Kwenzekeni
I-Guardian ibika ukuthi uphenyo lwasephalamende lase-Australia luthintwa ulwazi olungaqondile olukhiqizwe yi-AI, okuhlanganisa nezifundo eziqanjiwe nezingcaphuno ezingalungile ezishiwo izifundiswa nababhali bangempela. Ukuhlaziya kwayo kukhombe okungenani okuthunyelwe okungu-39 okuqukethe izinkomba ezibonakala zingekho, kuyilapho amaphepha angaphezu kwe-100 afaka omaka be-URL be-ChatGPT kuzixhumanisi zereferensi.
I-Guardian ibika ukuthi inqwaba yezinto ezithunyelwe kuyo yonke imikhakha yezombangazwe emibuzweni yamanje yasephalamende yase-Australia iqukethe izinkomba ezibonakala sengathi zakhiwe noma zashintshwa yizinhlobo zezilimi ezinkulu. Amaphutha abikiwe ahlanganisa izifundo eziqanjiwe, izindatshana ezingekho, imininingwane yekhasi engalungile, nabacwaningi bangempela abahlotshaniswa nomsebenzi abangawenzanga. Izethulo zibandakanya izindawo zenqubomgomo ezihlanganisa udlame lomndeni, ukuzibulala, ukungalingani kwezindlu kanye nolwazi olungaqondile lwesimo sezulu.
Lo mbiko uchaza ukuhanjiswa kophenyo olwenziwa ngodlame lomndeni nokuzibulala oluthi ireferensi engekho kanye nemiphumela yocwaningo engalungile ku-Divna Haslam, uprofesa osebenzisana naye waseNyuvesi yaseQueensland nodokotela wezengqondo. U-Haslam utshele i-Guardian ukuthi lokhu okucashuniwe kubukeka kuzwakala ngokwanele ukudukisa umuntu owenza isibuyekezo esingesihle. Inhlangano eyethulayo, i-Drilldown Reports, ithe isebenzise i-AI ohlelweni lwayo lokucwaninga futhi ngokuhamba kwesikhathi yabona amaphutha, kodwa yeqiwe umnqamulajuqu wokulungisa ukuthunyelwa kwasekuqaleni.
I-Guardian ithi yakhe uhlelo lwangokwezifiso olukhipha izinkomba emibhalweni yophenyo futhi yasesha ngokumelene ne-Crossref kanye ne-Google Scholar. Iphinde yahlola izihlonzi zezinto zedijithali lapho zitholakala khona futhi amadokhumenti abuyekezwa mathupha lapho okungenani u-20% wezithenjwa zingazange zifaniswe. Isibalo esibe umphumela kwaba okungenani izethulo ezingama-39 ezinereferensi ebonakala ihunyushwe. I-Guardian iphawule leyo nombolo njengelondoloza ngenxa yokuthi indlela ingathola kuphela amaphutha kumadokhumenti aqukethe izinkomba, hhayi umbhalo ongasekelwe okhiqizwe ngaphandle kwezingcaphuno.
Uphenyo luphinde lwathola amaphepha angaphezu kwe-100 aqukethe omaka be-URL be-ChatGPT kuzixhumanisi eziyisethenjwa. I-Guardian ixwayisa ngokuthi omaka abanjalo akufakazeli ukuthi umuntu usebenzise i-ChatGPT ngokuqondile, ngoba impahla kungenzeka ikopishwe emthonjeni wenkampani yangaphandle. Iphinde ithi izingcaphuno ezingafani zingakhombisa ukusetshenziswa kwe-AI kodwa zingakusunguli kuzo zonke izimo. Ngakho-ke umbiko wethula okutholakele njengobufakazi bamaphutha abonakalayo akhiqizwe yi-AI noma i-AI, hhayi njengesilinganiso esiqondile sakho konke ukusetshenziswa kwe-AI ezithulweni zasephalamende.
Imininingwane yomthombo: theguardian.com ↗
Kungani kubalulekile
Imibuzo yasePhalamende ihloselwe ukunikeza abenzi bomthetho ubufakazi obuvela kochwepheshe, izinhlangano kanye namalungu omphakathi. Uma izingcaphuno eziqanjiwe zingena kuleyo nqubo futhi zithathwe njengezinokwethembeka, izinqumo zingase zisekelwe ocwaningweni olungekho futhi ukuzethemba komphakathi ezikhungweni zentando yeningi kungase kube buthaka.
Iziphakamiso zophenyo zePhalamende zingabumba indlela amakomiti aqonda ngayo izinkinga zomphakathi futhi ahlole imithetho ehlongozwayo noma izinhlelo. I-Guardian ibika ukuthi eminye imibiko yekomiti icaphuna izethulo eziqukethe izinkomba ezibonakala zingamampunge. Lokho kudala ingozi yokuthi isimangalo esingasekelwe singathola isisindo sesikhungo ngenxa nje yokuthi singene ohlelweni olusemthethweni lokuqoqa ubufakazi.
Ukulimala okungokoqobo akukhawulelwe kumbhalo waphansi owodwa ongalungile. Abacwaningi abacashunwe yi-Guardian bathi izingcaphuno eziqanjiwe zinganqamula iketango phakathi kwesimangalo senqubomgomo kanye nobufakazi okuthiwa bayasisekela. Ezindaweni ezinjengodlame lwasekhaya, ezempilo, ukuthuthela kwelinye izwe kanye nenqubomgomo yesimo sezulu, izinqumo ezisekelwe ocwaningweni olungekho zingaqondisa kabi ukunaka, zihlanekezele inkulumo-mpikiswano noma zibukele phansi ukuvikelwa kwabantu abathintekile. Umthombo awuqinisekisi ukuthi noma iyiphi inqubomgomo ethile yase-Australia yenziwe ngenxa yengcaphuno ekhohliwe, ukuze umphumela uhlale uyingozi esikhundleni somphumela obonisiwe.
Isiqephu siphinde sibonisa iluphu yempendulo efaka i-AI ekhiqizayo namasistimu okusesha. I-Guardian ibika ukuthi isifinyezo se-AI se-Google kwesinye isikhathi siphatha isithenjwa esingumgunyathi njengesangempela, kuyilapho izinhlelo ze-ChatGPT kanye ne-Google zingasho ukuthunyelwa kombuzo okuqukethe iphutha. Lokho kungenza idokhumenti engalungile ibonakale iqinisekisiwe ngokuzimele ngisho nalapho ucwaningo oluyisisekelo lungekho. Umbiko awuhloli ngokuzimele ububanzi obugcwele bezimo lapho lezi zinhlelo zikhiqiza khona izifinyezo ezinjalo, kodwa ubhala ubungozi ngezibonelo ozihlolile.
Udaba olubanzi lwesikhungo wukuziphendulela. Isiqondiso seSenethi sase-Australia kubikwa ukuthi sixwayisa abahambisi ukuthi i-AI ingathinta ikhwalithi yolwazi nokuthi ukunemba kuhlala kuwumthwalo wemfanelo womhambisi. Usihlalo wekomiti ocashunwe yi-Guardian uthe amakomiti athola izinto ezisezingeni elihlukile futhi kufanele aphenye ubufakazi ngesikhathi sophenyo. Umthombo ushiya kusobala ukuthi amakomiti njengamanje anabo abasebenzi, amathuluzi ezobuchwepheshe noma izinqubo ezidingekayo ukuze enze lokho kuhlola ngokungaguquki.
I-Interactive Mechanism: Indlela Esebenza Ngayo Ngempela
Hlola ubuchwepheshe obuyisisekelo ngemuva kwalokhu kuthuthukiswa ngokuhlanganyela.
Why can ethical evaluation not be reduced to one model score?
Ongakubuka ngokulandelayo
Imibuzo esheshayo eyokuthi amakomidi ephalamende azokwethula imithetho eqinile yokuqinisekisa noma yokudalula, ukuthi abahambisi balungisa kanjani izinto esezifayiliwe, nokuthi izikhungo zikahulumeni zinezinsiza ezanele zokuhlola izinkomba. I-Guardian iphinde yafuna izimpendulo ku-OpenAI kanye ne-Google mayelana neqhaza labo ekuqiniseni izithenjwa ezingalungile ngamasistimu e-AI nezifinyezo zosesho.
Bheka izinguquko ezisemthethweni ekuqondisweni kokuthunyelwe kanye nezinqubo zokubuyekeza ikomidi. I-Guardian ibika ukuthi uChristian Downie, uprofesa wase-Australian National University, ucele iziqondiso ezintsha ezingakhuthaza ukuba neqiniso kuyilapho kulondolozwa inqubo yophenyo evulekile. Izinyathelo ezingenzeka zingabandakanya ukudalulwa okucacile kosizo lwe-AI, ukuhlola okuyisibopho kwemithombo ekhonjiwe, amawindi okulungisa ngemva kokugcwalisa, noma izidingo eziqinile zezinhlangano eziletha ubufakazi obusekelwe ocwaningweni. Umthombo awusho ukuthi kukhona izinguquko ezinjengalezi ezisamukelwe.
Buka ukuthi ingabe abacwaningi abathintekayo nabahambisi bayafuna yini ukulungiswa noma ukucaciselwa esidlangalaleni. I-Guardian ibika ukuthi uNicole Gurran wayexhunywe emsebenzini owawungekho nokuthi uMargaret Simons ngokusobala wayehlotshaniswa nesihloko esingekho. Imibiko ye-Drilldown ithe ithole amaphutha ekuthumeleni okulandelayo, kanti iNational Rational Energy Network ayizange iphendule esicelweni se-Guardian sokuphawula. Isimo sesiphakamiso ngasinye esiyingxabano kanye nokuthi amakomiti ayayichibiyela yini imibiko yawo kusewumbuzo obalulekile ongakaxazululwa.
Buka ubufakazi mayelana nesikali nokusatshalaliswa kwenkinga ngale kwamadokhumenti anezinkomba ezisemthethweni. Uhlelo lwe-Guardian alukwazi ukukhomba iphrozi ekhiqizwe yi-AI engaqukethe izingcaphuno, futhi ixwayisa ngokuthi inani layo elithunyelwe okungu-39 likubukela phansi ukusetshenziswa kwe-AI. Ngesikhathi esifanayo, izitholi ze-AI zezentengiselwano zingakhiqiza amaphuzu angamanga, ngakho-ke noma iyiphi isistimu yokuphoqelela yesikhathi esizayo izodinga ukuhlukanisa izinsolo ezingasekelwe ekuhlulekeni kwengcaphuno okuqinisekiswa.
Buka ukuthi i-OpenAI ne-Google ziphendula kanjani ku-loop yempendulo ebikiwe. I-OpenAI itshele i-Guardian ukuthi ukuncishiswa kwemibono engemihle kusalokhu kuyindawo yocwaningo eqhubekayo futhi yeluleka abasebenzisi ukuthi baphathe i-ChatGPT njengohlaka lokuqala esikhundleni somthombo wokugcina, oqinisekisa izingcaphuno, idatha kanye nezinkomba zangaphandle. I-Google ithe i-AI Overviews yayo ihlose ukufanisa okuqukethwe kwewebhu namagama okubuza, afana nosesho oluvamile. Lezo zimpendulo azixazululi ukuthi izifinyezo ze-AI kufanele ziveze noma zikhiqize kabusha izimangalo ezivela kumadokhumenti aqukethe izithenjwa eziqanjiwe, okushiya ukusebenza kwezivikelo ezikhona kungaqinisekile.