \u2014 roughly equivalent to running<br \/>\nan electric kettle for one or two hours.<\/p>\n<p>The new optical approach, however, is radically different.<\/p>\n<p>Instead of relying on energy-hungry processors, the system uses<br \/>\na beam of light. A digital encoder first creates a &#8220;noise circuit&#8221;<br \/>\n\u2014 a pattern that requires almost no power. This pattern is then<br \/>\nprojected onto a laser beam using a spatial light modulator (SLM).<br \/>\nThe light, carrying this noise pattern, passes through a second<br \/>\nSLM, which transforms it into a fully rendered image. In essence,<br \/>\nthe laser is doing all the heavy lifting, not the computer.<\/p>\n<p>\u201cOur optical generative models can synthesize countless images<br \/>\nwith virtually no computing power, offering a scalable and<br \/>\nenergy-efficient alternative to digital artificial intelligence<br \/>\nmodels,\u201d said Shiqi Chen, the study\u2019s lead author.<\/p>\n<p>The implications of this technology are vast. Thanks to its<br \/>\nultra-low energy consumption and high processing speed, it could be<br \/>\nintegrated into lightweight devices \u2014 from virtual and augmented<br \/>\nreality systems to smartphones, wearable gadgets, and even<br \/>\nAI-powered glasses.<\/p>\n<p>Researchers believe that by harnessing light, AI could finally<br \/>\nbecome both environmentally sustainable and accessible on a much<br \/>\nbroader scale. The combination of speed, energy efficiency, and<br \/>\nimage quality may signal the beginning of a new era \u2014 one where<br \/>\npowerful generative AI doesn&#8217;t have to come with a massive carbon<br \/>\nfootprint.<\/p>\n","protected":false},"excerpt":{"rendered":"By Alimat Aliyeva A team of researchers led by Aydogan Ozcan at the University of California, Los Angeles&hellip;\n","protected":false},"author":2,"featured_media":106592,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[9872,64,63,961,9874,9870,1348,174,1048,164,9873,9871,84,9869,105,940,1346],"class_list":["post-106591","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology","tag-analytics","tag-au","tag-australia","tag-azerbaijan","tag-azernews","tag-baku","tag-breaking-news","tag-business-news","tag-company-news","tag-economy","tag-financial-news","tag-oil-and-gas-news","tag-politics","tag-socar","tag-technology","tag-turkey","tag-world-news"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/106591","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/comments?post=106591"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/posts\/106591\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media\/106592"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/media?parent=106591"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/categories?post=106591"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/au\/wp-json\/wp\/v2\/tags?post=106591"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}