Kwenzekeni
I-South China Morning Post ibika ukuthi i-DeepSeek ifuna onjiniyela abasezingeni eliphezulu abangaba ngu-150 kulokho inkampani ekuchaza ngokuthi ukuqasha “okungakaze kubonwe ngaphambili”. Ukuqashwa kuhloselwe ukuthuthukisa, ukugcina kanye nokubhala kabusha ingqalasizinda ye-backend njengoba umsebenzi wabasebenzisi, umthwalo wekhompiyutha kanye nobunzima besistimu bukhula.
I-South China Morning Post ibika ukuthi i-DeepSeek iqasha onjiniyela abasezingeni eliphezulu abangaba ngu-150 ukuze balungise ingqalasizinda inkampani ethi isisondela emikhawulweni yayo. UCui Tianyi, uchwepheshe wangaphambili wezohwebo owajoyina i-DeepSeek ngoMashi ukuze ahole ithimba layo le-Harness, wabhala ku-X ukuthi ukuqashwa kwabangelwa ukwanda okusheshayo kwamavolumu edatha, izinombolo zemishini, umthwalo wokuqeqeshwa kanye nabasebenzisi abasebenzayo.
Ngokombiko, uCui uthe isilinganiso esikhulu sidala inkimbinkimbi enkulu kakhulu yekhompiyutha nokuthi izinhlelo ezingemuva ze-DeepSeek kumele zithuthukiswe kakhulu, zigcinwe futhi zibhalwe kabusha ukuze zisekele izikhombisi-ndlela ezintsha zobuchwepheshe. Umbiko uphinde uphawule ukuthi i-DeepSeek yamemezela ngoJuni ukuthi ihlele okungenani ukuphinda kabili ubukhulu bomnyango ngamunye futhi isivule izindima ezingama-33 kulo lonke ucwaningo, ubunjiniyela kanye nokuphathwa kwemikhiqizo.
Isibalo sokuqashwa kanye nezizathu zengqalasizinda kubikwa yiSouth China Morning Post evela engosini yezokuxhumana kaCui. Ukuqinisekiswa okubanzi kwe-DeepSeek uqobo, ububanzi obunembayo bokubhala kabusha, kanye nomugqa wesikhathi womsebenzi akuqinisekiswanga ngokuzimela emthonjeni onikeziwe. Akukho ukufinyelela komkhiqizo, intengo noma ushintsho lokutholakala olubikiwe.
Imininingwane yomthombo: scmp.com ↗
Kungani kubalulekile
Ukuqashwa kwe-DeepSeek okubikwayo kusikisela ukuthi ukukala isevisi ye-AI kudinga ukutshalwa kwezimali okukhulu ngale kocwaningo lwemodeli. Uma kunembile, ukunyakaza kungase kuthinte ukwethembeka namandla enkampani ye-AI enesisindo esivulekile esetshenziswa kabanzi njengoba iphatha abasebenzisi abaningi kanye nemithwalo yemisebenzi enzima yokubala. Umbiko awuqinisekisi ngokuzimela isimo sengqalasizinda ye-DeepSeek, ukukhula komsebenzisi noma izinhlelo zesevisi yezentengiso.
Intuthuko ibalulekile ngoba umthamo we-backend uyisithiyo esisebenzayo kumasevisi e-AI: abasebenzisi abaningi, umsebenzi wokuqeqesha kanye nemithwalo yemisebenzi yama-ejenti yekhompyutha ingakhuphula inkimbinkimbi yokusebenza ngisho nalapho imodeli ngokwayo ingakashintshi. Ukugxila okubikwe kwe-DeepSeek konjiniyela bengqalasizinda abakhulu kubonisa ukuthi ukukala isevisi yayo kungase kudinge umzamo omkhulu wobunjiniyela eduze kokuthuthukiswa kwemodeli.
I-South China Morning Post iveza ukuqashwa njengobufakazi bokuvuthwa kwamasu kusukela kusiphazamisi esigxile ocwaningweni ukuya endaweni yevolumu ephezulu. Lokho kuhumusha kuwukuhlaziywa kwe-outlet, hhayi iqiniso elizimele. Umthombo awuhlinzeki ngedatha yesikhathi esengeziwe, izinombolo zabasebenzisi, izivivinyo zokusebenza, izibalo zemali engenayo noma ubufakazi bokuthi izinsizakalo zezentengiso zezinga lebhizinisi ziqalisiwe.
I-Interactive Mechanism: Indlela Esebenza Ngayo Ngempela
Hlola ubuchwepheshe obuyisisekelo ngemuva kwalokhu kuthuthukiswa ngokuhlanganyela.
crm_get_transaction(id='4092').Which component of an AI application is the machine-learning model itself?
Ongakubuka ngokulandelayo
Buka ukuze uthole ukuqinisekiswa okuvela ku-DeepSeek, imininingwane mayelana nezindima nezindawo, kanye nobufakazi bokuthi ukulungiswa kwengqalasizinda kushintsha ukwethembeka kwesevisi, umthamo noma ukutholakala. Umbiko ukhomba ukuthi akukho ukwethulwa komkhiqizo omusha, ukushintshwa kokufinyelela noma ulwazi lwentengo, ngakho abasebenzisi akufanele bacabangele ukutholakala okubanzi noma imigomo yokuthengisa kusukela esimemezelweni sokuqasha.
Ukuqinisekisa okuwusizo kakhulu kungaba yikhasi lokuqasha le-DeepSeek, isitatimende senkampani noma umbiko olandelayo okhomba izikhundla eziyi-150, izibopho zabo kanye nezinhlelo ezakhiwa kabusha.
Abasebenzisi nonjiniyela kufanele babheke izinguquko ezibhalwe phansi ekuzinzeni kwesevisi, ukusebenza, imikhawulo yesilinganiso, ukufinyelela kwe-API noma umthamo wokunikeza imodeli. Alukho kulezi zinguquko olushiwo embikweni onikeziwe.
Uhlelo lokwandisa lwangoJuni lungase lubonise ukukhula kwenhlangano okuqhubekayo, kodwa umthombo awuqinisekisi ukuthi ukuqashwa kwamanje kwe-backend kuyingxenye yalolo hlelo noma impendulo ehlukile kuzithiyo ezisanda kukhonjwa.