Riaan IMicrosoft yenze isiqiniseko sokuthi ifaka i-Azure ku-AI-fest obekuyingqungquthela yonjiniyela Yakha 2023 kuleli sonto.
Njengoba amabhizinisi acabangela ukuhlola noma ukusebenzisa i-AI ekhiqizayo, angase abheke kumafu omphakathi kanye nengqalasizinda yekhompuyutha ekalayo efanayo kanye nengqalasizinda yesitoreji ukuze aqhube izinto ezifana namamodeli olimi olukhulu (LLMs).
Microsoft, ehlome nge I-ChatGPT, GPT-4, nezinye izinhlelo ze-OpenAI, sekunezinyanga icindezela amakhono e-AI kuwo wonke ama-nook kanye ne-cranny yombuso wayo. I-Azure ayihlukile – Isevisi ye-OpenAI iyisibonelo – nangemva kwayo Yakha inkomfaIfu lomphakathi le-Redmond manje selinokunikezwa okwengeziwe okufuniwe.
Okuphezulu ohlwini kunwetshiwe ukusebenzisana no-Nvidia, yona ngokwayo ejaha ukuzibonakalisa njengomhlinzeki wobuchwepheshe we-AI obalulekile, kusukela kuzisheshisi ze-GPU kuye kwisofthiwe. Kuleli sonto lilodwa umenzi we-chip wethu wethule inqwaba yobambiswano, njengoDell eDell Technologies World kanye nabakhi bamakhompiyutha amakhulu kwa-ISC23.
Ukuletha izinsiza ze-Nvidia ku-Azure
Ngokukhethekile, iMicrosoft ihlanganisa isoftware ye-Nvidia’s AI Enterprise suite, amathuluzi okuthuthukisa, izinhlaka, namamodeli aqeqeshwe kusengaphambili ku-Azure Machine Learning, idale lokho uTina Manghnani, umphathi womkhiqizo weplathifomu yamafu yokufunda umshini, akubiza ngokuthi “ibhizinisi lokuqala elilungele, elivikelekile, ukuphela. -ukuze kuphele inkundla yefu ukuze abathuthukisi bakhe, basebenzise, futhi baphathe izinhlelo zokusebenza ze-AI kufaka phakathi amamodeli wolimi amakhulu ngokwezifiso.”
Ngalo lolo suku, iMicrosoft yenza i-Azure Machine Learning amarejista – inkundla yokusingatha kanye nokwabelana ngamabhulokhi wokwakha wokufunda ngomshini njengeziqukathi, amamodeli nedatha kanye nethuluzi lokuhlanganisa i-AI Enterprise ku-Azure – ngokuvamile etholakalayo. I-AI Enterprise ku-Azure Machine Learning iyatholakala futhi ngokubuka kuqala okulinganiselwe kobuchwepheshe.
“Lokhu kusho ukuthi kumakhasimende anobudlelwano obukhona nobudlelwano ne-Azure, angasebenzisa lobo budlelwano – angasebenzisa izinkontileka zamafu asenazo – ukuthola iNvidia AI Enterprise futhi ayisebenzise ngaphakathi kwe-Azure ML ukuthola lokhu. isipiliyoni sebhizinisi elingenamthungo noma ngokwehlukana ezimeni abakhetha ukuzenza,” uManuvir Das, iphini likamongameli we-entreprise computing eNvidia, etshela izintatheli ezinsukwini ezimbalwa ngaphambi kokuvulwa kweBuild.
Ukuhlukanisa amanethiwekhi ukuze kuvikelwe idatha ye-AI
Amabhizinisi asebenzisa imisebenzi ye-AI efwini afuna ukuqinisekisa ukuthi idatha yawo ayivezwa kwezinye izinkampani, ukuhlukaniswa kwenethiwekhi kuyithuluzi elibalulekile. I-Microsoft inezici ezifana nendawo yokusebenza yesixhumanisi esiyimfihlo nokuvikelwa kokukhishwa kwedatha, kodwa ayikho inketho ye-IP yomphakathi yokubala izinsiza zezinkampani eziqeqesha amamodeli e-AI. KwaBuild, umthengisi umemezele ukuhlukaniswa kwenethiwekhi okuphethwe ku-Azure Machine Learning yokukhetha imodi yokuzihlukanisa elingana kangcono nezinqubomgomo zokuphepha zebhizinisi.
Ungaphuthelwa ukufakwa kwethu kwe-Build 2023
Ngokungamangalisi, amathuluzi omthombo ovulekile aya ngokuya eza esikhaleni se-AI. IMicrosoft ngonyaka odlule ibambisene neHugging Face ukuletha amaphoyinti wokufunda we-Azure Machine anikwa amandla ubuchwepheshe benkampani evula umthombo. KwaBuild, izinhlangano ezimbili kunwetshiwe ubudlelwano babo.
Ubuso Obukwangayo vele bunikeza a isethi ekhethiwe yamathuluzi nama-API kanye ne-a ihabhu elikhulu yamamodeli we-ML ongawalanda futhi uwasebenzise. Manje iqoqo lezinkulungwane zalawa mamodeli lizovela kukhathalogi yokufunda komshini we-Redmond’s Azure ukuze amakhasimende akwazi ukufinyelela futhi awasebenzise ezindaweni eziphethwe efwini le-Microsoft.
Izinketho eziningi zemodeli yesisekelo
I-Redmond nayo iyayenza amamodeli esisekelo ku-Azure Machine Learning etholakala ekubukeni kuqala komphakathi. Amamodeli esisekelo angamamodeli aqeqeshwe kusengaphambili anamandla futhi anekhono kakhulu izinhlangano ezingawenza ngokwezifiso ngedatha yazo ngokwezinjongo zazo futhi zikhiphe njengoba kudingeka.
Amamodeli esisekelo sezibaluleke kakhulu, njengoba zingasiza izinhlangano ukuthi zakhe izinhlelo zokusebenza ezinamandla ze-ML ezingasho lutho, ezilolongelwe ngokwezidingo zazo ezithile, ngaphandle kokuthi zichithe amakhulu ezigidi zamaRandi ziqeqesha amamodeli kusukela ekuqaleni noma ekulayisheni ukulayisha kanye nedatha yamakhasimende ebucayi iye efwini.
UNvidia ukhiphe a I-NeMo uhlaka olungase lube usizo kule ndawo, futhi le nyanga ine ebambisene nge-ServiceNow futhi – kuleli sonto – u-Dell in Iphrojekthi ye-Helix kuleyo migqa.
“Njengoba sisebenze nezinkampani zamabhizinisi ku-generative AI ezinyangeni ezimbalwa ezedlule, esikufundile ukuthi kunenqwaba yezinkampani ezingathanda ukusebenzisa amandla e-AI ekhiqizayo, kodwa zikwenze kuma-datacenters azo. noma ukwenze ngaphandle kwefu lomphakathi, ”kusho uNvidia’s Das.
Izinsiza ezifana nomthombo ovulekile kanye namamodeli esisekelo zithembisa ukunciphisa ubunkimbinkimbi nezindleko ukuze kuvunyelwe izinhlangano eziningi ukufinyelela ku-AI ekhiqizayo. ®
