Hijerarhijski model zaklju\u010divanja (HRM) kojeg koristi vje\u0161ta\u010dka inteligencija ra\u0111ena po uzoru na ljudski mozak, u zadacima logi\u010dkog zaklju\u010divanja nadma\u0161uje najmodernije velike jezi\u010dne modele poput ChatGPT-a.<\/p>\n
HRM je rezultat rada singapurske kompanije Sapient Intelligence i dizajniran je prema na\u010dinu na koji mozak obra\u0111uje informacije na vi\u0161e hijerarhijskih nivoa i vremenskih skala. Poput odnosa prefrontalnog korteksa i bazalnih ganglija u ljudskom mozgu, i HRM djeluje kroz dva odvojena, me\u0111usobno povezana modula: jedan za sporo, visokorangirano planiranje, i drugi za brze, detaljne prora\u010dune.<\/p>\n
Umjesto klasi\u010dne lan\u010dane dekompozicije problema (chain-of-thought), HRM koristi \u201citerativno rafiniranje\u201d: u vi\u0161e kratkih ciklusa “razmi\u0161ljanja” model predla\u017ee rje\u0161enje, ocjenjuje treba li ga pobolj\u0161ati i po potrebi ponavlja rafiniranje. Klju\u010dno je \u0161to zavr\u0161ava s vrlo ta\u010dnim odgovorima bez eksplicitnog nadzora svakog me\u0111ukoraka, objavio je bug.hr.\u00a0<\/a><\/p>\n","protected":false},"excerpt":{"rendered":" Hijerarhijski model zaklju\u010divanja (HRM) kojeg koristi vje\u0161ta\u010dka inteligencija ra\u0111ena po uzoru na ljudski mozak, u zadacima logi\u010dkog zaklju\u010divanja nadma\u0161uje najmodernije velike jezi\u010dne modele poput ChatGPT-a<\/p>\n","protected":false},"author":1,"featured_media":416676,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_seopress_robots_primary_cat":"none","_seopress_titles_title":"","_seopress_titles_desc":"","_seopress_robots_index":"","_et_pb_use_builder":"","_et_pb_old_content":"","_et_gb_content_width":"","inline_featured_image":false,"footnotes":""},"categories":[14],"tags":[],"class_list":["post-438784","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-vijesti"],"_links":{"self":[{"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/posts\/438784","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/comments?post=438784"}],"version-history":[{"count":1,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/posts\/438784\/revisions"}],"predecessor-version":[{"id":438787,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/posts\/438784\/revisions\/438787"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/media\/416676"}],"wp:attachment":[{"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/media?parent=438784"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/categories?post=438784"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.pcnen.com\/portal\/wp-json\/wp\/v2\/tags?post=438784"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}