Modeli m\u00eb i ri i inteligjenc\u00ebs artificiale, ChatGPT-5, ka b\u00ebr\u00eb nj\u00eb p\u00ebrparim t\u00eb duksh\u00ebm n\u00eb reduktimin e halucinacioneve, nj\u00eb fenomen q\u00eb ka shqet\u00ebsuar p\u00ebrdoruesit e tij. Sipas testeve t\u00eb fundit, ky model \u00ebsht\u00eb m\u00eb pak i prirur ndaj shpikjes s\u00eb informacionit t\u00eb rrem\u00eb, duke ofruar nj\u00eb nivel m\u00eb t\u00eb lart\u00eb sakt\u00ebsie dhe besueshm\u00ebrie.<\/b><\/p>\n
Rezultatet e analizave tregojn\u00eb se shkalla e halucinacioneve t\u00eb ChatGPT-5 \u00ebsht\u00eb 1.4 p\u00ebrqind, nj\u00eb p\u00ebrmir\u00ebsim i duksh\u00ebm krahasuar me paraardh\u00ebsit e tij, GPT-4o dhe GPT-4, t\u00eb cil\u00ebt kishin shkall\u00eb halucinacionesh prej 1.49% dhe 1.8% p\u00ebrkat\u00ebsisht. Kjo tregon se zhvilluesit e OpenAI kan\u00eb b\u00ebr\u00eb p\u00ebrpjekje t\u00eb konsiderueshme p\u00ebr t\u00eb adresuar k\u00ebt\u00eb problem kritik.<\/p>\n
Megjithat\u00eb, disa modele t\u00eb tjera kan\u00eb arritur rezultate edhe m\u00eb t\u00eb mira. P\u00ebr shembull, o3-mini High Reasoning ka regjistruar vet\u00ebm 0.795% halucinacione, nd\u00ebrsa GPT-4.5 Preview \u00ebsht\u00eb n\u00eb 1.2%. K\u00ebto t\u00eb dh\u00ebna sugjerojn\u00eb se ka ende hap\u00ebsir\u00eb p\u00ebr p\u00ebrmir\u00ebsim n\u00eb fush\u00ebn e inteligjenc\u00ebs artificiale.<\/p>\n
Nj\u00eb tjet\u00ebr konkurrent, Grok-4 nga xAI, ka dal\u00eb t\u00eb jet\u00eb m\u00eb i prirur ndaj fabrikimit t\u00eb informacionit, me nj\u00eb shkall\u00eb halucinacionesh deri n\u00eb 4.8%. Kjo e b\u00ebn at\u00eb nj\u00eb nga modelet m\u00eb problematike n\u00eb treg, duke ngritur shqet\u00ebsime p\u00ebr sakt\u00ebsin\u00eb e informacionit q\u00eb ofron.<\/p>\n
Edhe pse ChatGPT-5 \u00ebsht\u00eb teknikisht m\u00eb i avancuar, disa p\u00ebrdorues kan\u00eb shprehur shqet\u00ebsime p\u00ebr natyr\u00ebn e tij m\u00eb t\u00eb ftoht\u00eb dhe p\u00ebr p\u00ebrgjigjet m\u00eb t\u00eb shkurtra krahasuar me GPT-4o. Kjo ka \u00e7uar n\u00eb nj\u00eb debat t\u00eb gjer\u00eb mbi balancimin e sakt\u00ebsis\u00eb dhe kreativitetit n\u00eb inteligjenc\u00ebn artificiale.<\/p>\n
Drejtori ekzekutiv i OpenAI, Sam Altman, ka pranuar se kompania b\u00ebri nj\u00eb gabim duke hequr modelet e vjetra pa paralajm\u00ebrim. Ai njoftoi se GPT-4o do t\u00eb vihej p\u00ebrs\u00ebri n\u00eb dispozicion p\u00ebr nj\u00eb periudh\u00eb t\u00eb caktuar, duke ofruar p\u00ebrdoruesve mund\u00ebsin\u00eb p\u00ebr t\u00eb zgjedhur mes modeleve t\u00eb ndryshme.<\/p>\n
Altman gjithashtu premtoi p\u00ebrmir\u00ebsime t\u00eb reja, duke p\u00ebrfshir\u00eb nj\u00eb modalitet t\u00eb ri t\u00eb t\u00eb menduarit p\u00ebr detyra m\u00eb komplekse dhe nj\u00eb nd\u00ebrrim automatik m\u00eb t\u00eb mir\u00eb midis versioneve. K\u00ebto ndryshime pritet t\u00eb rrisin p\u00ebrvoj\u00ebn e p\u00ebrdoruesve dhe t\u00eb adresojn\u00eb shqet\u00ebsimet e shprehura deri tani.<\/p>\n
Megjithat\u00eb, Grok-4 \u00ebsht\u00eb kritikuar p\u00ebr modalitetin e tij “Pikant”, i cili ka gjeneruar p\u00ebrmbajtje t\u00eb pap\u00ebrshtatshme dhe materiale t\u00eb rreme, pavar\u00ebsisht filtrave t\u00eb integruar. Ky fenomen tregon se, pavar\u00ebsisht p\u00ebrparimeve teknologjike, sfidat e krijimit t\u00eb p\u00ebrmbajtjes s\u00eb besueshme mbeten nj\u00eb problem i r\u00ebnd\u00ebsish\u00ebm n\u00eb fush\u00ebn e inteligjenc\u00ebs artificiale.<\/p>\n","protected":false},"excerpt":{"rendered":"
Modeli m\u00eb i ri i inteligjenc\u00ebs artificiale, ChatGPT-5, ka b\u00ebr\u00eb nj\u00eb p\u00ebrparim t\u00eb duksh\u00ebm n\u00eb reduktimin e halucinacioneve, nj\u00eb fenomen q\u00eb ka shqet\u00ebsuar p\u00ebrdoruesit e tij. Sipas testeve t\u00eb fundit, ky model \u00ebsht\u00eb m\u00eb pak i prirur ndaj shpikjes s\u00eb informacionit t\u00eb rrem\u00eb, duke ofruar nj\u00eb nivel m\u00eb t\u00eb lart\u00eb sakt\u00ebsie dhe besueshm\u00ebrie. Rezultatet e […]<\/p>\n","protected":false},"author":1,"featured_media":343766,"comment_status":"","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[174],"tags":[],"class_list":["post-343767","post","type-post","status-publish","format-standard","has-post-thumbnail","category-teknologji"],"_links":{"self":[{"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/posts\/343767","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/comments?post=343767"}],"version-history":[{"count":0,"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/posts\/343767\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/media\/343766"}],"wp:attachment":[{"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/media?parent=343767"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/categories?post=343767"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infokosova.net\/new\/wp-json\/wp\/v2\/tags?post=343767"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}