Ye-BMS, BUS, Industrial, Instrumentation Cable.
Njengoko uMthendeleko weNtwasahlobo ufikelela esiphelweni, uchulumanco olujikeleze i-DeepSeek luhlala luqinile. Iholide yakutshanje ibonise ingqiqo ebalulekileyo yokhuphiswano ngaphakathi kwishishini le-tech, kunye nabaninzi baxoxa kunye nokuhlalutya le "catfish." ISilicon Valley ijongene nengxaki engazange ibonwe ngaphambili: abameli bomthombo ovulekileyo bavakalisa izimvo zabo kwakhona, kwaye i-OpenAI iphinda iphonononge ukuba isicwangciso sayo somthombo ovaliweyo yayiyeyona nto ilungileyo. I-paradigm entsha yeendleko eziphantsi zokubala ibangele ukusabela kwekhonkco phakathi kwee-chip giants ezifana ne-Nvidia, ekhokelela ekurekhodeni ilahleko yexabiso lemarike yosuku olunye kwimbali ye-stock market yase-US, ngelixa ii-arhente zikarhulumente ziphanda ukuthotyelwa kwee-chips ezisetyenziswe yi-DeepSeek. Phakathi kophononongo oluxubeneyo lwe-DeepSeek phesheya kolwandle, ngaphakathi, ifumana ukukhula okumangalisayo. Emva kokusungulwa kwemodeli ye-R1, i-app ehambelana nayo ibone ukunyuka kwezithuthi, ebonisa ukuba ukukhula kumacandelo ezicelo kuya kuqhubela phambili i-AI ecosystem. Umba oncomekayo kukuba i-DeepSeek iya kwandisa amathuba okusebenza, icebisa ukuba ukuthembela kwi-ChatGPT akuyi kuba yindleko enkulu kwixesha elizayo. Olu tshintsho lubonakaliswe kwimisebenzi yamva nje ye-OpenAI, kubandakanywa ukunikezelwa kwemodeli yokuqiqa ebizwa ngokuba yi-o3-mini kubasebenzisi abakhululekileyo ngokuphendula i-DeepSeek R1, kunye nophuculo olulandelayo olwenze ikhonkco lokucinga le-o3-mini yoluntu. Abasebenzisi abaninzi baphesheya kweelwandle bavakalise umbulelo kwi-DeepSeek ngolu phuhliso, nangona le khonkco lokucinga lisebenza njengesishwankathelo.
Ngethemba, kuyabonakala ukuba i-DeepSeek idibanisa abadlali basekhaya. Ngokugxila ekunciphiseni iindleko zoqeqesho, abavelisi be-chip abahlukeneyo abanyukayo, ababoneleli belifu abaphakathi, kunye noqalo oluninzi bajoyina ngokusebenzayo inkqubo yendalo, ephucula ukusebenza kakuhle kweendleko zokusebenzisa imodeli ye-DeepSeek. Ngokwamaphepha e-DeepSeek, uqeqesho olupheleleyo lwemodeli ye-V3 lufuna kuphela iiyure ze-2.788 yezigidi ze-H800 GPU, kwaye inkqubo yoqeqesho izinzile kakhulu. Uyilo lwe-MoE (Umxube weeNgcali) lubalulekile ekunciphiseni iindleko zoqeqesho lwangaphambili nge-factory yeshumi xa kuthelekiswa ne-Llama 3 kunye ne-405 yeebhiliyoni zeeparamitha. Okwangoku, i-V3 yimodeli yokuqala eyamkelwa esidlangalaleni ebonisa ukungabikho okuphezulu kwi-MoE. Ukongeza, i-MLA (Ingqalelo yoMaleko oMninzi) isebenza ngokubambisana, ngakumbi kwimiba yokuqiqa. "I-sparser i-MoE, inkulu isayizi yebhetshi efunekayo ngexesha lokuqiqa ukuze kusetyenziswe ngokupheleleyo amandla okubala, kunye nobukhulu be-KVCache ibe yeyona nto ibalulekileyo yokukhawulela; i-MLA inciphisa kakhulu ubungakanani be-KVCache," watsho umphandi ovela kwi-Chuanjing Technology kuhlalutyo lwe-AI Technology Review. Ngokubanzi, impumelelo ye-DeepSeek ilele ekudityanisweni kweetekhnoloji ezahlukeneyo, hayi nje enye. Abangaphakathi kwishishini bancoma ubuchule bobunjineli beqela le-DeepSeek, beqaphela ukugqwesa kwabo kuqeqesho oluhambelanayo kunye nokulungelelaniswa kwabasebenzisi, ukufikelela kwiziphumo ezisisiseko ngokusulungekisa zonke iinkcukacha. Indlela ye-DeepSeek yomthombo ovulekileyo iqhubela phambili ifuthe kuphuhliso lulonke lweemodeli ezinkulu, kwaye kucingelwa ukuba ukuba iimodeli ezifanayo ziyanda zibe yimifanekiso, iividiyo, kunye nokunye, oku kuya kuvuselela imfuno kulo lonke ishishini.
Amathuba eeNkonzo zeQela lesithathu lokuQiqa
Idatha ibonisa ukuba ukususela ekukhululweni kwayo, i-DeepSeek iqokelele i-22.15 yezigidi zabasebenzisi abasebenzayo bemihla ngemihla (DAU) kwiintsuku nje ze-21, ukufikelela kwi-41.6% yesiseko somsebenzisi we-ChatGPT kwaye idlula i-16.95 yezigidi zabasebenzisi abasebenzayo bemihla ngemihla ye-Doubao, ngaloo ndlela ibe sisicelo esikhula ngokukhawuleza kwihlabathi jikelele, phezulu kwi-Apple App Store.5 Nangona kunjalo, ngelixa abasebenzisi bethontelana kwimihlambi, abahlaseli be-cyber bebesoloko behlasela i-DeepSeek app, bebangela uxinzelelo olukhulu kwiiseva zayo. Abahlalutyi boshishino bakholelwa ukuba oku kungenxa yokuba i-DeepSeek ithumela amakhadi oqeqesho ngelixa ingenawo amandla aneleyo okuqiqa. Umzi-mveliso wangaphakathi wazisa uPhononongo lweTekhnoloji ye-AI, "Imiba rhoqo yeseva inokusonjululwa ngokulula ngokubiza iifizi okanye ukuxhasa ngemali ukuthenga oomatshini abaninzi; ekugqibeleni, kuxhomekeke kwizigqibo ze-DeepSeek." Oku kubonisa ukurhweba ngokugqalisela kubuchwephesha ngokuchasene nokuveliswa kwemveliso. I-DeepSeek ixhomekeke ubukhulu becala kubungakanani bobungakanani bokuziphilisa, ifumene inkxaso-mali encinci yangaphandle, ekhokelela kuxinzelelo oluphantsi lokuhamba kwemali kunye nokusingqongileyo kwetekhnoloji esulungekileyo. Okwangoku, ngenxa yeengxaki ezikhankanywe ngasentla, abanye abasebenzisi babongoza i-DeepSeek kumajelo eendaba ezentlalo ukuba iphakamise imida yokusetyenziswa okanye yazise iimpawu ezihlawulelwayo ukomeleza intuthuzelo yabasebenzisi. Ukongeza, abaphuhlisi baqalisile ukusebenzisa i-API esemthethweni okanye i-APIs yomntu wesithathu ukuze kuphuculwe. Nangona kunjalo, iqonga elivulekileyo le-DeepSeek lisanda kubhengeza, "Izixhobo zangoku zeseva zinqabile, kwaye ukutshaja kwenkonzo ye-API kunqunyanyisiwe."
Oku ngokungathandabuzekiyo kuvula amathuba amaninzi kubathengisi beqela lesithathu kwicandelo leziseko ezingundoqo ze-AI. Kutshanje, uninzi lwezigebenga zamafu zasekhaya nakwamanye amazwe ziphehlelele imodeli ye-DeepSeek ye-APIs-izigebenga zaphesheya iMicrosoft kunye neAmazon zaziphakathi kwabokuqala ukujoyina ukuphela kukaJanuwari. Inkokeli yasekhaya, iHuawei Cloud, yenza inyathelo lokuqala, ikhupha i-DeepSeek R1 kunye ne-V3 iinkonzo zokuqiqa ngokubambisana neSilicon-based Flow ngoFebruwari 1. Iingxelo ezivela kwi-AI Technology Review zibonisa ukuba iinkonzo ze-Silicon-based Flow ziye zabona ukungena kwabasebenzisi, ngokufanelekileyo "ukubetha" iqonga. Iinkampani ezinkulu ezintathu zobuchwepheshe-i-BAT (i-Baidu, i-Alibaba, i-Tencent) kunye ne-ByteDance-zikwakhuphe izibonelelo zexabiso eliphantsi, ixesha elilinganiselweyo eliqala ngoFebruwari 3, likhumbuza iimfazwe zexabiso lomthengisi wefu lonyaka ophelileyo eliye latshiswa yi-DeepSeek's V2 model launch, apho i-DeepSeek yaqala ukubizwa ngokuba "yi-price butcher." Izenzo ezingxamisekileyo zabathengisi bamafu ziphinda ubudlelwane obuqinileyo phakathi kweMicrosoft Azure kunye ne-OpenAI, apho ngo-2019, iMicrosoft yenza utyalo-mali oluyibhiliyoni yeedola kwi-OpenAI kwaye yafumana inzuzo emva kokusungulwa kwe-ChatGPT ngo-2023. Nangona kunjalo, obu budlelwane busondeleyo buqale ukonakala emva kwe-Meta evulekileyo ye-Llama, ivumela abanye abathengisi be-ecosystem enkulu ngaphandle kweMicrosoft. Kulo mzekelo, i-DeepSeek ayigqithanga nje i-ChatGPT ngokubhekiselele kubushushu bemveliso kodwa iphinde yazisa iimodeli zemithombo evulekileyo emva kokukhululwa kwe-o1, efana nolonwabo olujikeleze ukuvuselelwa kukaLlama kwe-GPT-3.
Ngokwenyani, ababoneleli ngamafu bazibeka njengamasango etrafikhi kwizicelo ze-AI, okuthetha ukuba ubunzulu bobudlelwane nabaphuhlisi buguqulela kuncedo lokuqala. Iingxelo zibonisa ukuba i-Baidu Smart Cloud inabathengi abangaphezu kwe-15,000 abasebenzisa imodeli ye-DeepSeek ngeqonga leQianfan ngosuku lokuqaliswa kwemodeli. Ukongeza, iifemu ezininzi ezincinci zibonelela ngezisombululo, kubandakanya iSilicon-based Flow, Luchen Technology, Chuanjing Technology, kunye nababoneleli abahlukeneyo be-AI Infra abaye basungula inkxaso yeemodeli ze-DeepSeek. Uphononongo lweTekhnoloji ye-AI lufundile ukuba amathuba angoku okuphucula ukuthunyelwa kwe-DeepSeek akhoyo ngokuyintloko kwiindawo ezimbini: enye ilungiselela iimpawu ze-sparsity yemodeli ye-MoE isebenzisa indlela exubeneyo yokuqiqa ukusebenzisa imodeli ye-MoE yeebhiliyoni ezingama-671 ekuhlaleni ngelixa kusetyenziswa i-GPU edibeneyo ye-CPU. Ukongeza, ukwenziwa ngcono kwe-MLA kubalulekile. Nangona kunjalo, iimodeli ezimbini ze-DeepSeek zisajongene nemiceli mngeni ethile ekuhambiseni kakuhle. "Ngenxa yobukhulu bemodeli kunye neeparamitha ezininzi, ukwenziwa kwayo kuntsonkothile ngokwenene, ngakumbi kusetyenziso lwasekhaya apho ukufikelela kwibhalansi efanelekileyo phakathi kokusebenza kunye neendleko kuya kuba ngumngeni," utshilo umphandi ovela kwiChuanjing Technology. Owona mqobo ubalulekileyo usekoyiseni imida yamandla enkumbulo. "Sisebenzisa indlela ebanzi yokusebenzisana ukuze sisebenzise ngokupheleleyo ii-CPU kunye nezinye izixhobo zokubala, sibeka kuphela iindawo ezingabelwanga kwi-matrix ye-MoE kwi-CPU/DRAM ukuze kusetyenzwe kusetyenziswa abaqhubi be-CPU abasebenzayo, ngelixa iindawo ezixineneyo zihlala kwi-GPU," wachaza ngakumbi. Iingxelo zibonisa ukuba inkqubo-sikhokelo ye-Chuanjing yomthombo ovulekileyo i-KTransformers igalela ikakhulu amaqhinga ohlukeneyo nabasebenzi kufezekiso lweziGuquli zentsusa ngethemplethi, eyandisa ngokubonakalayo isantya sokuthelekelela usebenzisa iindlela ezifana neCUDAGraph. I-DeepSeek idale amathuba kwezi ziqalo, njengoko iinzuzo zokukhula zibonakala; iifemu ezininzi ziye zaxela ukukhula kwabathengi okuqaphelekayo emva kokusungula i-DeepSeek API, ukufumana imibuzo kubathengi bangaphambili abafuna ukulungiswa. Abangaphakathi kwi-industry baye baqaphela, "Kwixesha elidlulileyo, amaqela athile abaxhasiweyo aye ahlala evalelwe kwiinkonzo ezisemgangathweni zeenkampani ezinkulu, zibotshelelwe ngokuqinileyo kwiinzuzo zabo zeendleko ngenxa yesikali. Nangona kunjalo, emva kokugqiba ukuthunyelwa kwe-DeepSeek-R1 / V3 ngaphambi koMnyhadala we-Spring, ngequbuliso safumana izicelo zentsebenziswano ezivela kubathengi abaliqela abaziwayo, kwaye sade sazisa abathengi bangaphambili kwi-DeepSeek-R1 / V3. Okwangoku, kubonakala ngathi i-DeepSeek yenza ukuba imodeli yokusebenza ibonakale ibaluleke kakhulu, kwaye ngokwamkelwa ngokubanzi kweemodeli ezinkulu, oku kuya kuqhubeka nokuphembelela uphuhliso kwishishini le-AI Infra kakhulu. Ukuba imodeli yenqanaba le-DeepSeek inokubekwa kwindawo ngexabiso eliphantsi, iya kunceda kakhulu urhulumente kunye neenzame zenguqu yedijithali yeshishini. Nangona kunjalo, imiceli mngeni iyaqhubeka, njengoko abanye abathengi banokulindela okuphezulu malunga nemodeli enkulu yezakhono, oko bekwenza kucace ngakumbi ukuba ukulungelelanisa ukusebenza kunye neendleko kubaluleke kakhulu ekuhanjisweni okusebenzayo.
Ukuvavanya ukuba ingaba i-DeepSeek ingcono na kune-ChatGPT, kubalulekile ukuqonda iiyantlukwano zabo eziphambili, amandla, kunye neemeko zokusebenzisa. Nalu uthelekiso olubanzi:
| Uphawu/Umbono | DeepSeek | NcokolaGPT |
|---|---|---|
| Ubunini | Iphuhliswe yinkampani yaseTshayina | Iphuhliswe yi-OpenAI |
| Imodeli yoMthombo | Vula Umnikezi | Ubunikazi |
| Iindleko | Isimahla ukuyisebenzisa; iinketho zofikelelo API zitshiphu | Umrhumo okanye ukuhlawula-ngamaxabiso okusetyenziswa |
| Ukwenza ngokwezifiso | Yenziwe ngokwezifiso kakhulu, ivumela abasebenzisi ukuba batshintshe kwaye bakhe phezu kwayo | Ulungelelwaniso oluqingqiweyo lukhoyo |
| Ukusebenza kwiMisebenzi ethile | Ii-Excels kwiindawo ezithile ezifana nohlalutyo lwedatha kunye nokufunyanwa kolwazi | Ukuguquguquka nokusebenza olomeleleyo kubhalo loyilo kunye nemisebenzi yencoko |
| Inkxaso yoLwimi | Ugxininiso olomeleleyo kulwimi lwesiTshayina kunye nenkcubeko | Inkxaso yolwimi olubanzi kodwa i-US-centric |
| Iindleko zoQeqesho | Iindleko zoqeqesho eziphantsi, ezilungiselelwe ukusebenza kakuhle | Iindleko zoqeqesho eziphezulu, ezifuna izixhobo ezininzi zokubala |
| Ukwahluka kweempendulo | Inganikezela ngeempendulo ezahlukeneyo, mhlawumbi ziphenjelelwa yimeko yezopolitiko | Iimpendulo ezifanayo ezisekelwe kwidatha yoqeqesho |
| Abantu ekujoliswe kubo | Ijolise kubaphuhlisi kunye nabaphandi abafuna ukuguquguquka | Ijolise kubasebenzisi ngokubanzi abafuna izakhono zokuthetha |
| Sebenzisa Amatyala | Isebenza ngokufanelekileyo ukwenza ikhowudi kunye nemisebenzi ekhawulezayo | Ilungele ukuvelisa isicatshulwa, ukuphendula imibuzo, kunye nokubandakanyeka kwingxoxo |
Umbono obalulekileyo malunga "nokuphazamisa iNvidia"
Okwangoku, ngaphandle kweHuawei, abavelisi betshiphu abaninzi basekhaya abanje ngeMoore Threads, Muxi, Biran Technology, kunye noTianxu Zhixin nabo baziqhelanisa neemodeli ezimbini zeDeepSeek. Umenzi we-chip uxelele i-AI Technology Review, "Isakhiwo se-DeepSeek sibonisa ukutsha, kodwa sihlala si-LLM. Ukulungelelaniswa kwethu kwi-DeepSeek kugxininise kakhulu kwizicelo zokuqiqa, ukwenza ukuphunyezwa kobugcisa ngokufanelekileyo ngokuthe ngqo kwaye ngokukhawuleza." Nangona kunjalo, indlela ye-MoE ifuna iimfuno eziphezulu malunga nokugcinwa kunye nokusasazwa, kunye nokuqinisekisa ukuhambelana xa kuthunyelwa ngeetshiphusi zasekhaya, inikezela imingeni emininzi yobunjineli efuna ukusonjululwa ngexesha lohlengahlengiso. "Okwangoku, amandla okubambisa asekhaya awahambelani noNvidia ekusebenziseni nasekuzinziseni, efuna intatho-nxaxheba yasefektri yokuqala ukuseta isoftware, ukulungisa iingxaki, kunye nokusetyenzwa kwesiseko," ingcali kwishishini yatsho ngokusekwe kumava asebenzayo. Kwangaxeshanye, "Ngenxa yomlinganiselo omkhulu weparameter ye-DeepSeek R1, amandla okubala asekhaya afuna ii-node ezingakumbi zokudityaniswa. Ukongezelela, iinkcukacha ze-hardware zasekhaya zisengasemva ngandlel' ithile; umzekelo, iHuawei 910B okwangoku ayikwazi ukuxhasa i-FP8 inference eyenziwe yi-DeepSeek." Enye yeembalasane zemodeli ye-DeepSeek V3 kukwaziswa kwesikhokelo soqeqesho oluchanekileyo oluxutyiweyo lweFP8, oluthe lwaqinisekiswa ngokusebenzayo kwimodeli enkulu kakhulu, ephawula impumelelo ebalulekileyo. Ngaphambili, abadlali abaphambili abanjengoMicrosoft kunye noNvidia bacebise umsebenzi ohambelanayo, kodwa amathandabuzo ahlala ngaphakathi kumzi mveliso malunga nokuba nokwenzeka. Kuyaqondwa ukuba xa kuthelekiswa ne-INT8, eyona nto iphambili ye-FP8 yeyokuba ubungakanani basemva koqeqesho bunokufezekisa ukuchaneka okungalahlekiyo ngelixa kuphuculwa kakhulu isantya. Xa kuthelekiswa ne-FP16, i-FP8 inokuqonda ukuya kuthi ga kabini ukukhawulezisa kwi-Nvidia's H20 nangaphezulu kwe-1.5 yamaxesha okukhawulezayo kwi-H100. Ngokucacileyo, njengoko iingxoxo ezijikeleze umkhwa wamandla okubala yasekhaya kunye neemodeli zasekhaya zikhula ngokukhawuleza, uqikelelo malunga nokuba iNvidia ingaphazamiseka na, kwaye nokuba umsele weCUDA unokugqithwa, uya uxhaphaka. Inyani enye engenakuphikiswa kukuba i-DeepSeek ibangele ukuhla okukhulu kwixabiso lentengiso ye-Nvidia, kodwa olu tshintsho luphakamisa imibuzo malunga nengqibelelo yamandla e-Nvidia ephezulu. Amabali angaphambili awamkelweyo malunga nokuqokelela okuqhutywa yinkunzi ayacelwa umngeni, kodwa kusenzima ukuba uNvidia atshintshwe ngokupheleleyo kwiimeko zoqeqesho. Uhlalutyo lwe-DeepSeek yokusetyenziswa okunzulu kweCUDA lubonisa ukuba ukuguquguquka-okufana nokusebenzisa i-SM kunxibelelwano okanye ukukhohlisa ngokuthe ngqo amakhadi womnatha-akunakwenzeka ukuba ii-GPU eziqhelekileyo zihlalise. Iimbono zeshishini zigxininisa ukuba i-Nvidia's moat ibandakanya yonke i-ecosystem ye-CUDA kunokuba nje i-CUDA ngokwayo, kunye ne-PTX (i-Parallel Thread Execution) imiyalelo eqeshwa yi-DeepSeek iseyinxalenye ye-ecosystem ye-CUDA. "Ngexesha elifutshane, amandla e-computational kaNvidia akakwazi ukugqithiswa-oku kucacile ngokukodwa ekuqeqesheni; nangona kunjalo, ukuhambisa amakhadi asekhaya ngokuqiqa kuya kuba lula, ngoko ke inkqubela iya kukhawuleza ngokukhawuleza. Ukulungelelaniswa kwamakhadi asekhaya ngokuyinhloko kugxininise kwi-inference; akukho mntu sele ekwazi ukuqeqesha imodeli yentsebenzo ye-DeepSeek kumakhadi asekhaya ahlalutya i-AI yokuhlaziya i-Technology. " Ngokubanzi, ngokwembono, iimeko ziyakhuthaza kwiichips ezinkulu zasekhaya. Amathuba abavelisi beetshiphu zasekhaya ngaphakathi kwendawo yokuthelekelela abonakala ngakumbi ngenxa yeemfuno zoqeqesho eziphezulu kakhulu, ezithintela ukungena. Abahlalutyi bathi kwanele ukusebenzisa amakhadi angaphakathi; ukuba kuyimfuneko, ukufumana umatshini owongezelelweyo kunokwenzeka, ngelixa imodeli yoqeqesho ibeka imingeni eyodwa-ukulawula inani elongezelelweyo loomatshini linokuba ngumthwalo, kwaye amazinga aphezulu eempazamo anokuba nefuthe elibi kwiziphumo zoqeqesho. Uqeqesho lukwanazo iimfuno zesikali seqela, ngelixa iimfuno kumaqela okuthelekelela azikho ngqongqo, ngaloo ndlela zenza lula iimfuno ze-GPU. Okwangoku, ukusebenza kwekhadi le-Nvidia elilodwa le-H20 alidluli iHuawei okanye iCambrian; amandla ayo akukudibanisa. Ngokusekelwe kwimpembelelo jikelele kwimarike yamandla okubambisayo, umseki weLuchen Technology, Wena Yang, waphawula kudliwano-ndlebe kunye ne-AI Technology Review, "I-DeepSeek inokujongela phantsi okwethutyana ukusekwa kunye nokuqeshwa kwamaqela e-ultra-large yoqeqesho lwe-computational. Ekuhambeni kwexesha, ngokunciphisa kakhulu iindleko ezinxulumene noqeqesho olukhulu lwemodeli, ukuqiqa, kunye nezicelo, imfuno yokuthengisa iya kuba yinto enokwenzeka. qhubeleka nokuqhuba imfuno ezinzileyo kwimakethi yamandla ekubalwa." Ukongeza, "Imfuno ephezulu ye-DeepSeek yokuqiqa kunye neenkonzo zohlengahlengiso zihambelana ngakumbi nembonakalo yomhlaba yasekhaya, apho amandla engingqi abuthathaka, anceda ukuthomalalisa inkunkuma evela kwimithombo engasebenziyo emva kokusekwa kweqela; oku kudala amathuba asebenzayo kubavelisi kumanqanaba ahlukeneyo endalo ekholiweyo yasekhaya." I-Luchen Technology ibambisene neHuawei Cloud ukusungula i-DeepSeek R1 uthotho lwee-APIs zokuqiqa kunye neenkonzo zokucinga zamafu ezisekelwe kumandla asekhaya wokubala. Wena Yang uvakalise ithemba ngekamva: "I-DeepSeek ifaka ukuzithemba kwizisombululo eziveliswe ekhaya, ikhuthaza umdlandla omkhulu kunye notyalo-mali kwizakhono zokubala zasekhaya ukuya phambili."
Ukuqukumbela
Ingaba i-DeepSeek "ingcono" kune-ChatGPT ixhomekeke kwiimfuno ezithile kunye neenjongo zomsebenzisi. Kwimisebenzi efuna ukuguquguquka, ixabiso eliphantsi, kunye nokwenza ngokwezifiso, i-DeepSeek inokuba phezulu. Ukubhala ngokuyilayo, ukubuza ngokubanzi, kunye nonxibelelwano olusebenzisekayo lwencoko, i-ChatGPT inokukhokela. Isixhobo ngasinye sisebenza ngeenjongo ezahlukeneyo, ngoko ke ukhetho luya kuxhomekeka kakhulu kumxholo osetyenziswa kuwo.
Iintambo zokulawula
Inkqubo yeCabling eyakhiweyo
Inethiwekhi kunye neDatha, iFiber-Optic Cable, iPatch Cord, iiModyuli, iFaceplate
Epreli.16th-18th, 2024 Middle-East-Energy eDubai
Epreli.16-18th, 2024 Securika eMoscow
May.9th, 2024 IIMVELISO EZINTSHA KUNYE NETEKNOLOJI ZIQALISA UMNYATHO eShanghai
Oct.22nd-25th, 2024 UKHUSELEKO CHINA e Beijing
Nov.19-20, 2024 IHLABATHI ELIQHELEKILEYO KSA
Ixesha lokuposa: Feb-10-2025