Ngokombiko we-Phys.org wangaphandle namuhla (ngomhla ka-24), ucwaningo olusha olwenziwe yiNyuvesi yaseCardiff neNyuvesi yaseMelbourne luthole ukuthi ubuhlakani bokwenziwa obukhiqizayo (GenAI) okwamanje abukakwazi ukuhlola ngokwethembeka imisebenzi ebhaliwe yabafundi ngendlela efana nothisha abangabantu.
Abacwaningi basebenzise izinguqulo ezimbili ze-ChatGPT ukuze banikeze amamaki ezindabeni ezingu-50 zabafundi abenza iziqu zesayensi yezinto eziphilayo, behlola ikhono lamamodeli amakhulu lokuhlola imisebenzi yabafundi. I-ChatGPT kwadingeka ihlole lezi zindatshana ngokwemigomo engu-7, kanti abacwaningi basebenzise nezindlela ezi-4 ezihlukene zokuyinikeza iziqondiso. Kamuva baqhathanisa amamaki amaphakathi nokuhlukahluka kwamamaki okunikezwe imodeli enkulu nalokho okunikezwe abantu.
UDkt William Kay weSikole Sesayensi Yezinto Eziphilayo eNyuvesi yaseCardiff uthe: “Imiphumela yocwaningo ikhombisa ukuthi amamaki anikezwe imodeli ahlukahluka kakhulu futhi awakwazi ukubikezela ngokunembile amamaki abantu. Kukhona umehluko obonakalayo lapho i-GenAI nabantu behlola iqoqo elifanayo lezindatshana. Uma kubhekwa kuphela imaki eliphelele lendatshana, imiphumela yabo isondelene; kodwa uma kubhekwa umgomo ngamunye wokumaka nendatshana ngayinye, umehluko phakathi kwemodeli enkulu nokuhlolwa kwabantu uba mkhulu kakhulu.”
Ngaphandle kwesimo esisodwa, amamodeli e-AI ngokuvamile anikeze amamaki amaphakathi aphezulu kunalawo abantu. Emaqhathanisweni amamaki amaphakathi, umehluko omkhulu ufinyelele kumaphuzu angu-16.1; ephepheni elilodwa, umehluko omkhulu ufinyelele kumaphuzu angu-40.
UWilliam Kay uphinde wathola ukuthi i-AI ivame ukwehlisa amamaki ezindatshana ezithole amamaki aphezulu, bese iphakamisa amamaki emisebenzi ethole amamaki aphansi, okwenza amamaki agxile ngokuhlelekile phakathi nendawo.
Imiphumela ikhombisa ukuthi, okungenani kulesi sigaba, ngisho nangemva kokuqeqeshwa okubanzi, i-AI ayikakwazi ukunikeza imisebenzi ebhaliwe enobungako obukhulu bokuzithoba amamaki athembekile afana nalawo abantu. “Amamodeli amakhulu esiwahlolile okwamanje awafanele ukuthatha indawo yothisha ekubikezeleni amamaki emisebenzi yabafundi.”
Abacwaningi baveze ukuthi umkhakha wemfundo ephakeme unentshisekelo enkulu ekutheni amamodeli amakhulu angasebenzisa yini ikhono lawo lokubona amaphethini ukuze enze ukumakwa kwemisebenzi yabafundi kube nenhloso kakhudlwana, ngesikhathi kukhuphula ukusebenza kahle kokumaka nokunciphisa umthwalo wabasebenzi bezemfundo. Kodwa lolu cwaningo lukhombisa ukuthi okwamanje lokhu akufanele kwenziwe.
Ngaphezu kwalokho, ukuhambisa umsebenzi ku amathuluzi e-AI ngaphandle kwemvume ecacile yomfundi kuhilela nezindaba zokuziphatha; amamodeli amakhulu awakwazi futhi akufanele kuncike kuwo ukuze amake imisebenzi emide ebhaliwe yabafundi. “Njengoba amamodeli amakhulu eqhubeka nokuthuthuka, ikhono lawo lokulingisa ukwahlulela kwabantu lingase lithuthuke esikhathini esizayo. Kodwa ngokwalolu cwaningo, ukwenza amamaki e-AI ahambisane ngempela nokuhlolwa kwabantu cishe ngeke kube lula.”