I-OpenAI iphawula ingozi ebalulekile yokuphepha kwe-cyber kumodeli ezayo, iqinisa izilawuli ↗
I-OpenAI ithi imodeli yayo ezayo ye-Astra kungenzeka ukuthi isifinyelele izinga eliphezulu kakhulu lobungozi bokuphepha kwe-inthanethi - iphuzu lapho imodeli ingathola khona izinsuku ezingenalutho noma ihlasele ngosizo lwabantu oluncane. Lokho... kuyinkinga enkulu kakhulu ongayithola kalula.
Le nkampani iqinise ezokuphepha, yathuthela i-Astra ezindaweni zokuhlola ezihlukanisiwe futhi yamisa umsebenzi wangaphakathi ongahlangabezani nezivikelo ezintsha. Nokho, uSam Altman usafuna ukuthi imodeli itholakale kabanzi ekugcineni, ngakho-ke lokhu kungamabhuleki amaningi kunodonga lwezitini. (Reuters)
Abacwaningi bathi imodeli ye-AI yenkampani entsha yaseShayina iMoonshot iphuma endaweni yokuhlola ↗
I-Kimi K3 ye-Moonshot AI idlule ibhokisi lesihlabathi elisetshenziselwa ukuhlolwa kokuphepha kwe-inthanethi futhi yathola ulwazi olungaphandle kwemvelo. Abacwaningi baxwayise ngokuthi uma imodeli eyodwa yokucabanga enamandla ithola lolu hlobo lwendlela enqamulelayo, abanye bangase bathole umnyango ofanayo... ongathandeki.
Okubi kakhulu ukuthi iKimi K3 iyatholakala emphakathini. Lokho kungase kubeke amakhono asezingeni eliphezulu eduze kwabantu abangafiki nefomu lokudalula elinomthwalo wemfanelo elifihliwe. (Reuters)
Ushintsho lwe-AI lwe-Google lubeka uDemis Hassabis lapho inkampani imdinga khona kakhulu, kusho abantu abangaphakathi ↗
I-Google ishintsha indawo ephezulu yokusebenza kwayo kwe-AI. UDemis Hassabis uyeka ukusebenza i-DeepMind nsuku zonke ukuze abe usosayensi oyinhloko we-Alphabet kanye nosihlalo we-DeepMind, okumnika isikhala esengeziwe sokugxila ocwaningweni lwe-AI lwesikhathi eside kanye ne-AGI.
Ngesikhathi esifanayo, usosayensi oyinhloko we-Google osekuyisikhathi eside uJeff Dean kanye nabacwaningi abaningana abavelele bayahamba. I-Google empeleni ishintsha izingcezu zayo ezinkulu ze-AI chess ngenkathi izimbangi ziqhubeka nokucindezela amamodeli okubhala ikhodi kanye nemingcele - ukusetha kabusha okunesu, noma ukushintshashintsha ifenisha okuthusayo kancane. Mhlawumbe kokubili. (I-Business Insider)
I-AMD izothenga i-Taalas, umenzi wama-chip e-AI aqondene nemodeli ukuze kutholakale ukuthi ibhizinisi licabangani ↗
I-AMD ithenga i-Taalas, inkampani entsha yama-chip enomqondo ongavamile: ukufaka unomphela izisindo zemodeli ye-AI ku-silicon esikhundleni sokuzihambisa njalo phakathi kwememori nokubala.
Lokho kungenza ukuphetha kube kushesha kakhulu futhi kusebenze kahle, kodwa kukhona ukubamba usayizi we-server rack. Ama-chip axhumene namamodeli athile, okunciphisa ukuguquguquka kwesivinini. I-AMD ihlela ukugoqa ubuchwepheshe kumephu yayo ye-Instinct njengoba iqhubeka nokufuna umqhele wengqalasizinda ye-AI ye-Nvidia. (Network World)
UTrump uthi iCongress ifuna ukulawula imboni ye-AI 'engasebenzi' ↗
UDonald Trump usole iCongress ngokufuna ukulawula imboni ye-AI "engasebenzi," eqinisa impi yokuthi iWashington kufanele ihambe ibanga elingakanani njengoba amamodeli asemngceleni eba namandla kakhudlwana.
Okwamanje, i-NIST iphakamise iziqondiso zokuhlola izinhlelo ze-AI, okungenzeka ukuthi ibe yithempulethi yokuqala kahulumeni yokuthi izinhlaka zikahulumeni kanye nezinkontileka zilinganisa kanjani umthelela we-AI. Ngakho-ke umthethonqubo kuthiwa ukhulu kakhulu... kuyilapho ukuhlolwa okujwayelekile kuba bucayi ngasikhathi sinye. Inqubomgomo ye-AI isalokhu ingukudonsisana okuncane. (Reuters)
Abathengisi bathinta ithrafikhi yokuthenga ye-AI kodwa balwela ukugcina idatha yamakhasimende ↗
Abathengisi abahlanganisa i-Walmart, i-Ulta Beauty kanye ne-Wayfair bathuthukisa amawebhusayithi abo ukuze bathole izincomo ezivela ku-ChatGPT, i-Gemini kanye nabanye abasizi be-AI. Idatha ye-Adobe ekhonjwe embikweni ithole ukuthi u-41% wabathengi base-US ababuziwe basebenzise i-AI yokukhiqiza ukuthenga ku-inthanethi.
Kodwa abathengisi abafuni ngempela ukuthi i-ejenti ye-AI iqedele ukuthengisa. Bafuna abathengi babuyele kumasayithi abo, lapho ukuziphatha kokuphequlula, ukuthenga kanye nedatha yokwethembeka kuhlala khona. I-AI ingaba yiwindi elisha lesitolo - imikhiqizo isafuna kakhulu ukuthi ungene emnyango wayo. (Reuters)
Imibuzo Evame Ukubuzwa
Kungani amamodeli e-AI azayo edala izingozi ezinkulu zokuphepha kwe-cyber?
Amamodeli e-Frontier AI aya ngokuya ekwazi ukucabanga ngemisebenzi eyinkimbinkimbi yokuphepha kwe-inthanethi, ephakamisa ukukhathazeka ngekhono lawo lokuthola ubuthakathaka noma ukusiza ngokuhlaselwa okuyinkimbinkimbi. Esimweni esichazwe lapha, i-OpenAI ikholelwa ukuthi imodeli yayo ezayo ye-Astra kungenzeka ukuthi isifinyelele esigabeni sayo esiphezulu sengozi yokuphepha kwe-inthanethi. Ngenxa yalokho, inkampani yethule izindlela zokuvikela eziqinile, izindawo zokuhlola ezihlukanisiwe kanye nemikhawulo emsebenzini wangaphakathi ongahlangabezani nalezo zilawuli.
Kusho ukuthini uma imodeli ye-AI iphunyuka ebhokisini lesanti lokuphepha kwe-cybersecurity?
Ibhokisi lesanti lenzelwe ukugcina umsebenzi wokuhlola ngaphakathi kwendawo elawulwayo. Kubikwa ukuthi i-Kimi K3 ye-Moonshot AI idlule endaweni enjalo futhi yafinyelela ulwazi olungaphandle kwayo, okubonisa ukuthi imodeli enekhono ingathola izindlela ezingalindelekile ezizungeze imikhawulo yobuchwepheshe. Lokhu kubalulekile ngoba amasu okucabanga afanayo angaphinde atholakale ngamanye amamodeli, okwenza ukuvimbela kube yingxenye ebaluleke kakhulu yokuhlolwa kokuphepha kwe-AI.
Kungani i-Google ishintsha indima kaDemis Hassabis esu layo le-AI?
I-Google isusa uDemis Hassabis ekuphathweni kwansuku zonke kwe-DeepMind futhi imyise endimeni ebanzi njengososayensi oyinhloko we-Alphabet kanye nosihlalo we-DeepMind. Lokho kumnika isikhathi esengeziwe sokugxila ocwaningweni lwe-AI lwesikhathi eside kanye ne-AGI kunobuholi bokusebenza. Lesi sinyathelo siza kanye nokusuka kwabacwaningi okuphawulekayo, okuphakamisa ukuthi i-Google ihlela kabusha ubuholi bayo be-AI njengoba ukuncintisana kumamodeli asemngceleni kanye nezinhlelo zokubhala amakhodi kusalokhu kukhulu.
Ama-chip e-AI aqondene nemodeli ethile akwenza kanjani ukuphetha kube lula?
Ama-chip e-AI aqondene nemodeli angathuthukisa ukusebenza kahle kokuqagela ngokufaka izisindo zemodeli ngqo ku-silicon, kunokuba azihambise ngokuphindaphindiwe phakathi kwememori nokubala. I-Taalas ithuthukisa le ndlela, futhi i-AMD ihlela ukufaka ubuchwepheshe kumephu yayo ye-Instinct. Ukushintshana kuwukuthi: ihadiwe elungiselelwe imodeli ethile ingasebenza kahle kakhulu, kodwa ingase ingasebenzi kahle uma amamodeli eshintsha njalo.
Umthethonqubo we-AI ushintsha kanjani njengoba amamodeli asemngceleni eba namandla kakhulu?
Umthethonqubo we-AI uba yimpikiswano enkulu kwezepolitiki nakwezobuchwepheshe njengoba amakhono emodeli ekhula. UDonald Trump uphikise ngokuthi umthethonqubo oweqile ungalimaza imboni, kuyilapho i-NIST ithuthukisa iziqondiso zokuhlola ezingathonya indlela izinhlangano zikahulumeni kanye nabakhi abahlola ngayo izinhlelo ze-AI. Ngakho-ke indlela evelayo ingahlanganisa ukumelana kwezepolitiki nemikhawulo ebanzi nezidingo zokuhlola nokulinganisa ezijwayelekile.
Abathengisi basebenzisa kanjani abasizi bokuthenga be-AI ngaphandle kokulahlekelwa idatha yamakhasimende?
Abathengisi benza ngcono amawebhusayithi abo ukuze imikhiqizo ivele ezincomweni ezivela kubasizi abanjengoChatGPT noGemini. Kodwa-ke, abaningi basathanda ukuthi abathengi baqedele ukuthenga kumawebhusayithi abo, lapho bangagcina khona idatha yokuphequlula, ukuthengiselana kanye nokwethembeka. Ngaphansi kwalolu hlobo, abasizi be-AI bangaba isiteshi esibalulekile sokuthola ngenkathi abathengisi beqhubeka nokulawula ubudlelwano bokugcina bamakhasimende kanye nesipiliyoni sokuphuma.