Isikhulu esiphezulu se-Hugging Face, uClément Delangue, muva nje sikhulumele amamodeli e-AI anesisindo esivulekile (Open-weight), ngesikhathi sixwayisa iMelika ngokuthi i-China isiqala ukuba nenzuzo emjahweni wezobuhlakani bokwenziwa.
Engxoxweni ayenze ngoMsombuluko nohlelo lwe-CNBC i-“Squawk on the Street”, uDelangue uthe i-China ihamba phambili emjahweni we-AI ngoba ivuleleke kakhulu kuneMelika kwezesayensi evulekile namamodeli avulekile.
Ngesikhathi umethuli uSara Eisen embuza ukuthi uyakholelwa yini ukuthi i-China iyawunqoba umjaho we-AI, uDelangue waphendula: “Ngicabanga ukuthi kunjalo ngempela. I-China yakhe isimiso semvelo lapho abantu befunda khona komunye nomunye futhi bathuthuke ngokubambisana, ngakho ubuchwepheshe buthuthuka ngokushesha kakhulu kuneMelika. Kanti amanye amalabhorethri aseMelika aphambili alwa wodwa, avaleleke komunye nomunye, futhi awabelani ngemiphumela yawo naso sonke isimiso semvelo.”
Uqhubeke wathi, uma kubhekwa ijubane lamanje lentuthuko, ekupheleni kwalo nyaka noma ngonyaka ozayo, i-China ayigcini nje ngokuqhubeka ihola emkhakheni wamamodeli avulekile, kodwa ingase ibuse ngisho nasemkhakheni wonke we-AI esezingeni eliphambili.
Impikiswano ngamamodeli e-AI anesisindo esivulekile ishube kakhulu muva nje. Lokhu kubangelwe yisigameko esenzeka ngoJulayi kulo nyaka: i-ejenti ehlakaniphile evela kwa-OpenAI yaphuma endaweni yokuqeqeshwa yabe isingena enkundleni ye-Hugging Face. UDelangue wake wadalula ukuthi ngaleso sikhathi inkampani yasebenzisa imodeli yomthombo ovulekile i-GLM 5.2, eyethulwe yi-Zhipu, ukusiza ukuvikela izinhlelo zenkampani.
Uthe: “Owasihlasela kwakuyimodeli yangasese eyayingakakhishwa futhi eyayithuthukiswa endaweni evaliwe. Thina sakwazi ukuzivikela kuphela ngokuthembela kumamodeli avulekile, ngoba imingcele yokuphepha ye-API (guardrails) yenza singakwazi ukusebenzisa lawo mamodeli avaliwe ukuze sibhekane nalokhu.”
UDelangue ukholelwa ukuthi ukuhlaselwa okuningi kwe-AI esikhathini esizayo kungase kuqhamuke kumamodeli angasese anomthombo ovaliwe, ngoba abahlaseli ngeke balandele imingcele yokuphepha noma imigomo yokusebenzisa izinsizakalo. Amamodeli avulekile yiwona okungenzeka athwale ingxenye enkulu yomsebenzi wokuvikela.
Lezi zitatimende zakhe zivele ngesikhathi abenzi bezinqubomgomo eMelika bexoxa ngokuthi kufanele yini kubekwe imingcele eqinile kumamodeli e-AI anesisindo esivulekile avela e-China.
Ekupheleni kukaJulayi kulo nyaka, izinkampani zobuchwepheshe ezingaphezu kwamashumi amabili, okuhlanganisa i-Nvidia, i-Microsoft, i-Meta ne-OpenAI, zasayina ngokuhlanganyela incwadi eya kuhulumeni waseMelika, zicela ukuthi kungacindezelwa noma kubekwe imingcele kumamodeli e-AI anesisindo esivulekile. Ezinsukwini ezimbalwa kamuva, i-CEO ye-Anthropic uDario Amodei naye wachaza isimo senkampani kokuthunyelwe kwebhulogi. Wabhala: “Amamodeli anesisindo esivulekile angenawo amakhono ayingozi ayinzuzo yomphakathi.”