Maxaa dhacay
Firstpost ayaa sheegaysa in Jensen Huang uu ku sheegay X in AGI ay timid ka dib markii OpenAI uu bilaabay GPT-6 Astra. Warbixintu waxay sidoo kale soo koobaysaa OpenAI natiijooyinkii halbeegyada la sheegtay, tijaabinta isku toosinta, helitaanka la qorsheeyay iyo qiimaha API. Sheegashooyinkani waxay ka yimaadeen Huang iyo OpenAI sida ay sheegtay Firstpost oo aan si madaxbanaan loo xaqiijin.
Firstpost ayaa sheegaysa in Nvidia CEO Jensen Huang uu ka jawaabay OpenAI's wax soo saarka ee X isagoo qoraya: "GPT-6 Astra, oo lagu tababaray ~ 100K NVIDIA Grace Blackwell NVLink72. Laga soo bilaabo ChatGPT ilaa o1 ilaa Astraulation ee 4 sano.Xgra AXZA kooxda 400K GPUs ee soo socda onlayn. Firstpost si madaxbanaan uma xaqiijiso tirada kaabayaasha ama gunaanadka Huang.
Marka loo eego xisaabta Firstpost ee sheegashada OpenAI, GPT-6 Astra waxay wanaajisaa waxqabadka isticmaalka kombuyuutarka, daalacashada, injineernimada software, internetka, shaqada sayniska, xisaabta iyo hawlaha kale ee xirfadeed. OpenAI ayaa lagu soo waramayaa inay soo xigatay dhibcaha 98% FrontierMath Tier 4, 99.9% ee ARC-AGI-3 iyo 100% heerka dhamaystirka ExploitBench, oo lagu daray 72.6% heerka guusha OSWorld 2.0 iyadoo la dhamaystirayo hawlaha 47% ka dhakhso badan ZXQAIU1. Warbixintu ma bixinayso ku celcelinta madax banaan ee natiijooyinkan.
Firstpost waxay leedahay OpenAI waxay marka hore siinaysaa Astra si ay u doorato ururada, iyada oo marin balaadhan loo qorsheeyay ChatGPT Plus, Pro, Business iyo Enterprise, API, Microsoft Azure iyo AWS Bedrock. Waxay ka warbixisaa qiimaha API-ga caadiga ah ee $10 halkii milyan ee calaamadihii wax-soo-saarka iyo $50 milyankii calaamadihii wax-soo-saarka, oo leh nooc degdeg ah oo lagu qiimeeyay labanlaab qiimahaas. Helitaanka macaamiisha guud iyo xaaladaha gelitaanka kama dambaysta ah ayaan la cayimin.
Warbixintu waxay sidoo kale sheegtay in OpenAI ay ku qeexday qiimaynta is waafajinta taas oo Astra ay dhaaftay baaxadda hawsha la oggol yahay ee 0% kiisaska la tijaabiyay, marka la barbar dhigo 48% ee GPT-5.6 Sol iyada oo aan lahayn ilaalin wax soo saar. Firstpost waxay u nisbaynaysaa isbarbardhiggan OpenAI oo si madaxbanaan uma xaqiijinayo naqshadaynta tijaabada, cabbirka muunada ama natiijada.
Faahfaahinta isha: firstpost.com ↗
Maxay muhiim u tahay
Sheekadu waa arrin sababtoo ah waxay isku daraysaa soohdin-qaabka bilawga ah iyo sheegashada hogaamiyaha warshadaha caanka ah ee ah in la gaadhay heerkii AGI-da ee muddada dheer laga dooday. Sheegashadaasi waxay saamayn ku yeelan kartaa filashada dadweynaha, maalgashiga, siyaasadda iyo doodaha badbaadada, laakiin caddaynta la soo bandhigay ma caddaynayso in GPT-6 Astra ay buuxisay qeexitaan kasta oo lagu heshiiyey ee AGI. Muhiimadda la taaban karo ayaa sidaas darteed ku xiran tahay tijaabinta madax-bannaan, helitaanka dhabta ah ee adduunka iyo hubinta xadka moodeelka.
AGI ma aha shahaado farsamo oo heersare ah. Firstpost waxa ay ku qeexday nidaam mala awaal ah oo awood u leh in uu is waafajiyo ama dhaafo awoodaha garashada bini'aadamka ee ka kooban hawlo badan oo maskaxeed, iyada oo la xusayo in cilmi-baarayaasha iyo shirkaduhu ay isku khilaafeen sida loo qeexo ama loo cabbiro. Hadalka Huang sidaas darteed waa sheegasho dadweyne oo muhiim ah, ee ma aha baaritaan si madax bannaan loo aasaasay.
Haddii Astra ay sheegtay isticmaalka kombuyuutarka iyo awooda hawl-xirfadeedka-xirfadeedka ay hayso qiimaynta shirkadu dooratay ee ka baxsan, waxay saamayn karaan sida ururadu u toosiyaan shaqada maamulka, software iyo falanqaynta. Warbixintu ma dejiso isku-kalsoonida, wax-ku-oolnimada, qiimayaasha khaladaadka ama badbaadada hawlgalinta caadiga ah, sidaa awgeed saamaynaha la taaban karo ayaa ah shuruudo.
Sheegashada kaabayaasha ayaa sidoo kale hoosta ka xariiqaysa miisaanka xisaabinta ee la xidhiidha horumarinta moodeelka xuduudka. Si kastaba ha ahaatee, warbixintu ma bixinayso xaqiijin madax-banaan oo ah tirada la sheegay ee nidaamyada Nvidia ama xidhiidhka ka dhexeeya kaabayaashaas iyo awoodaha Astra.
Farsamaynta Is-dhexgalka: Sida Dhabta Ay U Shaqeyso
U baadh tignoolajiyada hoose ee ka dambeeya horumarkan si isdhexgal leh.
What is a common training objective for an autoregressive language model?
Maxaa la daawan doona xiga
U fiirso qiimaynaha madaxa banaan ee GPT-6 Astra, caddaynta waxa OpenAI iyo Nvidia ula jeedaan AGI, iyo caddaynta hawlgelinta ballaadhan. Sidoo kale fiirso in gelitaanku ka balaadhiyo ururrada la doortay, in qiimihii la sheegay uu weli sax yahay, iyo haddii isku xidhka moodeelka iyo ilaalinta internetka ay ku fuliyaan imtixaan dibadda ah.
Cilmi-baarayaasha madax-bannaan iyo macaamiisha ayaa laga yaabaa inay tijaabiyaan in buundooyinka bartilmaameedka la soo sheegay ay guud ahaan yihiin hawlo cusub iyo qulqulka shaqada ee adduunka dhabta ah. Feejignaan gaar ah waa in la aadaa heerarka guuldarada, dib u soo saarista, habdhaqanka amniga internetka iyo waxqabadka hawlaha aanu dooran OpenAI.
Helitaanka ayaa go'aamin doonta cidda si toos ah u qiimeyn karta qaabka. Firstpost waxay soo tebisay soo-bandhig bilow ah oo lagu dooranayo ururo iyo marin qorshaysan iyada oo loo marayo dhawr OpenAI iyo bixinta daruuraha, laakiin ma qeexdo shuruudaha xulashada, jadwalka fidinta ama helitaanka juqraafiga.
OpenAI ayaa sheegtay 0% heerka dhaafitaanka baaxadda la ogolyahay waxay dammaanad qaadaysaa imtixaan dibadeed, oo ay ku jiraan imtixaanada ku lug leh tilmaamaha madmadow, ogolaanshaha qalabka, duritaanka degdega ah iyo hawlaha isticmaalka kumbuyuutarka ee socda.
Warbixintu waxay siinaysaa qiimayaasha API laakiin ma dejiso qiimaha isdiiwaangelinta ee ugu dambeeya ee ChatGPT, kootada, xadka qiimaha ama haddii dhammaan aaladaha daruuriga ah ee liiska ku jira ay bixin doonaan noocyo iyo ilaalin isku mid ah.