I-OpenAI yembula i-chip eyenziwe ngokwezifiso eyaklanywa nge-Broadcom ukuze ithuthukise ingqalasizinda yayo ye-AI ↗
I-OpenAI ibonise i-Jalapeño, i-chip yayo yokuqala ye-AI eyenziwe ngokwezifiso eyakhiwe nge-Broadcom. Yenzelwe ukuqagela - umsebenzi ongakhangi kangako kodwa obaluleke kakhulu wokuphendula izimpendulo zomsebenzisi ngezinga elikhulu.
Le nkampani ithi amasampula asevele esebenza kumalebhu ayo, alungiselelwe imodeli yayo yokubhala ikhodi, futhi ahloselwe kuphela izinhlelo ze-OpenAI. Ngakho-ke, yebo, okunye ukushintsha kusuka ekuthembekeni ngokuphelele kweNvidia… noma okungenani umzamo oqinile.
Iphinde ithi i-AI isize ekusheshiseni izingxenye zomklamo we-chip. Iphindaphinda kahle: I-AI isiza ekuklameni ihadiwe ephakela i-AI eningi. Inyoka idla isangweji elinomumo we-GPU.
I-Qualcomm izothenga iModular entsha ngama-dollar ayizigidi eziyizinkulungwane ezine ngesoftware ye-AI push ↗
I-Qualcomm ithenga i-Modular ngesivumelwano esiphelele esibiza cishe amaRandi ayizigidi eziyizinkulungwane ezine, ihlose ngqo ungqimba lwesofthiwe olusiza amamodeli e-AI ukuthi asebenze kuma-chip ahlukene.
Lokho kubalulekile ngoba umsele we-CUDA we-Nvidia useyinto eyisimanga. Iphimbo likaModular liguquguquka kakhudlwana: vumela abathuthukisi basebenzise isiphetho ngaphandle kokubhala kabusha konke kweprosesa ngayinye.
Kalula nje, i-Qualcomm izama ukungabi “yinkampani yama-chip ocingweni” kodwa ibe “yinkampani encintisana kakhulu nesikhungo sedatha se-AI.” Isifiso esikhulu, isikhathi esijuluka kancane.
U-Anthropic uthi i-Alibaba ikhiphe ngokungemthetho amandla emodeli ye-Claude AI ↗
I-Anthropic isole abaqhubi abaxhumene ne-Alibaba ngokusebenzisa ama-akhawunti angamanga angaba ngu-25,000 ukuze bakhiphe amakhono kaClaude ngokusebenzisa "i-distillation," lapho imodeli ebuthakathaka ifunda emiphumeleni yemodeli enamandla.
Ubukhulu babubukhulu: U-Anthropic uthe lo mkhankaso udale ukuhwebelana okungaphezu kwezigidi ezingu-28.8 noClaude. Lokho akusikho ukuphenya nje okungahleliwe - lokho kuyiqembu ledijithali elisebenzisa imali eningi.
Lesi simangalo sifika phakathi kokukhathazeka phakathi kwe-US ne-China ngamamodeli athuthukile, amathuluzi e-cyber, nokuthi ubani othola ukufinyelela ezinhlelweni ze-AI ezibukhali kakhulu. Kuthukile, akumangazi.
I-RBI iphakamisa iziqondiso zamabhange zokuphatha izingozi ze-AI ↗
Ibhange eliphakathi laseNdiya liphakamise imithetho ecindezela amabhange ukuthi aqinise ukuqapha amamodeli e-AI kanye nokufunda komshini.
Lolu hlelo luzodinga izinhlaka zezingozi ezivunyiwe yibhodi, uhlu lwezimpahla zamamodeli, kanye nokuhlolwa okuqhubekayo kokubili imodeli ngayinye kanye nakulo lonke inhlangano.
Ngokuyisisekelo: sebenzisa i-AI, yebo, kodwa ungayilahli endaweni yokugcina impahla njenge-raccoon ene-calculator. Amabhange azodinga ukulawulwa okucacile ngaphambi kokuba ukwenza ngokuzenzakalelayo kusakazeke kakhulu ekubolekeni imali, ukukhwabanisa, kanye nokusebenza.
Bekufanele i-AI ibulale imisebenzi yobunjiniyela, kodwa idatha entsha iphakamisa ukuthi yizona eziqine kakhulu ↗
Ukuhlaziywa okusha kokuqashwa okuvela ku-SignalFire kusikisela ukuthi izindima zobunjiniyela zingase zibambe kangcono kunokulindelekile, naphezu komsindo mayelana namathuluzi okubhala ikhodi ye-AI athatha indawo yabathuthukisi.
Okubi ukuthi ubunjiniyela besofthiwe bekufanele bube yinto yokuqala ku-wood chipper. Esikhundleni salokho, idatha ikhomba ubunjiniyela njengenye yemisebenzi eqinile.
Lokho akusho ukuthi akekho ocindezelwayo. Kumane kwenza kube nzima indaba encane ehlanzekile ethi “i-AI ibulala abakopishi”, eyayihlala ihlelekile kakhulu.
I-AI isiza ekufundeni umqulu we-papyrus oshisiwe waba lukhuni ngesikhathi sokuqhuma kweVesuvius ↗
Abacwaningi basebenzise i-AI kanye ne-X-ray imaging ukuze bavule cishe umqulu we-Herculaneum oshile ngaphandle kokuwuvula ngokoqobo.
Bathola izinhlu ezingu-20 zombhalo ofihliwe ephepheni elingaphezu kwemitha elilodwa le-papyrus, kanti umbhalo ubonakala uxoxa ngemibono yamaStoyiki mayelana nokuziphatha, ukuziphatha, kanye nokuhlakanipha okusebenzayo.
Isimangaliso esincane, ngempela. Umqulu wasendulo oshile usufundeka kalula ngoba amamodeli anamuhla angabona izinsalela zeyinki emaphethini efayibha - umlando uphuma ngembobo ye-software.
I-Big tech ichithe izigidi emncintiswaneni owodwa we-US Congress. Ngeke kube okokugcina ↗
Isinqumo sokuqala seCongress yaseNew York saba yisibonelo sokuhlola imali ye-AI kwezepolitiki, kwathi imali engaphezu kwezigidi ezingama-24 zamaRandi yangena emncintiswaneni ivela emaqenjini asekelwa ubuchwepheshe.
Ingxenye enkulu yezindleko yayizungeze u-Alex Bores, owaxhasa umthethosivivinywa wokuphepha kwe-AI, kodwa wahlulwa nguMichael Lasher - naye owasekela umthethosivivinywa we-AI. Ngakho-ke imephu yemali yezobuchwepheshe yagoqwa ngendlela engavamile.
Iphuzu elikhulu licacile: Izinkampani ze-AI kanye nabagxeki bazo manje sebephatha ukhetho njengezinkundla zempi yenqubomgomo. Imodeli iyipolitiki, kodwa inephunga eliningi lesikhungo sedatha.
I-Met ithola ukwelulwa kwephrojekthi ye-Palantir AI ngemuva kokuthi uSadiq Khan evimbe isivumelwano ↗
Amaphoyisa aseMetropolitan aseLondon athole isikhathi sezinyanga ezingu-12 somshayeli we-AI onePalantir ngenkathi eqhuba inqubo entsha yokuthenga.
Le phrojekthi isetshenziselwa ukuqoqa idatha yangaphakathi kanye nokubeka uphawu ekuziphatheni kabi okungenzeka, inhlalakahle, noma ukukhathazeka ngamasiko ngaphakathi kwebutho.
Kuyasebenza, kuyaphikisana, futhi kuyasakaza imimoya kwezepolitiki - i-AI ejwayelekile emkhakheni womphakathi, ngokuyisisekelo. Ihhovisi likameya lalivimbe isivumelwano eside ngemuva kokukhathazeka ngokuthengwa kwempahla.
Imibuzo Evame Ukubuzwa
Iyini i-chip ye-OpenAI ye-Jalapeño AI esetshenziselwa yona?
I-chip ye-Jalapeño ye-OpenAI yenzelwe ukuqagela kwe-AI, okusho ukuthi isebenzisa amamodeli aphendula izimpendulo zomsebenzisi ngezinga. Ngokusho kwesihloko, yakhiwe nge-Broadcom futhi okwamanje ihloselwe izinhlelo ze-OpenAI kuphela. Amasampula asevele esebenza kumalebhu e-OpenAI, kanti ukulungiswa kugxile kumodeli yayo yokubhala ikhodi.
Kungani i-Qualcomm ithenga i-Modular ngesu layo le-AI?
I-Qualcomm ithenga i-Modular ukuze iqinise i-software yayo ye-AI, ikakhulukazi yokusebenzisa amamodeli ezinhlotsheni ezahlukene zama-chip. Lesi sivumelwano sikhomba ezifisweni ze-Qualcomm ngale kwama-processor efoni kanye nengqalasizinda yesikhungo sedatha se-AI. Isiphakamiso se-Modular ukuthi abathuthukisi bangakwazi ukwenza isiphetho ngendlela eguquguqukayo ngaphandle kokubhala kabusha konke kwe-processor ngayinye.
Kusho ukuthini ukuhluzwa kwemodeli ye-AI empikiswaneni ye-Anthropic ne-Alibaba?
Ukuhluzwa kwemodeli ye-AI yilapho imodeli ebuthakathaka ifunda emiphumeleni yemodeli enamandla. Kulokhu, i-Anthropic isola abaqhubi abaxhumene ne-Alibaba ngokusebenzisa ama-akhawunti amaningi mbumbulu ukuze bakhiphe amakhono kaClaude ngokusebenzisa ukuxhumana okuphindaphindiwe. Lesi sihloko sibonisa lokhu njengengxenye yokukhathazeka okubanzi mayelana nokufinyelela ezinhlelweni ze-AI ezithuthukisiwe kanye nokuncintisana kwamazwe ngamazwe.
Amabhange kulindeleke ukuthi aphathe kanjani izingozi ze-AI ngaphansi kwesiphakamiso se-RBI?
Isiphakamiso se-RBI sizocindezela amabhange ukuthi akhe ukuqapha okuqinile kwamamodeli e-AI kanye nokufunda komshini. Lokhu kufaka phakathi izinhlaka zezingozi ezivunyelwe yibhodi, izinqolobane zamamodeli, kanye nokuhlolwa okuvamile kokubili emazingeni omodeli ngamunye kanye nenhlangano yonke. Umgomo ukwenza i-AI isetshenziswe ngokulawulwa kakhudlwana njengoba isakazekela ekubolekeni, ekutholeni ukukhwabanisa, kanye nokusebenza.
Ingabe amathuluzi okubhala ikhodi ye-AI athatha indawo yemisebenzi yobunjiniyela ngempela?
Lesi sihloko sithi ukuhlaziywa okusha kokuqashwa kwe-SignalFire kwenza kube nzima umqondo wokuthi amathuluzi okubhala ikhodi ye-AI athatha indawo yabathuthukisi. Izindima zobunjiniyela zibonakala ziqine kakhulu kunalokho obekulindelwe, yize ubunjiniyela besofthiwe babuvame ukubikezelwa ukuthi buzodalulwa kakhulu. Lokho akusho ukuthi akukho ngcindezi, kodwa isithombe sibonakala silula kakhulu kunesithi “i-AI ibulala ababhali bekhodi.”
I-AI yasiza kanjani ekufundeni umqulu we-papyrus waseHerculaneum wasendulo?
Abacwaningi basebenzise i-AI kanye ne-X-ray imaging ukuze bavule cishe umqulu we-Herculaneum oshile ngaphandle kokuwuvula ngokoqobo. Le ndlela yasiza ekutholeni umbhalo ofihliwe kusuka ku-papyrus engaphezu kwemitha. Kubonakala sengathi umbhalo uxoxa ngemibono yamaStoic mayelana nokuziphatha, ukuziphatha, kanye nokuhlakanipha okusebenzayo, okubonisa ukuthi i-AI ingasekela kanjani ucwaningo lomlando olubuthakathaka.