I-AWS Ihlose Ukuthuthukisa Ibhizinisi Le-AI Nge-Cerebras Chip Deal I-Amazon Web Services (AWS) yenza umnyakazo wamasu wokukhulisa amandla ayo obuhlakani bokwenziwa ngokubambisana okusha ne-AI chip startup Cerebras Systems. Lokhu kubambisana kuhlose ukukhulisa kakhulu ukusebenza kwezinhlelo zokusebenza ze-AI eziyinkimbinkimbi ezisebenza emafini. Isevisi ezayo, ezokwethulwa phakathi nezinyanga, izohlanganisa i-Hardware eyingqayizivele ye-Cerebras endaweni ye-AWS's Bedrock yokwakha isicelo, okuphawula umzuzu obalulekile emjahweni wengqalasizinda wamafu we-AI. Lesi sivumelwano sigcizelela ukuncintisana okukhulu phakathi kwabahlinzeki bamafu ukuze banikeze ukuqeqeshwa kwe-AI okunamandla kakhulu futhi okusebenzayo nezindawo zokuthintwa. Ngokusebenzisa ubuchwepheshe be-Cerebras obusha be-wafer-scale chip, i-AWS izimisa ukuze ihehe amabhizinisi nonjiniyela abasebenza esizukulwaneni esilandelayo samamodeli ezilimi ezinkulu kanye ne-AI ekhiqizayo.
Ukuqopha i-AWS kanye ne-Cerebras Partnership Umnyombo walesi simemezelo yisevisi entsha yamafu enikwa amandla i-Hardware ye-Cerebras ekhethekile. Lesi akusona nje esinye isibonelo somshini obonakalayo; imele isixazululo esididiyelwe ngokujulile esiklanyelwe kusukela phansi kuye phezulu ukuze kube nemithwalo yemisebenzi ye-AI yesikali esikhulu. Le nkonzo izotholakala ngqo nge-Amazon Bedrock, insizakalo ephethwe yinkampani yokwakha izinhlelo zokusebenza ezikhiqizayo ze-AI. Lokhu kuhlanganiswa kusho ukuthi abathuthukisi bangakwazi ukufinyelela amandla ekhompyutha e-Cerebras ngaphandle kokuphatha ingqalasizinda eyisisekelo. Kwenza kube lula inqubo yokuqeqesha amamodeli asezingeni eliphezulu, okungase kunciphise isikhathi sokuya emakethe ngemikhiqizo namasevisi anamandla e-AI. Ubambiswano lubonisa ukuzibophezela kwe-AWS ekunikezeni ukukhetha nokusebenza okuphambili ngale kwe-silicon yayo yasendlini, njenge-Trainium ne-Inferentia.
Kungani Cerebras? I-Wafer-Scale Engine Engine Advantage I-Cerebras Systems izihlukanise emakethe ye-chip ye-AI ngefilosofi yayo yokuklama eqinile. Esikhundleni sokusebenzisa ama-chips amaningi amancane, i-Cerebras yakha iphrosesa eyodwa, enkulu elingana ne-silicon wafer yonke. Le Wafer-Scale Engine (WSE) iyi-chip enkulu kunazo zonke ezake zenziwa futhi ihlukile ngokwezakhiwo. Izinzuzo ezibalulekile zale ndlela yezicelo zebhizinisi le-AI zijulile:
I-Massive Memory On-Chip: I-WSE ifaka inani elikhulu lememori enesivinini esikhulu ngqo kuphrosesa. Lokhu kubalulekile ekuqeqesheni amamodeli amakhulu olimi, njengoba kunciphisa ukunyakaza kwedatha okunensayo phakathi kwe-chip nememori yangaphandle. Umkhawulokudonsa ongakaze ubonwe: Ukuxhumana phakathi kwama-cores ku-wafer eyodwa kuyashesha kakhulu kunaphakathi kwama-chips ahlukene axhunywe kunethiwekhi. Lokhu kuqeda ibhodlela elikhulu ekuqeqeshweni kwe-AI esabalalisiwe. Ukuhlela Okwenziwa Lula: Onjiniyela bangahlela le divayisi enkulu njengesistimu eyodwa, ehlangene esikhundleni seqoqo eliyinkimbinkimbi lezinkulungwane zama-GPU, okwenza kube lula ukuthuthukiswa kwemodeli nokusetshenziswa.
Umthelela Ekusebenzeni Kwamafu AI Nokufinyelela Isithembiso esiyinhloko sale sevisi ye-AWS-Cerebras siwukuthuthuka okumangalisayo ekusebenzeni kokuqeqeshwa nokusebenzisa amamodeli e-AI. Kumabhizinisi, lokhu kuhumusha imijikelezo yokuqamba esheshayo kanye nezindleko eziphansi zokubala. Imisebenzi ebikade ithatha amasonto ku-hardware evamile ingaqedwa ngezinsuku noma amahora. Lokhu kugxuma kokusebenza kwenza ucwaningo lwe-AI oluthuthukisiwe nokuthuthukiswa kufinyeleleke kakhudlwana. Iziqalo nezikhungo zemfundo ezingakwazi ukuzenzela amakhompuyutha azo amakhulu manje sezingakwazi ukuqasha isikhathi kuhardware ye-AI esezingeni lomhlaba ngokusebenzisa isixhumi esibonakalayo samafu. Kwenza intando yeningi ukufinyelela ku-firepower edingekayo ukuze uqhudelane enkundleni ye-AI. Sibone ukusebenzisana okufanayo kwamasu kusheshisa ukusungulwa kweminye imikhakha yezobuchwepheshe. Isibonelo, Amasonto ayisithupha asendle womdali we-NanoClaw aholele esivumelwaneni no-Docker akhombisa ukuthi ukuqondana nesiqhwaga sepulatifomu kunganikeza kanjani amandla amakhulu kubuchwepheshe obukhethekile.
Ukuhlanganiswa ne-Amazon Bedrock: A Strategic Play Ukusingatha isevisi ye-Cerebras ngaphakathi kwe-Amazon Bedrock kuwumsebenzi omkhulu. I-Bedrock iyisevisi ephethwe ye-AWS yamamodeli ayisisekelo, ehlinzeka ngekhithi yamathuluzi ehlanganisiwe yokwakha izinhlelo zokusebenza ezikhiqizayo ze-AI. Ngokungeza ama-Cerebras njengenketho ye-backend, i-AWS ifinyelela imigomo yamasu amaningana:
Isiphakamiso Senani Esithuthukisiwe: Senza i-Bedrock ibe isitolo esiphoqelela kakhulu sokumisa okukodwa, esinikeza amamodeli e-AI aphambili futhi manje, ihadiwe yokuqeqeshwa ye-elite-tier. I-Ecosystem Lock-in: Ikhuthaza onjiniyela bakhe, baqeqeshe, futhi bakhiphe amamodeli abo ngaphakathi kwe-ecosystem ye-AWS, ukusuka ekugcinweni kwedatha kuye ekusingatheni izinhlelo zokusebenza kokugcina. Umehluko Wokuncintisana: Ibala ngokuqondile ukunyakaza okufanayoizimbangi ezifana ne-Microsoft Azure ne-Google Cloud, nazo ezisebenzisana kabi nabasunguli be-chip.
Imithelela Ebanzi Yemboni ye-AI neTech Lesi sivumelwano siyinsimbi yekusasa lengqalasizinda ye-AI. Iqinisekisa ukuthi ayikho inkampani eyodwa, ngisho ne-titan yamafu njenge-Amazon, engaba umnikazi waso sonke isitaki. Abasunguli bezingxenyekazi zekhompuyutha abakhethekile abanjengoCerebras bazodlala indima ebalulekile ekuphusheni imingcele yalokho okungenzeka. Umjaho wokuphakama kwe-AI ulwelwa emikhawulweni eminingi, kusukela ekwakhiweni kwama-chip kuya ekwakhiweni kwamamodeli. Lokhu kubambisana kugqamisa ukuthi ukusebenza kwenkundla yefu—ikhono lokuhlanganisa ubuchwepheshe obusezingeni eliphezulu ngokushesha—kubaluleke njengokuchitha kwe-R&D okungaphekiwe. Izinkundla eziwinayo kuzoba yilezo ezinganikeza uhlu olubanzi lwamathuluzi anamandla, asebenziseka kalula. Lo mkhuba wokusebenzisa ubuchwepheshe obukhethekile ukuze uthole ukuncintisana awukhawulelwe kukhompyutha yamafu. Sikubona ekuhlaziyweni kwedatha yokuzijabulisa, njengoba kuhlolisiswa kokuthi Ungayenza Kanjani Imali Ukubikezela U-Oscar Uyawina kanye Ne-Box Office Hits Ngomkhiqizo Omusha We-MoviePass, futhi ngisho nakwezemidlalo, njengoba kubonakala nge-Kings League Ekhula Ngokushesha Ibukeka Inqoba I-America Nge-Lean Approach to Pro Sports.
Ongakulindela Ezinyangeni Ezizayo Njengoba kulindeleke ukuthi kwethulwe maduze, imboni izobe ibheke imininingwane ebalulekile. Amamodeli entengo, izinhlobo zezibonelo ezithile, kanye nedatha yokusebenza kwebhentshimakhi izobaluleka ukuze yamukelwe. Amakhasimende okufinyelela kusenesikhathi cishe azofaka amalebhu ocwaningo e-AI namabhizinisi amakhulu anamasethi edatha obunikazi. Impumelelo izokalwa ngokuthi isevisi ihlangana kanjani ngaphandle komthungo ekugelezeni komsebenzi we-AI okukhona kakade nokuthi ingabe iyaletha yini isithembiso sayo sejubane nesikali esingakaze sibonwe. Uma kwenzeka, kungase kubangele igagasi elisha lokuqanjwa kabusha kwe-AI, okwenza amamodeli angasebenzi okwamanje ngenxa yezinkinga zekhompiyutha.
Isiphetho: Isahluko Esisha se-Cloud-Native AI Isivumelwano se-chip ye-AWS ne-Cerebras singaphezu kokwethulwa komkhiqizo omusha; yiphuzu lokuguquguquka lamasu. Imele ukuvuthwa kwemakethe ye-AI yefu, lapho ukuthuthukiswa kokusebenza ngehadiwe ekhethekile kuba inkundla yempi eyinhloko. Kumabhizinisi, lokhu kusho izindlela ezisheshayo, ezingabizi kakhulu zokuthumela izixazululo ezinamandla ze-AI. Ukuhlala uphambili kuzwe lobuchwepheshe elithuthukayo kudinga imininingwane kulobu budlelwano obubalulekile kanye noshintsho lwemakethe. Ukuze uthole ukuhlaziya okwengeziwe ngamasu abumba ikusasa lobuchwepheshe nebhizinisi, hlola ukuphawula kochwepheshe kanye nemibiko ejulile etholakala ku-Seemless.
