{"id":15186,"date":"2025-04-02T09:51:51","date_gmt":"2025-04-02T07:51:51","guid":{"rendered":"https:\/\/www.slovensko.ai\/?p=15186"},"modified":"2025-04-02T09:51:51","modified_gmt":"2025-04-02T07:51:51","slug":"mysli-ai-ako-clovek-nova-studia-ukazuje-ze-niekedy-az-prilis","status":"publish","type":"post","link":"https:\/\/www.slovensko.ai\/en\/mysli-ai-ako-clovek-nova-studia-ukazuje-ze-niekedy-az-prilis\/","title":{"rendered":"Mysl\u00ed AI ako \u010dlovek? Nov\u00e1 \u0161t\u00fadia ukazuje, \u017ee niekedy a\u017e pr\u00edli\u0161"},"content":{"rendered":"
<\/span> 900<\/span> Viden\u00ed<\/span><\/div>

Umel\u00e1 inteligencia sa \u010dasto pova\u017euje za chladn\u00fd a logick\u00fd n\u00e1stroj, ktor\u00fd dok\u00e1\u017ee rozhodova\u0165 bez em\u00f3ci\u00ed a iracion\u00e1lnych sklonov. Ale \u010do ak najnov\u0161ie AI syst\u00e9my, ako napr\u00edklad ChatGPT, kop\u00edruj\u00fa aj na\u0161e ment\u00e1lne skratky a chyby v uva\u017eovan\u00ed? Ned\u00e1vna \u0161t\u00fadia z Manufacturing & Service Operations Management<\/em> nazna\u010duje, \u017ee pr\u00e1ve to sa deje.<\/p>\n

Vedci zistili, \u017ee pri ur\u010dit\u00fdch rozhodnutiach jazykov\u00e9 modely ako ChatGPT vykazuj\u00fa rovnak\u00e9 kognit\u00edvne skreslenia<\/strong> ako \u013eudia. Tieto skreslenia, napr\u00edklad hr\u00e1\u010dsky omyl<\/strong>, averzia k riziku<\/strong> alebo efekt r\u00e1mcovania<\/strong>, sa be\u017ene objavuj\u00fa pri \u013eudskom rozhodovan\u00ed \u2013 \u010dasto vplyvom em\u00f3ci\u00ed alebo skryt\u00fdch predpokladov. U AI v\u0161ak ich pr\u00edtomnos\u0165 vyvol\u00e1va ot\u00e1zku: S\u00fa tieto syst\u00e9my naozaj \u201eneutr\u00e1lne\u201c, alebo len kop\u00edruj\u00fa chyby na\u0161ich my\u0161lienkov\u00fdch vzorcov?<\/strong><\/p>\n

\n

Ke\u010f AI nie je tak\u00e1 objekt\u00edvna, ako sa zd\u00e1<\/h3>\n

Vedci analyzovali odpovede AI na s\u00e9riu scen\u00e1rov ur\u010den\u00fdch na testovanie racion\u00e1lneho myslenia. Zistili, \u017ee ChatGPT:<\/p>\n

    \n
  • v 49 % pr\u00edpadov<\/strong> pod\u013eahol rovnak\u00fdm skresleniam ako \u013eudia,<\/li>\n
  • mal probl\u00e9my s rozhodnutiami, kde neexistovala jednozna\u010dn\u00e1 spr\u00e1vna odpove\u010f,<\/li>\n
  • a v niektor\u00fdch pr\u00edpadoch vykazoval nadmern\u00fa sebad\u00f4veru<\/strong> vo vlastn\u00e9 nespr\u00e1vne odpovede.<\/li>\n<\/ul>\n

    Zauj\u00edmav\u00e9 je, \u017ee ke\u010f v\u00fdskumn\u00edci explicitne po\u017eiadali model o \u201eracion\u00e1lne\u201c rozhodnutie, v\u00fdsledky sa zlep\u0161ili \u2013 \u010do nazna\u010duje, \u017ee AI dok\u00e1\u017ee tieto skreslenia do ur\u010ditej miery prekona\u0165, ak je spr\u00e1vne usmer\u0148ovan\u00e1<\/strong>.<\/p>\n

    \n

    Pre\u010do je to probl\u00e9m?<\/h3>\n

    V praxi sa tieto poznatky m\u00f4\u017eu prejavi\u0165 v r\u00f4znych oblastiach: od prij\u00edmania zamestnancov, cez schva\u013eovanie \u00faverov, a\u017e po anal\u00fdzu pr\u00e1vnych alebo zdravotn\u00fdch d\u00e1t. Ak AI syst\u00e9m automatizuje rozhodnutia, ktor\u00e9 s\u00fa zalo\u017een\u00e9 na skreslen\u00fdch \u00fadajoch alebo nepresn\u00fdch vzorcoch<\/strong>, m\u00f4\u017ee:<\/p>\n

      \n
    • replikova\u0165 diskrimin\u00e1ciu<\/strong>,<\/li>\n
    • uprednost\u0148ova\u0165 nespr\u00e1vne rie\u0161enia<\/strong>,<\/li>\n
    • alebo dokonca zd\u00f4vodni\u0165 nespr\u00e1vne rozhodnutie \u201elogicky\u201c<\/strong>, \u010do e\u0161te s\u0165a\u017euje jeho identifik\u00e1ciu.<\/li>\n<\/ul>\n

      AI tak m\u00f4\u017ee fungova\u0165 ako zrkadlo spolo\u010dnosti<\/strong>, ale nie v\u017edy zrkadl\u00ed to najlep\u0161ie.<\/p>\n

      \n

      M\u00f4\u017eeme tomu zabr\u00e1ni\u0165?<\/h3>\n

      \u00c1no \u2013 ale iba vtedy, ak si budeme vedom\u00ed riz\u00edk. Odborn\u00edci odpor\u00fa\u010daj\u00fa:<\/p>\n

        \n
      • zav\u00e1dza\u0165 auditovanie AI rozhodnut\u00ed<\/strong>, podobne ako kontrolujeme rozhodnutia \u013eud\u00ed,<\/li>\n
      • zabezpe\u010di\u0165 transparentnos\u0165 algoritmov<\/strong>, aby bolo jasn\u00e9, na z\u00e1klade \u010doho AI rozhodla,<\/li>\n
      • kombinova\u0165 AI s \u013eudsk\u00fdm doh\u013eadom<\/strong>, najm\u00e4 v oblastiach s ve\u013ek\u00fdm etick\u00fdm alebo spolo\u010densk\u00fdm dopadom.<\/li>\n<\/ul>\n

        Z\u00e1rove\u0148 je d\u00f4le\u017eit\u00e9, aby v\u00fdvoj\u00e1ri tr\u00e9novali AI na diverzifikovan\u00fdch d\u00e1tach<\/strong>, ktor\u00e9 zah\u0155\u0148aj\u00fa r\u00f4zne uhly poh\u013eadu a neobsahuj\u00fa nevedom\u00e9 predsudky.<\/p>\n

        \n

        AI ako partner, nie sudca<\/h3>\n

        Je l\u00e1kav\u00e9 necha\u0165 rozhodovanie na stroje, ktor\u00e9 nepoznaj\u00fa \u00fanavu, stres \u010di osobn\u00e9 preferencie. No ak tieto stroje preberaj\u00fa aj na\u0161e my\u0161lienkov\u00e9 skratky, mus\u00edme sa p\u00fdta\u0165: Chceme, aby n\u00e1s hodnotila inteligencia, ktor\u00e1 mysl\u00ed ako \u010dlovek \u2014 ale bez schopnosti pochopi\u0165 d\u00f4sledky?<\/strong><\/p>\n

        Bud\u00facnos\u0165 AI nie je len o v\u00fdpo\u010dtoch a algoritmoch. Je o zodpovednosti<\/strong> a kritickom myslen\u00ed<\/strong>. Preto\u017ee k\u00fdm sa nau\u010d\u00edme spr\u00e1vne vyu\u017e\u00edva\u0165 umel\u00fa inteligenciu, mus\u00edme si najprv prizna\u0165, ako ve\u013emi je dnes podobn\u00e1 n\u00e1m \u2013 so v\u0161etk\u00fdmi na\u0161imi chybami.<\/p>","protected":false},"excerpt":{"rendered":"

        900 Viden\u00edUmel\u00e1 inteligencia sa \u010dasto pova\u017euje za chladn\u00fd a logick\u00fd n\u00e1stroj, ktor\u00fd dok\u00e1\u017ee rozhodova\u0165 bez em\u00f3ci\u00ed a iracion\u00e1lnych sklonov. Ale \u010do ak najnov\u0161ie AI syst\u00e9my, ako napr\u00edklad ChatGPT, kop\u00edruj\u00fa aj na\u0161e ment\u00e1lne skratky a chyby v uva\u017eovan\u00ed? Ned\u00e1vna \u0161t\u00fadia z Manufacturing & Service Operations Management nazna\u010duje, \u017ee pr\u00e1ve to sa deje. Vedci zistili, \u017ee pri […]<\/p>\n","protected":false},"author":5,"featured_media":15235,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[51],"tags":[77,63,377,54,79,53],"class_list":["post-15186","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog","tag-ai","tag-ai-inovacie","tag-ai-novinky","tag-ai-technologie","tag-slovensko-ai","tag-umela-inteligencia"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts\/15186"}],"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=15186"}],"version-history":[{"count":1,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts\/15186\/revisions"}],"predecessor-version":[{"id":15236,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/posts\/15186\/revisions\/15236"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/media\/15235"}],"wp:attachment":[{"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/media?parent=15186"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/categories?post=15186"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.slovensko.ai\/en\/wp-json\/wp\/v2\/tags?post=15186"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}