
Ngokusho komyalezo oshicilelwe ku-akhawunti ye-WeChat yakwa-Zhiyuan AGIBOT, imiphumela yakamuva yokuhlolwa kohlu olugunyaziwe lwe-DailyOmni lokuqonda zonke izinhlobo zemidiya isanda kumenyezelwa. Imodeli enkulu ye-embodied intelligence yendabuko enokuqonda zonke izinhlobo zemidiya, i-WITA-Omni Preview eyakhiwe yi-Zhiyuan, ngenxa yamakhono ayo ahamba phambili okuqonda okuhlanganisiwe komsindo nevidiyo kanye nokucabanga ngokulandelana kwesikhathi, ithole amaphuzu aphelele angu-85.21 yaphuma phambili. Idlule amamodeli ahamba phambili asekhaya naphesheya, okuhlanganisa i-Qwen, i-Gemini, i-Doubao ne-NVIDIA, yaphinde yathola indawo yokuqala ezinkomba eziyisithupha kweziyisishiyagalombili ezinemininingwane.

Ngokwazi kwe-IT之家, i-DailyOmni iwuhlu olugunyaziwe nolwaziwa kabanzi embonini lokuhlola ukuhambisana kwesikhathi komsindo nevidiyo kanye nekhono lokucabanga okuhlanganisiwe kwezinhlobo ezahlukene zemidiya. Ngokungafani nemisebenzi evamile yokuhlola egxile ezithombeni ezinombhalo namavidiyo amafushane, lolu hlu lusebenzisa kakhulu izinto zomsindo nevidiyo ezivela ezindaweni zangempela ezivulekile. Luhlola ngokuyinhloko ikhono lem modeli lokuhlanganisa izithombe ezibonakalayo nolwazi lomsindo wendawo, ukubona ngokunembile ubudlelwano phakathi komsindo nesithombe, ukulandelana kwezenzakalo kanye nencazelo ehlanganisa izinhlobo ezahlukene zemidiya. Lokhu kuhambisana kakhulu namakhono ayisisekelo okubona adingwa amarobhothi anomzimba womuntu lapho enza ukuxhumana phakathi kwabantu nemishini ezindaweni zangempela zomzimba.

Ngokusho kwesitatimende esisemthethweni, imfihlo yokuhamba phambili kwe-WITA-Omni ukuthi ayigcini ngokufaka imodeli yengxoxo erobhothini, kodwa iphakamisa izenzo zomzimba nezinkulumo zobuso zibe yimiphumela eyisisekelo esezingeni elifanayo nenkulumo. Isebenzisa imodeli eyodwa yendabuko esebenza kusukela ekuqaleni kuya ekugcineni ukuze iqhube ukubona, ukwenza izinqumo nokuveza imizwa ngokuhlangene. Ngakho-ke, isakhiwo se-Thinker–Talker sinwetshwe saba isakhiwo se-Thinker–Talker–Actor esigxile emiphumeleni ye-embodied intelligence.
Ukuhamba phambili kwe-WITA-Omni akuncikile nje kubukhulu bemodeli, kodwa kuvela ohlelweni lwedatha olunezigaba nezindlela zokuqeqesha eziqondene nomsebenzi, olwakhiwe ngokukhethekile ukuxhumana kwe-embodied intelligence kuzo zonke izinhlobo zemidiya. Esigabeni sokuqeqeshwa okuphakathi, i-WITA-Omni isebenzisa idatha yemidiya eminingi evulekile neyayo engokwayo efinyelela ezigidini zamahora. Lokhu kuthuthukisa isisekelo esiqinile sombhalo nokubona sibe imodeli ekwazi “ukuzwa, ukuqonda nokwenza ukucabanga okuhlanganisiwe komsindo nesithombe”.
Ngaphezu kwalokho, idatha yomsindo nevidiyo yomphakathi ivame ukungabi nolwazi lokushintshana kwemijikelezo ekuxhumaneni kwangempela, isikhathi sokuphendula, izenzo kanye nezinkulumo zobuso, ngakho kuba nzima ukuyisebenzisa ngokuqondile ukuqeqesha imodeli yokuxhumana kwe-embodied intelligence esebenza kusukela ekuqaleni kuya ekugcineni. Ngenxa yalokho, i-Zhiyuan yakha iqoqo ledatha elikhulu, elisezingeni eliphezulu nelinemidiya yonke, elibheke kumuntu futhi elenzelwe izimo zokuxhumana zangempela. Leli qoqo ligcina ngokuphelele ubudlelwano bemvelo bokulandelana kwesikhathi phakathi komsindo, isithombe, ulimi, izenzo nezinkulumo zobuso.
