USam Altman, oyi-CEO ye-OpenAI, uthe ukhathazekile ngokuthi ngelinye ilanga ubuhlakani bokwenziwa bungase bulawulwe yizinkampani, amamodeli noma abantu abambalwa, okushiye iningi labathengi lingenalo izwi ekutheni ubuchwepheshe buzowakha kanjani umhlaba.
“Into efanele ukuthi sifuna abantu bakwazi ukulawula ikusasa ngokujulile,” kusho u-Altman ohlelweni lwe-podcast lukaDavid Senra, olushicilelwe ngeSonto ngesikhathi sendawo, “umphakathi namamodeli bazoguquka ndawonye.”
U-Altman uthe ukhathazekile nangokuthi ubuhlakani bokwenziwa bungase buphunyuke emandleni abantu ababulawulayo. Okuphikisanayo ukuthi, u-Altman uthi ukwesaba lokhu kungenza ukuthi lokhu okunye kugcine kwenzeka.
“Baningi abantu abakhathazeke kakhulu ngobukhulu balezi zingozi, abalawulwa yile mizwa baze babone sengathi kufanele bavikele umhlaba,” kusho yena, “ngokwesibonelo, uyazi, ‘kufanele sinikele ngenkululeko eningi ukuze sithole ukuphepha.’”
Kungenzeka ukuthi u-Altman wayecabanga ngoDario Amodei, oyi-CEO ye-Anthropic.
U-Amodei uke waxwayisa ngezingozi ezilethwa ukuthuthuka okusheshayo kobuhlakani bokwenziwa kuma-podcast, ezihlokweni nakumabhulogi enkampani. Esihlokweni sangoJuni, u-Amodei wabhala ukuthi imodeli entsha yenkampani, i-Mythos, iyingozi “yangempela kakhulu” kwezokuphepha ku-inthanethi, emkhakheni wezezimali, engqalasizinda ebalulekile kanye nasekuphepheni kwezwe.
U-Amodei usebenzise lezi zixwayiso ukweseka ukuthi uhulumeni adlale indima enkulu ekulawuleni ubuhlakani bokwenziwa. Ngokombono ka-Altman, lokhu kungaba ukushintsha inkululeko kube ukuphepha.
I-Anthropic futhi iyona kuphela inkampani enkulu yobuhlakani bokwenziwa enyangeni edlule engazange isayine incwadi ehlanganyelwe eyeseka amamodeli anesisindo esivulekile. Amamodeli anesisindo esivulekile avumela onjiniyela ukuthi babelane, badawunilode, baguqule futhi basebenzise amamodeli obuhlakani bokwenziwa engqalasizinda yabo.
I-China isivele yamukele amamodeli anesisindo esivulekile, okuyenza ikwazi ukulandela i-Silicon Valley ngezindleko eziphansi kakhulu. Ezinye izinkampani zaseMelika, okuhlanganisa i-Nvidia, i-Meta ne-Microsoft, zikholelwa ukuthi iMelika nayo idinga ukwenza lokhu, kungenjalo ingase inikele ngezikhundla zayo emncintiswaneni wobuhlakani bokwenziwa kubancintisana nayo.
Nakuba i-OpenAI ingazange ikhiphe ikhodi yomthombo yamamodeli ayo aphambili, ngonyaka odlule yakhipha amamodeli amabili anesisindo esivulekile: gpt-oss-120b kanye ne-gpt-oss-20b.
Kuleyo podcast, u-Altman uphinde wathi abaholi baseMelika bezobuchwepheshe bobuhlakani bokwenziwa behlulekile ukukhombisa umphakathi ukuthi lobu buchwepheshe bungabanikeza kanjani abantu ukuzimela okukhulu ezimpilweni zabo.
“Sizobona igagasi elikhulu kunawo wonke lokusungulwa kwamabhizinisi amancane emlandweni,” kusho yena, “ngicabanga ukuthi ubuhlakani bokwenziwa yibo obuqhuba lokhu.”