B\u00f6y\u00fck dil modell\u0259rin\u0259 \u0259saslanan s\u00fcni intellekt (AI) s\u0259hvl\u0259r ed\u0259, t\u0259k bir cavab daxilind\u0259 ziddiyy\u0259t t\u0259\u015fkil ed\u0259 v\u0259 m\u00fcxt\u0259lif sui-q\u0259sd n\u0259z\u0259riyy\u0259l\u0259ri d\u0259 daxil olmaqla z\u0259r\u0259rli dezinformasiya yaya bil\u0259r.<\/p>\n
Bu n\u0259tic\u0259y\u0259 \u201cChatGPT\u201d chatbotunun m\u00fcxt\u0259lif informasiya t\u0259sirl\u0259rin\u0259 qar\u015f\u0131 m\u00fcqavim\u0259tini t\u0259dqiq ed\u0259n Vaterloo Universitetinin kanadal\u0131 dil\u00e7il\u0259ri g\u0259libl\u0259r. T\u0259dqiqat Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP) elmi m\u0259qal\u0259l\u0259r toplusunda d\u0259rc olunub.<\/p>\n
M\u00fct\u0259x\u0259ssisl\u0259r GPT-3 modelini alt\u0131 kateqoriya \u00fczr\u0259 m\u00fcxt\u0259lif ifad\u0259l\u0259ri ba\u015fa d\u00fc\u015fm\u0259k \u00fc\u00e7\u00fcn s\u0131naqdan ke\u00e7iribl\u0259r: sui-q\u0259sd n\u0259z\u0259riyy\u0259l\u0259ri, ziddiyy\u0259tl\u0259r, yanl\u0131\u015f t\u0259s\u0259vv\u00fcrl\u0259r, stereotipl\u0259r, fantastika v\u0259 faktlar. S\u00fcni intellekt\u0259 1,2 mind\u0259n \u00e7ox m\u00fcxt\u0259lif ifad\u0259l\u0259r t\u0259klif edilib v\u0259 onlardan h\u0259r birini d\u00f6rd meyara g\u00f6r\u0259 qiym\u0259tl\u0259ndirm\u0259yi xahi\u015f edibl\u0259r: fakt v\u0259 ya uydurma, onun real d\u00fcnyada olub-olmamas\u0131, elmi n\u00f6qteyi-n\u0259z\u0259rd\u0259n do\u011fru olub-olmamas\u0131 v\u0259 olub-olmamas\u0131. subyektiv bax\u0131mdan do\u011frudur.<\/p>\n
Cavablar\u0131n t\u0259hlili g\u00f6st\u0259rdi ki, GPT-3 kateqoriyadan as\u0131l\u0131 olaraq 26%-\u0259 q\u0259d\u0259r yalan ifad\u0259l\u0259ri t\u0259sdiql\u0259yib. T\u0259hlil g\u00f6st\u0259rdi ki, sual\u0131n m\u0259tnind\u0259ki ki\u00e7ik d\u0259yi\u015fiklikl\u0259r bel\u0259 neyron \u015f\u0259b\u0259k\u0259nin cavab\u0131na t\u0259sir ed\u0259 bil\u0259r.<\/p>\n
M\u0259nb\u0259: Haberler.com<\/p>\n","protected":false},"excerpt":{"rendered":"
B\u00f6y\u00fck dil modell\u0259rin\u0259 \u0259saslanan s\u00fcni intellekt (AI) s\u0259hvl\u0259r ed\u0259, t\u0259k bir cavab daxilind\u0259 ziddiyy\u0259t t\u0259\u015fkil ed\u0259 v\u0259 m\u00fcxt\u0259lif sui-q\u0259sd n\u0259z\u0259riyy\u0259l\u0259ri d\u0259 daxil olmaqla z\u0259r\u0259rli dezinformasiya yaya bil\u0259r. Bu n\u0259tic\u0259y\u0259 \u201cChatGPT\u201d chatbotunun m\u00fcxt\u0259lif informasiya t\u0259sirl\u0259rin\u0259 qar\u015f\u0131 m\u00fcqavim\u0259tini t\u0259dqiq ed\u0259n Vaterloo Universitetinin kanadal\u0131 dil\u00e7il\u0259ri g\u0259libl\u0259r. T\u0259dqiqat Proceedings of the 3rd Workshop on Trustworthy Natural Language Processing (TrustNLP) […]<\/p>\n","protected":false},"author":2,"featured_media":37640,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[36,1],"tags":[],"class_list":["post-37639","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-media","category-society"],"yoast_head":"\n