Umkhiqizi omakhalekhukhwini, amaphilisi kanye nobuchwepheshe I-Huawei isebenza kanzima ukufaka i-DeepSeek nama-chips ayo athuthukile. I-DeepSeek iwukubheja okusha kwamaShayina komkhakha wezobunhloli bokwenziwa, futhi iyimbangi ethakaselekayo ye-OpenAI kanye Gemini. Ngokusekelwa kanye nokufakwa okuzenzakalelayo kwamadivayisi we-Huawei, singase sikhuluma ngomunye wemifelandawonye yamaShayina enamandla kakhulu emkhakheni oqhudelana kakhulu wemakethe yezobuchwepheshe.
Izimakethe zamasheya zisanda kuzanyazanyiswa yintuthuko ye-DeepSeek kanti uHuawei ugxumela ku-bandwagon. Umthelela wale phrojekthi entsha ye-AI uguqula imakethe yaseShayina futhi usakazeka ngokushesha. Ihlanganiswe kanjani?
I-DeepSeek kaHuawei kanye nokunye ukuthuthuka ekuhlanganisweni kwe-AI
IHuawei ingenye yezinkampani ezinkulu zobuchwepheshe zaseChina, futhi kulandela isiphakamiso se-DeepSeek, isivele ibheke ukufaka i-AI kuma-chips ayo. Lona umzamo ohlanganyelwe wokuhlanganisa imodeli yomthombo ovulekile ezinhlelweni zayo zokusebenza namasevisi, ngaleyo ndlela kuzuzwe ukunwetshwa okukhulu kwemodeli ye-DeepSeek kubasebenzisi emhlabeni jikelele.
Ekuqaleni kuka February UHuawei umemezele ukuthi uzosebenzisa i-DeepSeek kumodeli yakho yehadiwe yekhompuyutha. Ama-Ascend processors azoba ngabokuqala ukufaka ukwesekwa kwezici namathuluzi axhunywe ku-DeepSeek. Lezi zindaba bezihlanganisa kakhulu emkhakheni we-AI njengoba manje amaphrosesa anamandla e-NVIDIA awadingeki ukuze imodeli ye-Artificial Intelligence isebenze. Lesi isinyathelo esisha esiya phambili salolu hlobo lwesiphakamiso.
Ukungqubuzana ne-United States
Ngokombiko wabahlaziyi bakwa- I-Berstein, inkampani ezinikele ekutshalweni kwezimali nokucwaningaIsinyathelo sikaHuawei siphikisana nezijeziso zase-US. Kubonisa ukuthi i-China inamandla okuletha ukusebenza kwe-AI okuncintisanayo emhlabeni jikelele, nge-software ne-hardware ethuthukiswe e-China futhi ngaphandle kwesidingo sokuthenga emakethe yaseNtshonalanga. Ama-chips e-NVIDIA azothathelwa indawo ngokuqondile umndeni wakwaHuawei we-Ascend.
Kusukela ekupheleni kuka-2022, abaphathi beBiden babeke imijikelezo ehlukahlukene yokulawula ukuthunyelwa kwe-Chinese. Inhloso yalezi zinyathelo kwakuwukuphuca izwe ubuchwepheshe iWashington eyesaba ukuthi iBeijing ingabusebenzisa ukwakha izikhali zempi kanye nezinhlelo ze-AI ezithuthukisiwe.
Nokho, impumelelo yemibhikisho ye-DeepSeek ngemodeli ye-R1 iphakamisa ukungabaza. Izindleko zokuqeqesha ziyingxenyana uma kuqhathaniswa nalokho okudingekayo kumodeli ye-ChatGPT. Impumelelo ye-DeepSeek iholele u-Huawei nezinye izikhondlakhondla zaseShayina ukuthi zibheje emsamo. Abakhiqizi abakhulu be-chip abanjengoMoore Threads, Enflame (Tencent), Kunlunxin (Baidu) kanye neHygon Information Technology sebememezele ukusekelwa kwenkampani. Ama-chips abo ekhompiyutha azothola ukwesekwa ekusebenziseni i-DeepSeek nokusebenzisa ngokugcwele lokho okuhlinzekwa yimodeli ye-AI.
I-Huawei, i-DeepSeek nokuthi yini elele ngale kwama-chips
Ngokungeziwe ekusekeleni i-DeepSeek ngokucubungula ama-chips asekela ukusebenza nemodeli, ezinye izindawo zobuchwepheshe nazo zizothuthukela ekukhiqizeni ukuhambisana. Ingalo yekhompyutha yamafu yezinkampani ezinkulu izophinde isekele umsebenzi ne-DeepSeek. Ngale ndlela, i-Alibaba, i-Bytedance, i-Tencent ne-Baidu yenza i-AI nemisebenzi yayo itholakale emphakathini wabasebenzisi.
I-Los opharetha bezingcingo abathathu abakhulu eChina kanye neLenovo kanye nefemu yezimoto i-Geely yamukela i-DeepSeek kumapulatifomu abo ahlukahlukene ahlakaniphile namafu. Ikusasa le-Artificial Intelligence kumdondoshiya wase-Asia lihamba ngokuqondile ekuhloleni nasekuthuthukisweni okuhlangene kwamathuluzi ne-DeepSeek. Futhi yonke into ikhomba ukuthi kunentshisekelo enkulu ekwenzeni iphrojekthi isebenze. Njengoba iwuhlelo lomthombo ovulekile, lwenza ukuzethemba okukhulu phakathi kwengxenye enkulu yomphakathi wamakhompyutha.
Iyini i-DeepSeek?
Lesi isiqalo esisekelwe e-Hangzhou esasungulwa ngo-2023 futhi ngenyanga edlule savusa isithakazelo somhlaba wonke lapho sembula imodeli yaso ye-Artificial Intelligence. Uhlelo lokusebenza lomsizi we-AI ngokushesha lwaba uhlelo lokusebenza olulandwe kakhulu emhlabeni jikelele, lwaze lwadlula i-ChatGPT ekupheleni kukaJanuwari.
A Ezinsukwini ezingu-20 nje ngemva kokwethulwa kwayo, abasebenzisi abasebenzayo nsuku zonke badlula izigidi ezingu-22 ngokwemithombo esemthethweni evela kwabezindaba zezwe laseChina. Ukuphumelela kwalolu hlelo kuphinde kwenza umsunguli walo, uLian Wenfeng, acishe abe yiqhawe lesizwe namuhla.
Yize kunethemba kulo mkhakha, abahlaziyi baphinde bakhombise izingqinamba ezithile uma kukhulunywa ngokuthuthukiswa kwe-chip ye-AI eChina. Isithiyo esikhulu yimikhawulo yokuthumela ngaphandle ebekwe uhulumeni wase-United States.
Izinzuzo nezinselelo zomkhakha
Ukuthi abathuthukisi be-DeepSeek bakwazile ukufaka amamodeli ekwakhiweni kwama-chip ahlukene kuwumsebenzi wesoftware osezingeni eliphansi. Futhi okuningi. Nakuba lokhu kuwumphumela omangalisayo, akuyixazululi inkinga yokushoda kwama-chip i-United States eyiphoqelela ezweni lase-Asia.
Ukwanda ku ukukhiqizwa kwe-chip okuthuthukisiwe uhlala eboshwe ukwesaba kanye nemikhawulo yase-US. Kodwa lokho akusona ukuphela kwesithiyo. I-Taiwan ne-Australia isanda kuvimbela ukusetshenziswa kwe-DeepSeek ngezizathu zokuthi idatha isengcupheni. Ngale ndlela, bafuna ukunciphisa amandla okukhula kwezobuchwepheshe anikezwa yiChina emphakathini womhlaba.
Ukuvinjelwa kuphinde kwabikwa kwezinye iminyango kahulumeni waseNingizimu Korea kanye neminyango. Ezombusazwe ziyangena emkhakheni wezobuchwepheshe kanye nomnotho njengoba izwe le-AI lithuthuka ukuze linikeze abasebenzisi izinketho ezintsha. UHuawei ubheke ukukhulisa intuthuko yasendaweni esifundeni, kanti i-DeepSeek iwuphawu lokujoyina impi kulo mkhakha.
Okwamanje, I-Gemini, i-ChatGPT namanye amathuluzi e-AI i-generative iyaqhubeka nokuguquguquka. Ngokuvamile, singachaza umkhakha ophumelelayo, onabadlali abavela ezizindeni ezihlukene abazama ukudala izethameli kanye nesevisi ezohlinzeka umphakathi ngamasu ahlukene, ububanzi kanye nemikhawulo. Kusazobonakala ukuthi inselelo iqhubeka kanjani.