{"id":166425,"date":"2026-03-26T17:13:13","date_gmt":"2026-03-26T15:13:13","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/intervyu-s-gregom-kroa-hartmanom-ob-otchyotah-ob-oshibkah-sgenerirovannyh-ai"},"modified":"2026-03-26T17:13:13","modified_gmt":"2026-03-26T15:13:13","slug":"intervyu-s-gregom-kroa-hartmanom-ob-otchyotah-ob-oshibkah-sgenerirovannyh-ai","status":"publish","type":"post","link":"https:\/\/prohoster.info\/sq\/blog\/news\/intervyu-s-gregom-kroa-hartmanom-ob-otchyotah-ob-oshibkah-sgenerirovannyh-ai","title":{"rendered":"Intervista me Greg Kroa-Hartmanin mbi raportet e gabimeve, t\u00eb gjeneruara nga AI","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
N\u00eb konferenc\u00ebn KubeCon Europe, edicioni i The Register intervistoi Greg Kroah-Hartman, i cili \u00ebsht\u00eb p\u00ebrgjegj\u00ebs p\u00ebr mb\u00ebshtetjen e deg\u00ebve stabile dhe 'staging' t\u00eb kernelit Linux dhe mban postin e maintainer-it n\u00eb 16 n\u00ebn-sistemat e kernelit. N\u00eb intervist\u00eb, Greg ndan pik\u00ebpamjet e tij rreth raporteve t\u00eb gabimeve q\u00eb gjendet nga AI. AI n\u00eb kernel tashm\u00eb \u00ebsht\u00eb p\u00ebrdorur n\u00eb rishikimin e ndryshimeve n\u00eb n\u00ebn-sistemin e rrjetit, eBPF dhe DRM, p\u00ebr m\u00eb tep\u00ebr, p\u00ebr kontrollin e ndryshimeve q\u00eb d\u00ebrgohen, \u00ebsht\u00eb integruar s\u00eb fundmi nj\u00eb mjet i quajtur Sashiko nga Google. <\/p>\n
Disa nga shprehjet e Gregut: <\/p>\n
\n
\u2018Disa muaj m\u00eb par\u00eb merrnim at\u00eb q\u00eb e quanim plehra nga AI \u2014 raporte mbi sigurin\u00eb t\u00eb gjeneruara nga AI, t\u00eb cilat ishin duksh\u00ebm t\u00eb gabuara ose me cil\u00ebsi t\u00eb ul\u00ebt. Ishte madje arg\u00ebtuese. Nuk ishim shum\u00eb t\u00eb shqet\u00ebsuar p\u00ebr k\u00ebt\u00eb... Nj\u00eb muaj m\u00eb par\u00eb ndodhi di\u00e7ka, dhe situata u ndryshua ndjesh\u00ebm. Tani na vijn\u00eb raporte reale.\u2019\n
\u2018Nj\u00eb situat\u00eb e till\u00eb nuk v\u00ebrehet vet\u00ebm n\u00eb Linux \u2014 t\u00eb gjitha projektet me kod t\u00eb hapur po marrin raporte reale, t\u00eb krijuara nga AI, dhe tani ato jan\u00eb me cil\u00ebsi dhe valide. Ekipet p\u00ebrgjegj\u00ebse p\u00ebr sigurin\u00eb n\u00eb projektet m\u00eb t\u00eb m\u00ebdha t\u00eb hapura, gjat\u00eb bisedave informale mes vete, regjistrojn\u00eb tendenc\u00ebn e nj\u00ebjt\u00eb.\u2019\n
Kur u pyet se \u00e7far\u00eb e shkaktonte k\u00ebt\u00eb, Greg u p\u00ebrgjigj: \u2018Nuk e dim\u00eb. Duket se askush nuk e di. Ose shum\u00eb mjete jan\u00eb b\u00ebr\u00eb ndjesh\u00ebm m\u00eb t\u00eb mira, ose njer\u00ebzit kan\u00eb filluar t\u00eb thon\u00eb: \u2018Hej, le t\u00eb merremi me k\u00ebt\u00eb.\u2019 Duket se kjo i p\u00ebrket shum\u00eb grupeve dhe kompanive t\u00eb ndryshme. Sa i p\u00ebrket kernelit, ne do ta p\u00ebrballojm\u00eb k\u00ebt\u00eb. Ekipa jon\u00eb \u00ebsht\u00eb b\u00ebr\u00eb shum\u00eb m\u00eb e madhe, \u00ebsht\u00eb shum\u00eb e shp\u00ebrndar\u00eb n\u00eb mbar\u00eb bot\u00ebn, dhe rritja jon\u00eb \u00ebsht\u00eb nj\u00eb realitet dhe nuk po ngadal\u00ebsohet. K\u00ebto jan\u00eb gj\u00ebra t\u00eb vogla, asgj\u00eb serioze, por ndihma n\u00eb k\u00ebt\u00eb \u00e7\u00ebshtje mund t\u00eb jet\u00eb e nevojshme p\u00ebr t\u00eb gjith\u00eb projektet e hapura. Projeketet m\u00eb t\u00eb vogla kan\u00eb shum\u00eb m\u00eb pak mund\u00ebsi p\u00ebr t\u00eb p\u00ebrballuar nj\u00eb fluks papritur t\u00eb raporteve t\u00eb gabimeve dhe dob\u00ebsive t\u00eb gjeneruara nga AI, n\u00eb t\u00eb cilat p\u00ebrmenden gabime t\u00eb v\u00ebrteta, e jo plehra.\u2019\n
Gregu i tregoi se kur k\u00ebrkoi nga AI t\u00eb gjej\u00eb gabimet nga grupi i propozimeve \u2014 ai gjeti 60 dhe ofroi patch p\u00ebr t'i rregulluar ato. Vet\u00ebm nj\u00eb e treta e gabimeve t\u00eb gjetura ishin t\u00eb tilla, dhe nga patch-et vet\u00ebm 2\/3 rezultuan t\u00eb ishin zgjidhje t\u00eb sakta, q\u00eb nuk k\u00ebrkonin p\u00ebrmir\u00ebsim, por \u00e7do gj\u00eb ishte shum\u00eb larg duke qen\u00eb e padobishme. Sipas Gregut, mbajt\u00ebsit nuk mund ta injorojn\u00eb k\u00ebt\u00eb, sidomos tani q\u00eb rezultatet e pun\u00ebs s\u00eb AI po b\u00ebhen gjithnj\u00eb e m\u00eb t\u00eb mira. P\u00ebr t\u00eb sh\u00ebnuar patch-et e krijuara n\u00ebp\u00ebrmjet AI, u shtua nj\u00eb etiket\u00eb \"Co-developed:\". Megjithat\u00eb, p\u00ebrkund\u00ebr p\u00ebrpjekjeve t\u00eb ve\u00e7anta p\u00ebr t\u00eb aplikuar AI p\u00ebr t\u00eb krijuar funksionalitete t\u00eb reja, kryesisht AI n\u00eb thelb p\u00ebrdoret p\u00ebr rishikimin e ndryshimeve.\n
Nj\u00eb nga p\u00ebrfitimet m\u00eb t\u00eb spikatur t\u00eb AI quhet reduktimi i koh\u00ebs s\u00eb trajtimit t\u00eb patch-eve. Kur asistenti AI identifikon probleme t\u00eb dukshme, autor\u00ebt e patch-eve marrin nj\u00eb feedback shum\u00eb p\u00ebrpara se mbajt\u00ebsi real t\u00eb ket\u00eb koh\u00eb t\u00eb lexoj\u00eb patch-in: \"N\u00ebse un\u00eb shoh q\u00eb sistemi reagon p\u00ebr di\u00e7ka, ai i jep feedback autorit m\u00eb shpejt se sa mund ta b\u00ebnte mbajt\u00ebsi, dhe kjo \u00ebsht\u00eb shum\u00eb e shk\u00eblqyer. Ne kemi gjithashtu disa bota q\u00eb kontrollojn\u00eb patch-et. N\u00ebse un\u00eb v\u00eb re se ata japin nj\u00eb gabim, un\u00eb menj\u00ebher\u00eb kuptoj se si mbajt\u00ebs, nuk kam nevoj\u00eb t\u00eb shikoj fare. Dhe zhvilluesi mendon 'O, mund t\u00eb b\u00ebj nj\u00eb version tjet\u00ebr nes\u00ebr', q\u00eb ndihmon pak p\u00ebr t\u00eb p\u00ebrmir\u00ebsuar feedback-un.\" <\/ul>\n