🧠 I-ByteDance yaseShayina ikhipha imodeli ye-Doubao 2.0 AI 'yesikhathi se-agent' ↗
I-ByteDance iqalise i-Doubao 2.0 njengendlela yokushintsha kusuka ku-"chat that answer" kuya ku-"AI that does stuff" - imisebenzi enezinyathelo eziningi, ukucabanga okuqinile, kanye nemisebenzi eminingi yokusekela ama-agent (njengoba wonke umuntu manje esakha ama-agent).
Baphinde bathi: ikhono elifanayo namamodeli aphezulu, kodwa kushibhile ukuwasebenzisa. Leyo ndlela yokubiza iyindlela… noma insimbi yokuqala yokulwa kokudla okulinganiselayo.
🎆 Amamodeli e-AI aseShayina agubha uMkhosi WaseNtwasahlobo ngonyaka ngemuva kokushaqeka kwe-DeepSeek ↗
Indawo ye-AI yaseShayina yenza lokho "okukhulu kokwethulwa kwamaholide" - izibuyekezo eziningi zamamodeli, ukuphishekela okuningi komfutho, kanye nomuzwa ocacile wokuthi akekho ofuna ukumangala futhi.
Ingxubevange ibanzi: ama-chatbot, ukuthuthukiswa kwesikhathi eside, amamodeli acindezelwe asebenziseka kalula kumaselula, ukukhishwa kwemithombo evulekile, kanye nokukhiqizwa kwamavidiyo okudonsela ukunaka ngaphandle kweShayina. Kufana nombukiso weziqhumane lapho yonke irokhethi imemeza khona ithi “mina olandelayo, mina olandelayo.”
🪙 Ukwesaba Amabhamuza E-AI Kudala Izinto Ezintsha Ezivela Kuwo ↗
Abatshalizimali bezikweletu bayaxakeka ngokuthi bangaboleka malini abadlali bezobuchwepheshe abakhulu ukuze baxhase umncintiswano wezikhali ze-AI - ngakho-ke, ngokwemvelo, imakethe isungula izindlela ezintsha zokuvimbela lokho kukhathazeka.
Kuhle kakhulu ukuthi “uma kukhona ukwesaba, othile uzokuqinisekisa,” okuyinto ehlakaniphile futhi eqalekisiwe kancane… njengokumboza amafu esiphepho bese uwathengisa ngokubhalisa.
🪖 Umbiko uthi amasosha ase-US asebenzise imodeli ye-Anthropic's AI Claude ekuhlaselweni kweVenezuela ↗
Umbiko uthi uClaude wasetshenziswa emsebenzini oyimfihlo waseVenezuela ngesiteshi sokubambisana - esikhipha izimpikiswano "zenqubomgomo yemodeli" elabhorethri zaya enkundleni yokusebenza, ngokushesha.
Ngisho noma imininingwane igcina incane kunezihloko (ngokuvamile injalo), iphuzu elikhulu liyanamathela: uma amamodeli esexhumene nemisebenzi yokuzivikela, "ubani olawula lokho" uyaxaka - hhayi ngendlela enhle, eqalayo.
🇮🇳 🧩 Inkampani ithi i-CEO ye-Nvidia uHuang ngeke aye engqungqutheleni ye-AI yase-India ngesonto elizayo ↗
UNvidia uthi uJensen Huang ngeke aye eNgqungqutheleni ye-AI Impact yaseNdiya ngenxa “yezimo ezingalindelekile,” kanti ithimba eliphezulu lizoya esikhundleni salokho. Lokho kuyishintsho eliphawulekayo, ngoba ukuba khona kwakhe empeleni kwakuyinto ekhangayo.
Umcimbi usabonakala uhlelekile, impela - kodwa ukukhanselwa okudumile kuhlala kushintsha izinga lokushisa ekamelweni kancane, noma ngabe wonke umuntu wenza sengathi akunjalo.
Imibuzo Evame Ukubuzwa
I-Doubao 2.0 ka-ByteDance kanye noshintsho "lwesikhathi sama-agent"
I-Doubao 2.0 ihlelwe njenge-pivot kusukela ku-"chat that answer" kuya ku-"AI that does stuff," kugcizelelwa imisebenzi enezinyathelo eziningi kanye nemisebenzi eminingi ehambisana ne-agent. Empeleni, ivezwa njengenamandla ekuhleleni, ekucabangeni ngokulandelana, nasekuthwaleni umsebenzi ezigabeni eziningana kunokunikeza impendulo eyodwa. Ilebula elithi "inkathi ye-agent" libonisa iphethini ebanzi: amaqembu amaningi akha izinhlelo ezisebenzayo, hhayi nje ukukhuluma.
Indlela "ukusebenza okusekelwe ku-agent" okubonakala ngayo ezimweni zokusetshenziswa kwe-Doubao 2.0
Ezindleleni eziningi, ukusethwa kwesitayela se-ejenti kuhlukanisa umgomo ube yizinyathelo, kuhlole imiphumela ephakathi, futhi kuphinde kuphele. Lokho kuvame ukusho ukubhala uhlelo, ukukhiqiza imisebenzi emincane, nokukhiqiza umphumela wokugcina ohlelekile kakhulu kunempendulo yesikhathi esisodwa. Enye indlela evamile yokuhlola lokhu ukwenza imisebenzi edinga izinqumo eziningi - isibonelo, ukubhala uhlu lokuhlola, ukulucwenga, bese uluguqula lube yinto ehlanzekile engalethwa.
Indlela yokuhlola ukuthi i-Doubao 2.0 ifana “namamodeli aphezulu” ngentengo ephansi
Kuphathe njengokuhlola umkhiqizo, hhayi isihloko. Qhathanisa i-Doubao 2.0 nemisebenzi ethile oyikhathalelayo - ukunemba, ukuhambisana, izindlela zokwehluleka, nokuthi kuthatha ukuzama okungakanani - kanye nezindleko zokubambezeleka kanye nokusetshenziswa. "Okushibhile" kungaguquka uma ugcina udinga izikhuthazo ezengeziwe, izithiyo ezisindayo, noma ukubuyekezwa okwengeziwe komuntu. Ekulinganiseni, qaphela izivivinyo ezikhethwe ngendlela ekhangayo ezingafani nemisebenzi yangempela yokusebenza.
Kungani amamodeli e-AI aseShayina eqala ngamandla kangaka ngesikhathi seNtwasahlobo Yomkhosi
Umbiko uchaza lokhu “njengokuqoqwa kokwethulwa kwemidlalo”: izibuyekezo eziqoqwe ngesikhathi samaholide amakhulu, okudala ukwanda okukhulu komfutho. Kuphinde kulandele ingcindezi yokuncintisana - akekho ofuna ukubhajwa engalindele ngemuva “kokwethuka” kwangaphambilini endaweni. Umphumela uba yinhlanganisela yeziqhumane zokwethulwa kwemidlalo kanye nokuthuthukiswa kwayo, namaqembu azama ukuheha ukunaka ngokushesha.
Yiziphi izinguquko ezishintshayo ezihambisana nomongo omude, amamodeli eselula acindezelwe, kanye nokukhishwa komthombo ovulekile kwabakhi
Ukuthuthukiswa komongo omude ngokuvamile kuhlose ukusingatha umbhalo noma umlando omningi kuseshini eyodwa, okungathuthukisa ukuhlaziywa noma ukuqhubeka kokufakwayo okukhulu. Amamodeli acindezelwe avame ukuphathelene nokwenza i-AI ibe usizo kakhulu kumadivayisi anesabelomali sokubala esiqinile. Ukukhishwa komthombo ovulekile kunganciphisa izindleko zokuhlola futhi kwandise ukwamukelwa, kodwa futhi kudlulisela imithwalo yemfanelo yokuthunyelwa, ukulawula ukuphepha, kanye nokugcinwa kumsebenzisi.
Lokho okushiwo yi-"AI bubble" hedges kanye nokusetshenziswa kwe-AI yezempi okubikiwe ekubuseni nasekuhleleni ubungozi
I-engeli ye-derivatives ikhomba ukukhathazeka ngokwezezimali mayelana nokuthi ingakanani isikweletu esingase sixhase umjaho wezikhali ze-AI, okuphoqa izimakethe ukuthi zisungule uthango lwalokho kungaqiniseki. Ngokwehlukile, ukusetshenziswa kwezempi okubikiwe kwemodeli efana noClaude kugcizelela ukuthi izingxoxo "zenqubomgomo yemodeli" zingaphenduka ngokushesha kangakanani zibe yimibuzo yokusebenza. Kubaholi, umphumela osebenzayo ukuphatha okunamandla: imikhawulo yokusetshenziswa ecacile, ukulawulwa kwabathengisi nabalingani, izindlela zokuhlola, kanye nezindlela zokukhulisa lapho ukuthunyelwa kwezingozi ezinkulu kungena esithombeni.