{"@type":"StructuredNewsArticle","access":{"license":"neupai_standard","structured_data":"free","full_text_access":null,"full_text_available":false,"attribution_required":true},"content":{"claims":[{"id":"c1","type":"fact","as_of":"2026-03-24","figures":null,"as_of_raw":"March 24","statement":"Google Research unveiled a compression algorithm called 'TurboQuant' that dramatically reduces AI model memory usage on March 24","comparison":null,"expiry_hint":null,"source_type":"company_disclosure","as_of_explicit":true},{"id":"c2","type":"future_plan","as_of":"2026","figures":null,"as_of_raw":"this year","statement":"TurboQuant is scheduled to be presented at the AI academic conference ICLR 2026 held this year","comparison":null,"expiry_hint":"2026-12","source_type":"company_plan","as_of_explicit":false},{"id":"c3","type":"fact","as_of":"2026-03","figures":{"unit":"비트","value":3,"converted":null,"approximate":false},"as_of_raw":"March 2026","statement":"TurboQuant compressed key-value cache down to 3 bits while maintaining no accuracy loss whatsoever","comparison":null,"expiry_hint":null,"source_type":"research_paper","as_of_explicit":false},{"id":"c4","type":"fact","as_of":"2026-03","figures":{"unit":"배","value":8,"converted":null,"approximate":false},"as_of_raw":"March 2026","statement":"When 4-bit TurboQuant was applied on NVIDIA H100 GPU, it recorded up to 8 times faster processing speed compared to 32-bit uncompressed method","comparison":"32bit_uncompressed","expiry_hint":null,"source_type":"research_paper","as_of_explicit":false},{"id":"c5","type":"fact","as_of":"2026-03","figures":{"unit":"배","value":6,"converted":null,"approximate":true},"as_of_raw":"March 2026","statement":"Memory usage was reduced by more than 6 times compared to existing methods","comparison":"previous_method","expiry_hint":null,"source_type":"research_paper","as_of_explicit":false}],"topics":["artificial intelligence","algorithm","technology","google"],"summary":"Google Research has unveiled a compression algorithm called 'TurboQuant' that reduces AI model memory usage by more than 6 times and improves processing speed by 8 times. This technology is expected to significantly improve the efficiency of large language models and vector search engines.","entities":[{"name":"Google Research","type":"organization","metadata":{"parent":"corp:us:alphabet"},"canonical_id":"org:us:google-research","role_in_article":"primary_subject"},{"name":"TurboQuant","type":"product","metadata":{"parent":null},"canonical_id":"product:us:turboquant","role_in_article":"primary_subject"},{"name":"ICLR 2026","type":"organization","metadata":{"parent":null},"canonical_id":"org:xx:iclr-2026","role_in_article":"mentioned"},{"name":"NVIDIA","type":"company","metadata":{"parent":null,"ticker":"NVDA"},"canonical_id":"corp:us:nvidia","role_in_article":"mentioned"},{"name":"Gemini","type":"product","metadata":{"parent":"corp:us:alphabet"},"canonical_id":"product:us:gemini","role_in_article":"mentioned"}],"headline":"Google unveils AI compression algorithm 'TurboQuant'... reduces memory by 6x and improves speed by 8x","geography":["US"]},"@context":"https://neupai.io/schema/v0.2","identity":{"ai_url":null,"author":"버트","language":"en","publisher":{"name":"테크42","type":"online","domain":"tech42.co.kr"},"article_id":"tech42_20260325_google-turboquant-ai-compression","updated_at":null,"originality":"self_produced","article_type":"straight_news","published_at":"2026-03-25T22:45:34.000Z","canonical_url":"https://www.tech42.co.kr/%ea%b5%ac%ea%b8%80-ai-%ec%95%95%ec%b6%95-%ec%95%8c%ea%b3%a0%eb%a6%ac%ec%a6%98-%ed%84%b0%eb%b3%b4%ed%80%80%ed%8a%b8-%ea%b3%b5%ea%b0%9c%eb%a9%94%eb%aa%a8%eb%a6%ac-6%eb%b0%b0-%ec%a4%84/"},"temporal":{"freshness":"archival","next_update_expected":null},"provenance":{"source_chain":["primary_reporting"],"related_articles":[],"original_source_url":null}}