{"id":53029,"date":"2026-04-25T01:15:10","date_gmt":"2026-04-25T01:15:10","guid":{"rendered":"https:\/\/www.europesays.com\/sk\/53029\/"},"modified":"2026-04-25T01:15:10","modified_gmt":"2026-04-25T01:15:10","slug":"pytas-sa-umelej-inteligencie-na-zdravie-ak-ano-robis-chybu-ai-casto-odpoveda-zle-no-tvari-sa-velmi-presvedcivo-vosveteit-sk","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/sk\/53029\/","title":{"rendered":"P\u00fdta\u0161 sa umelej inteligencie na zdravie? Ak \u00e1no, rob\u00ed\u0161 chybu. AI \u010dasto odpoved\u00e1 zle, no tv\u00e1ri sa ve\u013emi presved\u010divo | Vosveteit.sk"},"content":{"rendered":"

Umel\u00e1 inteligencia dnes odpovie na zdravotn\u00e9 ot\u00e1zky za p\u00e1r sek\u00fand. Probl\u00e9m je, \u017ee odpove\u010f s\u00edce p\u00f4sob\u00ed presved\u010divo, no \u010dasto obsahuje chyby. Nov\u00e1 \u0161t\u00fadia publikovan\u00e1 v BMJ Open uk\u00e1zala, \u017ee realita je menej spo\u013eahliv\u00e1, ne\u017e sa na prv\u00fd poh\u013ead zd\u00e1.<\/p>\n

V\u00fdskumn\u00edci otestovali p\u00e4\u0165 popul\u00e1rnych chatbotov, ChatGPT, Gemini, Grok, Meta AI a DeepSeek. Ka\u017ed\u00fd z nich dostal 50 ot\u00e1zok z oblasti zdravia, od rakoviny cez vakc\u00edny a\u017e po v\u00fd\u017eivu \u010di \u0161portov\u00fd v\u00fdkon. Odpovede n\u00e1sledne hodnotili dvaja odborn\u00edci.<\/p>\n

<\/p>\n

<\/p>\n

Odoberaj Vosveteit.sk cez Telegram a prihl\u00e1s sa k odberu spr\u00e1v
\n <\/a><\/p>\n

V\u00fdsledky nep\u00f4sobili pr\u00e1ve upokojuj\u00faco. Takmer 20 % odpoved\u00ed ozna\u010dili ako v\u00e1\u017ene problematick\u00e9. Pribli\u017ene polovica spadala do kateg\u00f3rie \u201eprobl\u00e9mov\u00e9\u201c a zvy\u0161ok s\u00edce nevyzeral kriticky, no st\u00e1le obsahoval nedostatky. In\u00fdmi slovami, \u00faplne spo\u013eahliv\u00fa odpove\u010f si dostal len zriedka,\u00a0informuje port\u00e1l The Conversation.<\/a><\/p>\n

Umel\u00e1 inteligencia ti sebavedomo odpovie, aj ke\u010f polo\u017e\u00ed\u0161 nespr\u00e1vnu ot\u00e1zku<\/p>\n

Zauj\u00edmav\u00e9 je, \u017ee chatboty odpovedali takmer v\u017edy. Z 250 ot\u00e1zok odmietli odpove\u010f iba dvakr\u00e1t. To je jeden z najd\u00f4le\u017eitej\u0161\u00edch detailov. Skuto\u010dn\u00fd lek\u00e1r ti v mnoh\u00fdch pr\u00edpadoch povie, \u017ee nem\u00e1 dos\u0165 \u00fadajov alebo \u017ee odpove\u010f nie je jednozna\u010dn\u00e1. AI sa v\u0161ak sna\u017e\u00ed odpoveda\u0165 takmer v\u017edy.<\/p>\n

\u201eModel m\u00e1 tendenciu vyhovie\u0165 aj v situ\u00e1ci\u00e1ch, kde by odborn\u00edk priznal neistotu,\u201c nazna\u010duj\u00fa autori v\u00fdskumu.<\/p>\n

\"Zla\"ZlaZdroj: OpenClipart-Vectors z Pixabay<\/a><\/p>\n

Pr\u00e1ve tento \u201enedostatok pokory\u201c vytv\u00e1ra probl\u00e9m najm\u00e4 pri t\u00e9mach, kde veda nem\u00e1 jasn\u00fa odpove\u010f. V\u00fd\u017eiva, doplnky stravy<\/a> alebo \u0161portov\u00fd v\u00fdkon s\u00fa pln\u00e9 protichodn\u00fdch n\u00e1zorov. Chatbot si z nich vyberie jednu verziu a pod\u00e1 ju ako presved\u010div\u00fd v\u00fdsledok, aj ke\u010f realita je ove\u013ea zlo\u017eitej\u0161ia.<\/p>\n

Neprehliadni
\n <\/p>\n

\"Tvoje\"Tvoje <\/p>\n

\n\t\t\t\t\tPopul\u00e1rna spravodajsk\u00e1 aplik\u00e1cia pre iOS a Android dostala ve\u013ek\u00fa aktualiz\u00e1ciu. Pribudlo mno\u017estvo nov\u00fdch funkci\u00ed, ktor\u00e9 mus\u00ed\u0161 vysk\u00fa\u0161a\u0165 <\/p>\n

<\/a><\/p>\n

Do toho vstupuje e\u0161te jeden faktor, \u0161t\u00fdl komunik\u00e1cie. AI modely navrhli tak, aby p\u00f4sobili pr\u00edvetivo, pokojne a sebavedomo. To znie ako v\u00fdhoda, no v medic\u00edne to m\u00f4\u017ee zav\u00e1dza\u0165. Ke\u010f dostane\u0161 odpove\u010f, ktor\u00e1 p\u00f4sob\u00ed milo a z\u00e1rove\u0148 ve\u013emi istotne, prirodzene jej viac ver\u00ed\u0161, aj vtedy, ke\u010f obsahuje chyby.<\/p>\n

\u201ePresved\u010div\u00fd t\u00f3n pom\u00e1ha zakry\u0165 faktick\u00e9 nedostatky,\u201c upozor\u0148uj\u00fa v\u00fdskumn\u00edci.<\/p>\n

Rozdiely medzi jednotliv\u00fdmi syst\u00e9mami neboli dramatick\u00e9, no existovali. Najhor\u0161ie dopadol Grok<\/a>, kde viac ne\u017e polovica odpoved\u00ed obsahovala probl\u00e9my. ChatGPT a Meta AI skon\u010dili len o nie\u010do lep\u0161ie. Celkovo sa v\u0161ak uk\u00e1zalo, \u017ee nejde o probl\u00e9m jednej konkr\u00e9tnej platformy, ale o \u0161ir\u0161\u00ed jav.<\/p>\n

Ve\u013ek\u00fa rolu zohr\u00e1va typ ot\u00e1zky. Chatboty zvl\u00e1dli lep\u0161ie t\u00e9my ako vakc\u00edny alebo rakovina<\/a>, kde existuje ve\u013ea kvalitn\u00fdch d\u00e1t. Naopak, v\u00fd\u017eiva a \u0161portov\u00fd v\u00fdkon ich dostali do probl\u00e9mov. Internet je v t\u00fdchto oblastiach pln\u00fd protichodn\u00fdch r\u00e1d, tak\u017ee syst\u00e9m \u010dasto vyberie nie\u010do, \u010do s\u00edce znie dobre, no nemus\u00ed by\u0165 spr\u00e1vne.<\/p>\n

E\u0161te v\u00e4\u010d\u0161\u00ed probl\u00e9m nastal pri otvoren\u00fdch ot\u00e1zkach. Ke\u010f si polo\u017eil jednoduch\u00fa ot\u00e1zku typu \u201eje toto pravda alebo nie\u201c, chybovos\u0165 bola relat\u00edvne ni\u017e\u0161ia. No pri ot\u00e1zkach ako \u201eak\u00e9 doplnky s\u00fa najlep\u0161ie pre zdravie\u201c sa situ\u00e1cia zhor\u0161ila. Autori v\u00fdskumu podot\u00fdkaj\u00fa, \u017ee pr\u00e1ve takto sa \u013eudia p\u00fdtaj\u00fa naj\u010dastej\u0161ie.<\/p>\n

Ve\u013ek\u00fdm varovan\u00edm s\u00fa aj odkazy na zdroje. Chatboty \u010dasto pridaj\u00fa cit\u00e1cie, ktor\u00e9 p\u00f4sobia d\u00f4veryhodne. Probl\u00e9m je, \u017ee tieto odkazy b\u00fdvaj\u00fa ne\u00fapln\u00e9, chybn\u00e9 alebo \u00faplne vymyslen\u00e9. Pri testovan\u00ed dosiahla presnos\u0165 referenci\u00ed len okolo 40 %. \u017diadny syst\u00e9m nedok\u00e1zal vytvori\u0165 \u00faplne spr\u00e1vny zoznam zdrojov.<\/p>\n

\u201eCit\u00e1cie vyzeraj\u00fa ako d\u00f4kaz, no laik nem\u00e1 d\u00f4vod pochybova\u0165, aj ke\u010f s\u00fa nespr\u00e1vne,\u201c upozor\u0148uj\u00fa autori \u0161t\u00fadie: \u201ePr\u00e1ve to rob\u00ed tieto odpovede nebezpe\u010dn\u00fdmi, p\u00f4sobia d\u00f4veryhodne aj vtedy, ke\u010f nie s\u00fa.\u201c<\/p>\n

Umel\u00e1 inteligencia nedok\u00e1\u017ee hodnoti\u0165 d\u00f4kazy ako kvalifikovan\u00fd lek\u00e1r<\/p>\n

D\u00f4vod, pre\u010do sa to deje, je pomerne jednoduch\u00fd. Jazykov\u00e9 modely nefunguj\u00fa ako lek\u00e1ri. Nehodnotia d\u00f4kazy ani nerobia odborn\u00e9 rozhodnutia. Predpovedaj\u00fa najpravdepodobnej\u0161ie slov\u00e1 na z\u00e1klade d\u00e1t, na ktor\u00fdch sa u\u010dili. A tieto d\u00e1ta zah\u0155\u0148aj\u00fa nielen vedeck\u00e9 \u0161t\u00fadie, ale aj diskusie z f\u00f3r \u010di blogov.<\/p>\n

\"AI\"AIZdroj: Morrowind \/ Shutterstock.com<\/p>\n

\u010eal\u0161ie v\u00fdskumy uk\u00e1zali podobn\u00fd obraz. \u0160t\u00fadia v Nature Medicine zistila zauj\u00edmav\u00fd paradox. Chatbot dok\u00e1zal spr\u00e1vne odpoveda\u0165 takmer v 95 % pr\u00edpadov, ke\u010f v\u0161ak rovnak\u00fd n\u00e1stroj pou\u017eili be\u017en\u00ed \u013eudia, spr\u00e1vnu odpove\u010f z\u00edskali menej ne\u017e v 35 % pr\u00edpadov. Probl\u00e9m teda nespo\u010d\u00edva iba v samotnej odpovedi, ale aj v tom, ako ju pochop\u00ed\u0161 a pou\u017eije\u0161.<\/p>\n

In\u00fd v\u00fdskum publikovan\u00fd v JAMA Network Open uk\u00e1zal, \u017ee AI m\u00e1 probl\u00e9m s diagnostikou, ke\u010f dostane len z\u00e1kladn\u00e9 inform\u00e1cie. Presnos\u0165 v\u00fdrazne st\u00fapla a\u017e po doplnen\u00ed detailn\u00fdch d\u00e1t, ako s\u00fa v\u00fdsledky testov.<\/p>\n

Z toho vypl\u00fdva jednoduch\u00fd z\u00e1ver. AI vie pom\u00f4c\u0165, no nemala by by\u0165 jedin\u00fdm zdrojom rozhodovania. M\u00f4\u017ee ti vysvetli\u0165 probl\u00e9m, pom\u00f4c\u0165 pripravi\u0165 ot\u00e1zky pre lek\u00e1ra alebo zhrn\u00fa\u0165 inform\u00e1cie. Nemala by v\u0161ak nahradi\u0165 odborn\u00edka.<\/p>\n

Tieto n\u00e1stroje tu zostan\u00fa a bud\u00fa sa zlep\u0161ova\u0165. Zatia\u013e v\u0161ak plat\u00ed, \u017ee r\u00fdchla odpove\u010f e\u0161te neznamen\u00e1 spr\u00e1vnu odpove\u010f. Pri zdrav\u00ed<\/a> sa oplat\u00ed spomali\u0165 a overi\u0165 si, \u010do \u010d\u00edta\u0161.<\/p>\n


\n
\n
\n
\n
\n
\n P\u00e1\u010dil sa v\u00e1m \u010dl\u00e1nok? Sledujte n\u00e1s na Facebooku
\n
\n <\/a> <\/p>\n

<\/p>\n","protected":false},"excerpt":{"rendered":"Umel\u00e1 inteligencia dnes odpovie na zdravotn\u00e9 ot\u00e1zky za p\u00e1r sek\u00fand. Probl\u00e9m je, \u017ee odpove\u010f s\u00edce p\u00f4sob\u00ed presved\u010divo, no…\n","protected":false},"author":2,"featured_media":53030,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[32],"tags":[223,21580,21581,21582,6085,8716,21583,34,37,21584,33,43,40,39,42,41,36,5342,21585,38,35,463],"class_list":{"0":"post-53029","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-spravy","8":"tag-ai","9":"tag-ai-doktor","10":"tag-ai-model","11":"tag-ai-nedava-spravne-informacie","12":"tag-chatgpt","13":"tag-gemini","14":"tag-grok","15":"tag-headlines","16":"tag-hlavne-spravy","17":"tag-nespravne-informacie","18":"tag-news","19":"tag-sk","20":"tag-slovak","21":"tag-slovakia","22":"tag-slovencina","23":"tag-slovensko","24":"tag-spravy","25":"tag-studia","26":"tag-testovanie-presnosti","27":"tag-titulky","28":"tag-top-stories","29":"tag-zdravie"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@sk\/116462703317365350","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/posts\/53029","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/comments?post=53029"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/posts\/53029\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/media\/53030"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/media?parent=53029"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/categories?post=53029"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/sk\/wp-json\/wp\/v2\/tags?post=53029"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}