🧱 ULutnick uthi iNvidia kumele iphile ngokuvikela ukuthengiswa kwayo kwama-chip e-AI eChina ↗
UNobhala Wezohwebo wase-US uHoward Lutnick uthe iNvidia ingathengisa ama-chip athile e-AI athuthukile eShayina - kodwa kuphela ngaphansi kwemibandela eqinile yelayisensi. Akusikho ukuthi “ungathengisi,” kodwa “uthengise, futhi ufakazele ukuthi uyakufanelekela.”
Imininingwane eyodwa ebabayo: kubikwa ukuthi la magama afaka phakathi izilawuli ezifana nokuhlolwa kwesitayela se-Know-Your-Customer ukunciphisa ingozi yokuthi ama-chips agcine esesetshenziswa empini. Ukuhoxisa i-Nvidia kuzwakala kungalindeleka, kodwa isikhathi sokuthobela imithetho sifika noma ngayiphi indlela.
💼 Umthombo uthi iBlackstone ikhuphule isabelo sayo kwinkampani entsha ye-AI i-Anthropic safinyelela cishe ku-$1 billion ↗
Kubikwa ukuthi iBlackstone yandise ukuchayeka kwayo ku-Anthropic kwaba cishe yi-$1B, yanezela imali eyengeziwe njengengxenye yomjikelezo wokuxhasa ngezimali obanzi. I-Big finance iyaqhubeka nokuthenga “abenzi bemodeli” sengathi bayizingqalasizinda, hhayi izinhlelo zokusebenza.
Inkulumo yokulinganisa okubikiwe yingxenye eyenza amashiya akho enze ukugxuma okuncane, okungahleliwe. Futhi, ukukhishwa okusha kwemodeli ye-Anthropic kuhlala ngemuva njengokuthi - siyathumela, qhubeka nokukhokha amasheke.
🧠 I-Cadence yethula i-ejenti ye-AI ukusheshisa ukwakheka kwama-chip ekhompyutha ↗
I-Cadence yethule i-ChipStack AI Super Agent, ngokuyisisekelo ithumela umsizi “we-agent” wokuklama nokuqinisekisa ama-chip - onjiniyela abancibilikisa ubuchopho abachitha isikhathi eside bekwenza lokho. Le nkampani ithi ingasheshisa eminye imisebenzi kakhulu ngokwakha “imodeli yengqondo” esebenzayo yomklamo bese iqhubeka nokuhlola nokulungisa amaphutha.
Kuyinguquko yesikhathi se-AI: ama-chip athuthuke kakhulu aklanywa ngokushesha… yi-AI… ukuze sikwazi ukwakha i-AI eyengeziwe. Inyoka idla umsila wayo, kodwa ngendlela ekhiqizayo ngendlela exakile.
🎬 I-Runway entsha yevidiyo ye-AI iqoqe u-$315M ngenani lika-$5.3B, ifuna amamodeli omhlaba anekhono elingcono ↗
I-Runway iqoqe uchungechunge olukhulu lwe-E futhi yabeka uxhaso njengophethiloli "wamamodeli omhlaba" - hhayi nje ukukhiqiza iziqeshana, kodwa ukwakha izinhlelo ezimele izindawo kahle ngokwanele ukuhlela nokulingisa. Lokho kuyisifinyezo esikhulu, kodwa isiqondiso sifundeka kahle: ividiyo eqondakala kakhudlwana, imihlaba ehambisana kakhudlwana, ubuso obuncane obuncibilikayo obungavamile (ngethemba).
Futhi bayanda ngale kwemidiya nezikhangiso beya ezintweni ezifana nemidlalo namarobhothi, okuyingxenye ezwakala ikhulu buthule… amamodeli evidiyo njengetshe lokuya emishinini eqonda izigcawu, hhayi nje ukuziveza.
🧩 Ihadiwe ye-AI kaJony Ives ihlehlisiwe ibe ngu-2027 futhi ngeke ibizwe ngokuthi io ↗
Icala lenkantolo lisikisela ukuthi iphrojekthi yehadiwe ye-OpenAI ehlobene noJony Ive isusiwe - futhi igama elithi "io" liyalahlwa ngenxa yokungqubuzana kwezimpawu zokuhweba. Ikusasa, elikhubekayo ngenxa yomkhiqizo, lizwakala liyinto exakile.
I-engeli yokulibaziseka ibalulekile ngoba ukuthandwa kwehadiwe bekulokhu kuzulazula isikhathi eside, futhi lolu hlobo lokusetha kabusha okulindelwe. Akubulali iphrojekthi - kumane kuyidonsela kuleyo ndawo enenkungu "ekugcineni" lapho imikhiqizo ilala khona.
🕵️ Izingxoxo 'ezingaziwa' zika-Anthropic ziphazanyiswe nguprofesa one-LLM ↗
Uprofesa waseNyakatho-mpumalanga ubonise indlela yokususa igama lengxoxo ekhishwe kuphrojekthi ye-Anthropic's Interviewer esebenzisa i-LLM engaziwa. Akuzona zonke - kodwa zanele ukwenza iphuzu lifike ngokushesha.
Kuyisikhumbuzo sokuthi “umbhalo ongaziwa” uvame ukufana “nombhalo ofihliwe kancane,” ikakhulukazi lapho amamodeli engakwazi ukunquma ubuwena kusukela ezingcezu zomongo. Ubumfihlo abuphulwa ngesinye isikhathi esimangalisayo - buyaphela.
🧾 Umthethosivivinywa omusha ungaphoqa izinkampani zobuchwepheshe ukuthi zibike ukuthi zisebenzisa okuqukethwe okune-copyright ekuqeqeshweni kwe-AI ↗
Isiphakamiso sezinhlangothi zombili (uMthetho we-CLEAR) singacindezela izinkampani ukuthi zidalule imisebenzi enamalungelo okushicilela esetshenziswa ekuqeqesheni amamodeli e-AI. Akuyona imfuneko yelayisensi eqondile - kufana nokuphoqelela izibani zikhanye ekamelweni elifiphaliswe ngamabomu.
Uma kwenzeka noma yikuphi, kungaguqula isimo sokulwa kwe-copyright: ukungasithembi kakhulu kodwa “ukukhombisa umsebenzi wakho wesikole.” Umbuzo omkhulu ukuthi lokho kuyaphoqelelwa yini, futhi ngandlela thile, iphuzu lonke.
Imibuzo Evame Ukubuzwa
Kusho ukuthini ngempela “ukuvikelwa” ekuthengisweni kwama-chip e-AI e-Nvidia eShayina?
Zibonisa ukuthi ukuthengisa kungase kuqhubeke, kodwa kuphela ngaphansi kwemibandela eqinile yelayisensi yoMnyango Wezohwebo wase-US. Esikhundleni sokuvinjelwa ngokuphelele, ukuma kuseduze nokuthi “uthengise, kodwa ufakazele ukuthi uyakufanelekela.” Empeleni, abathumeli ngaphandle kungadingeka babonise ukuthi ubani othengayo, ukuthi ama-chip azosetshenziswa kanjani, nokuthi yiziphi izinyathelo ezikhona zokunciphisa ingozi yokuphambuka.
Kubukeka kanjani ukuthobela imithetho ethi “Yazi Ikhasimende Lakho” ekuthumeleni kwamanye amazwe ama-chip e-AI athuthukile?
Ngokuvamile kuhilela ukuhlola abathengi, abalamuli, kanye nabasebenzisi bokugcina ngobudlova obukhulu kunasekuthengisweni kwebhizinisi okujwayelekile. Incwadi yokudlala evamile ihlanganisa ukuqoqa ulwazi oluqinile lobunikazi kanye nobunikazi, ukuqinisekisa ukusetshenziswa kokugcina okushiwo, kanye nokubheka izimpawu zokuthengisa kabusha noma amaphethini okuthunyelwa angavamile. Inhloso ukunciphisa amathuba okuba ama-chip agcine esekela ukusetshenziswa kwezempi noma okunye okunqunyelwe, kuyilapho kusavumela ukuthunyelwa kwezentengiselwano okuvunyelwe.
Kungani izinkampani ezifana neBlackstone zifaka imali ecishe ibe yi-$1 billion ku-Anthropic kanye nabanye abakhiqizi bamamodeli?
Abatshalizimali abakhulu baphatha izinkampani zamamodeli asemngceleni njengengqalasizinda: kuyabiza ukwakha, kubaluleke kakhulu, futhi kungenzeka kube yinto ebalulekile emikhiqizweni eminingi esezansi. Ukutshalwa kwezimali okulandelayo okubikiwe kungabonisa nesifiso sokugcina ukudalulwa njengoba izinga lezilinganiso likhula. Ngokuvamile, ukubheja ukuthi amandla emodeli, ukusatshalaliswa, kanye nokwamukelwa kwebhizinisi kuhlanganiswe ngokuhamba kwesikhathi - noma ngabe izindleko zesikhathi esifushane zihlala ziphezulu.
Ngingawachaza kanjani amanani amakhulu e-AI startup uma inkampani nayo ithumela amamodeli amasha ahamba phambili?
Inkulumo yokulinganisa ivame ukulandelela okulindelwe ngamandla emakethe yesikhathi esizayo kanye nemali engenayo yamanje. Ukuthumela amamodeli aqinile kungaqinisa umqondo wokuthi inkampani iyasebenza, hhayi nje ukuqoqa imali. Noma kunjalo, isignali ecacile ivame ukuba ukuheha: amakhasimende aphindaphindayo, ukusebenza okuthembekile, kanye nokuphepha emakethe. Indlela evamile ukubheka ukusetshenziswa komkhiqizo kanye nezibopho zebhizinisi kanye nezinombolo zezihloko.
Iyini i-Cadence's ChipStack AI Super Agent, futhi yiziphi izingxenye zomklamo wama-chip ezingazisheshisa?
Ibekwa njengomsizi "we-ejenti" wokuklama nokuqinisekisa ama-chip, kugxilwe emsebenzini ohamba kancane, ohlanganisa ukungqubuzana okukhulu njengokuhlola, ukulungisa amaphutha, kanye nokuphindaphinda imiklamo eyinkimbinkimbi. Umqondo uwukuthi ithuluzi lithuthukisa ukuqonda okusebenzayo komklamo, bese lisiza ekuqhubekiseni ukuhlola nokuthola izinkinga ngokushesha. Ezindleleni eziningi zokusebenza, izithiyo zokuqinisekisa yilapho isikhathi nomzamo wobunjiniyela kuqoqana khona.
Ayini “amamodeli omhlaba” kuvidiyo ye-AI, futhi kungani izinkampani ezintsha zibheja kuwo?
“Amamodeli omhlaba” ngokuvamile abhekisela ezinhlelweni ezimelela izindawo ezihlala njalo ngokwanele ukuhlela, ukulingisa, nokugcina izigcawu zihambisana ngokuhamba kwesikhathi. Ekudalweni kwamavidiyo, lokho kungahumusha kube amaphutha ambalwa okuqhubeka kanye nabalingiswa abazinzile, izinto, kanye nokunyakaza. Amandla afanayo angadlulela ngale kwemidiya - okuvame ukuxoxwa ngawo emidlalweni, ekulingiseni, kanye naserobhothini - ngoba kumayelana nokuqonda izigcawu, hhayi nje ukuveza ozimele.
Kungani amaphrojekthi ehadiwe e-AI ebambezeleka futhi eqanjwa kabusha, njengendaba yedivayisi ye-Jony Ive/OpenAI?
Izikhathi zehadiwe ziyashintsha ngenxa yezizathu eziningi: ama-prototype, imikhawulo yokuhlinzeka, ukuhlolwa kokusebenziseka, kanye nobunzima bokufanisa ikhono lesofthiwe nesici sefomu elibonakalayo. Ukuqamba izinguquko kungalandela ukungqubuzana kophawu lwentengiso noma ushintsho kusu lokufaka uphawu. Ukulibaziseka akubonisi ngokuzenzakalelayo ukuthi iphrojekthi ifile; kuvame ukukhombisa ukuthi ithimba lilungisa kabusha ububanzi, isisekelo sezomthetho, kanye nokulungela komkhiqizo ngaphambi kokuba kusakazwe emphakathini.
Umbhalo wengxoxo ye-AI "engaziwa" ungasuswa kanjani igama, futhi yini uMthetho we-CLEAR ohlose ukuyishintsha?
Umbhalo ungaveza ubunikazi ngezinkomba zomongo - okuhlangenwe nakho okuhlukile, izindawo, izikhathi, noma amagama - ngakho-ke i-LLM ngezinye izikhathi ingaphetha ngokuthi umuntu ungubani ngisho noma amagama esusiwe. Yingakho "ukungaziwa" kuvame ukudinga ukuvikelwa okuqinile kunokuhlelwa kabusha okulula. Ngokwehlukile, uMthetho we-CLEAR ohlongozwayo uzocindezela izinkampani ukuthi zidalule imisebenzi enamalungelo okushicilela esetshenziswa ekuqeqeshweni, ushukumise izingxoxo kusukela "ekuthembeni" ziye ekubonakaleni okulinganiselwe kakhudlwana.