{"id":9313,"date":"2024-07-22T08:35:48","date_gmt":"2024-07-22T08:35:48","guid":{"rendered":"https:\/\/www.slovensko.ai\/?p=9313"},"modified":"2024-07-22T08:35:48","modified_gmt":"2024-07-22T08:35:48","slug":"vznik-koalicie-pre-bezpecnu-ai-cosai","status":"publish","type":"post","link":"https:\/\/www.slovensko.ai\/en\/vznik-koalicie-pre-bezpecnu-ai-cosai\/","title":{"rendered":"Vznik koal\u00edcie pre bezpe\u010dn\u00fa AI – CoSAI"},"content":{"rendered":"
<\/span> 637<\/span> Viden\u00ed<\/span><\/div>

Niektor\u00e9 z najv\u00fdznamnej\u0161\u00edch mien v Big Tech sa spojili, aby spolu zalo\u017eili Coalition for Secure AI (CoSAI).<\/strong><\/p>\n

T\u00e1to nov\u00e1 iniciat\u00edva s otvoren\u00fdm zdrojov\u00fdm k\u00f3dom, organizovan\u00e1 glob\u00e1lnym \u0161tandardiza\u010dn\u00fdm org\u00e1nom OASIS, m\u00e1 za cie\u013e \u0161tandardizova\u0165 a zdie\u013ea\u0165 najlep\u0161ie postupy s\u00favisiace s bezpe\u010dn\u00fdm v\u00fdvojom a nasaden\u00edm umelej inteligencie (AI).<\/p>\n

Ciele a v\u00fdznam CoSAI<\/h2>\n

CoSAI <\/strong>sa zameriava na rie\u0161enie mnoh\u00fdch bezpe\u010dnostn\u00fdch v\u00fdziev spojen\u00fdch s v\u00fdvojom a nasaden\u00edm AI. Medzi tieto v\u00fdzvy patr\u00ed:<\/p>\n

    \n
  • Kr\u00e1de\u017e modelu<\/li>\n
  • D\u00e1ta otrava tr\u00e9ningov\u00fdch d\u00e1t<\/li>\n
  • Zav\u00e1dzanie \u0161kodliv\u00fdch vstupov prostredn\u00edctvom r\u00fdchleho vstrekovania<\/li>\n
  • \u0160k\u00e1lovan\u00e1 prevencia zneu\u017e\u00edvania<\/li>\n
  • \u00datoky na \u010dlenstvo<\/li>\n
  • Modelovanie inverzn\u00fdch \u00fatokov alebo gradientn\u00fdch inverzn\u00fdch \u00fatokov na odvodenie s\u00fakromn\u00fdch inform\u00e1ci\u00ed<\/li>\n
  • Extrahovanie d\u00f4vern\u00fdch inform\u00e1ci\u00ed z tr\u00e9ningov\u00fdch d\u00e1t<\/li>\n<\/ul>\n

    Vyhraden\u00e9 t\u00e9my<\/h2>\n

    Charta CoSAI jasne uv\u00e1dza, \u017ee projekt sa nebude zaobera\u0165 nasledovn\u00fdmi t\u00e9mami: dezinform\u00e1cie, halucin\u00e1cie, nen\u00e1vistn\u00fd alebo ur\u00e1\u017eliv\u00fd obsah, zaujatos\u0165, generovanie \u0161kodliv\u00e9ho softv\u00e9ru, generovanie phishingov\u00e9ho obsahu alebo in\u00e9 t\u00e9my v oblasti bezpe\u010dnosti obsahu.<\/p>\n

    Pr\u00edklady existuj\u00facich r\u00e1mcov<\/h2>\n

    Google u\u017e m\u00e1 svoj Google Secure AI Framework (SAIF) a OpenAI m\u00e1 svoj Aligning AI projekt. Napriek t\u00fdmto individu\u00e1lnym iniciat\u00edvam v\u0161ak neexistovalo \u017eiadne f\u00f3rum na kombinovanie osved\u010den\u00fdch postupov bezpe\u010dnosti AI, ktor\u00e9 by hr\u00e1\u010di v tomto odvetv\u00ed vyvinuli nez\u00e1visle, a\u017e do vzniku CoSAI.<\/p>\n

    Podpora pre mal\u00e9 spolo\u010dnosti<\/h2>\n

    Videli sme mal\u00e9 startupy, ako napr\u00edklad Mistral, ktor\u00e9 za\u017eili raketov\u00fd vzostup s modelmi AI, ktor\u00e9 vyrobili. Mnoh\u00e9 z t\u00fdchto men\u0161\u00edch spolo\u010dnost\u00ed v\u0161ak nemaj\u00fa zdroje na financovanie bezpe\u010dnostn\u00fdch t\u00edmov AI. CoSAI bude preto cenn\u00fdm bezplatn\u00fdm zdrojom osved\u010den\u00fdch postupov v oblasti bezpe\u010dnosti AI pre mal\u00fdch aj ve\u013ek\u00fdch hr\u00e1\u010dov v tomto odvetv\u00ed.<\/p>\n

    Podpora od priemyseln\u00fdch l\u00eddrov<\/h2>\n

    Heather Adkinsov\u00e1, viceprezidentka a d\u00f4stojn\u00ed\u010dka pre odolnos\u0165 vo\u010di kybernetickej bezpe\u010dnosti v spolo\u010dnosti Google, zd\u00f4raznila v\u00fdznam CoSAI: \u201eUmel\u00fa inteligenciu pou\u017e\u00edvame u\u017e mnoho rokov a vid\u00edme pretrv\u00e1vaj\u00faci potenci\u00e1l pre obrancov, ale tie\u017e si uvedomujeme jej pr\u00edle\u017eitosti pre protivn\u00edkov. CoSAI pom\u00f4\u017ee organiz\u00e1ci\u00e1m, ve\u013ek\u00fdm aj mal\u00fdm, bezpe\u010dne a zodpovedne integrova\u0165 AI \u2013 pom\u00f4\u017ee im vyu\u017ei\u0165 jej v\u00fdhody a z\u00e1rove\u0148 zmierni\u0165 rizik\u00e1.\u201c<\/p>\n

    Nick Hamilton, ved\u00faci oddelenia pre riadenie, rizik\u00e1 a dodr\u017eiavanie predpisov v OpenAI, dodal: \u201eV\u00fdvoj a nasadenie technol\u00f3gi\u00ed AI, ktor\u00e9 s\u00fa bezpe\u010dn\u00e9 a d\u00f4veryhodn\u00e9, je \u00fastredn\u00fdm bodom poslania OpenAI. Sme presved\u010den\u00ed, \u017ee na zaistenie bezpe\u010dn\u00e9ho a zodpovedn\u00e9ho pou\u017e\u00edvania umelej inteligencie je nevyhnutn\u00e9 vyvin\u00fa\u0165 spo\u013eahliv\u00e9 \u0161tandardy a postupy, a preto sme odhodlan\u00ed spolupracova\u0165 v r\u00e1mci cel\u00e9ho odvetvia. Prostredn\u00edctvom na\u0161ej \u00fa\u010dasti v CoSAI sa sna\u017e\u00edme prispie\u0165 na\u0161imi odborn\u00fdmi znalos\u0165ami a zdrojmi k vytvoreniu bezpe\u010dn\u00e9ho ekosyst\u00e9mu AI, ktor\u00fd bude pr\u00ednosom pre ka\u017ed\u00e9ho.\u201c<\/p>\n

    V\u00fdzvy a o\u010dak\u00e1vania<\/h2>\n

    D\u00fafajme, \u017ee odborn\u00edci ako I\u013eja Sutskever a \u010fal\u0161\u00ed, ktor\u00ed odi\u0161li z OpenAI kv\u00f4li obav\u00e1m o bezpe\u010dnos\u0165, dobrovo\u013ene prines\u00fa svoje sk\u00fasenosti a pr\u00edspevky pre CoSAI. T\u00fdm by mohli v\u00fdrazne prispie\u0165 k rozvoju bezpe\u010dnostn\u00fdch \u0161tandardov a prakt\u00edk v oblasti AI.<\/p>\n

    Z\u00e1ver<\/h2>\n

    CoSAI predstavuje d\u00f4le\u017eit\u00fd krok vpred pre bezpe\u010dnos\u0165 a zodpovedn\u00e9 nasadenie umelej inteligencie. Spojen\u00edm \u00fasilia ve\u013ek\u00fdch technologick\u00fdch spolo\u010dnost\u00ed a poskytovan\u00edm zdrojov pre men\u0161ie firmy, CoSAI pom\u00e1ha vytv\u00e1ra\u0165 bezpe\u010dnej\u0161\u00ed a efekt\u00edvnej\u0161\u00ed ekosyst\u00e9m pre v\u00fdvoj a implement\u00e1ciu AI technol\u00f3gi\u00ed.<\/p>\n

     <\/p>\n

    By: slovensko.ai<\/strong><\/p>","protected":false},"excerpt":{"rendered":"

    637 Viden\u00edNiektor\u00e9 z najv\u00fdznamnej\u0161\u00edch mien v Big Tech sa spojili, aby spolu zalo\u017eili Coalition for Secure AI (CoSAI). T\u00e1to nov\u00e1 iniciat\u00edva s otvoren\u00fdm zdrojov\u00fdm k\u00f3dom, organizovan\u00e1 glob\u00e1lnym \u0161tandardiza\u010dn\u00fdm org\u00e1nom OASIS, m\u00e1 za cie\u013e \u0161tandardizova\u0165 a zdie\u013ea\u0165 najlep\u0161ie postupy s\u00favisiace s bezpe\u010dn\u00fdm v\u00fdvojom a nasaden\u00edm umelej inteligencie (AI). Ciele a v\u00fdznam CoSAI CoSAI sa zameriava na […]<\/p>\n","protected":false},"author":5,"featured_media":9314,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[51],"tags":[254],"class_list":["post-9313","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog","tag-coalition-for-secure-ai-cosai"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts\/9313"}],"collection":[{"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/comments?post=9313"}],"version-history":[{"count":1,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts\/9313\/revisions"}],"predecessor-version":[{"id":9315,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts\/9313\/revisions\/9315"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/media\/9314"}],"wp:attachment":[{"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/media?parent=9313"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/categories?post=9313"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/tags?post=9313"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}