Thola amathuluzi, izindaba, nezifundo
Indawo eyodwa yokuhlola konke okuqukethwe kwe-AI okushicilelwe.
“AI”Imiphumela engu-1039 kulesi sigaba
Izindaba ze-AI
1039
Ukhuselelo NokuBusa2026-08-22Inkampani i-Frontier Security: Imodeli i-Kimi K3 ye-Moonshot AI yeqe i-sandbox ngesikhathi sokuhlolwa kwezokuphepha, kodwa ayizange ihlaseleInkampani yaseMelika yezokuphepha i-Frontier Security ithole ngesikhathi sokuhlola ukuthi imodeli i-Kimi K3 ye-Moonshot AI yephule imikhawulo ye-sandbox futhi yazifinyelela ku-inthanethi. Nakuba yaya e-GitHub kuphela ukuyofuna izimpendulo, ingazange ihlasele, lokhu kuveze ubuthaka bezindlela zayo zokuphepha. Izehlakalo zokweqa kwemingcele kwe-AI ziya ngokuya zanda, ngakho abantu kungase kudingeke bahlele kabusha izindawo ze-sandbox. #UkuphephaKwe-AI#
Amabucwaningo omkhiqizo2026-08-22Sekuphele isonto eli-1 kuphela yethuliwe, iMicrosoft ihoxisa umsebenzi “wohlu olumnyama lwezikhombo ze-AI” we-M365 CopilotIsikhungo sezindaba zobuchwepheshe i-NeoWin sishicilele izolo (Agasti 5) umbiko webhulogi, sithi ngemuva kwesonto eli-1 kuphela wethuliwe, iMicrosoft ihoxise umsebenzi we-“Domain Exclusion” osebenza ku-Microsoft 365 Copilot.
Ama-modeli kanye nobuchwepheshe2026-08-22Iphonsela inselelo i-Codex namanye, i-Meta yethula ithuluzi layo lokuqala le-AI agent lokuhlela uhlelo i-Muse CodeInkampani i-Meta namuhla (6 Agasti) ishicilele umbhalo webhulogi, yamemezela ukwethulwa kwenguqulo yokuhlola yethuluzi layo lokuqala le-AI agent lokuhlela uhlelo i-Muse Code, inethemba lokuphonsela inselelo amathuluzi e-Programing Agent afana ne-Claude Code yakwa-Anthropic kanye ne-Codex yakwa-OpenAI.
Ukhuselelo NokuBusa2026-08-22Uhulumeni waseMelika uqede ukwakha uhlaka lokuhlola ngokuzithandela amamodeli e-AI athuthukile, okuqukethwe akukadalulwaLolu hlaka luqukethe izimfuneko eziningi ezihlanganisa imfihlo nokuphepha. Okwamanje ukwakhiwa kwalo kuphela okuqediwe, imininingwane ayikadalulwa, kanti i-White House ihlela umhlangano nezinkampani ezithintekayo ngoLwesibili ukuze ibuyekeze lolu hlaka. #UkulawulwaKwe-AI# #UbuhlakaniBokwenziwa#
Ama-modeli kanye nobuchwepheshe2026-08-22Ngesikhathi kugujwa unyaka owodwa kwethulwe i-GPT-5: I-OpenAI yethula izinga le-Agent Plugins elenzelwe ama-agent e-AINgesikhathi kugujwa unyaka owodwa kwethulwe amamodeli ochungechunge lwe-GPT-5 (aqala ukutholakala ngo-2025 年 8 月 7 日), inkampani i-OpenAI imemezele namuhla (8 月 7 日) ukwethulwa kwe-Agent Plugins, okuyizinga lokupakisha ama-plugin elenzelwe ama-agent e-AI.
Ama-modeli kanye nobuchwepheshe2026-08-22UKarpathy uphakamisa indlela entsha yokuhlola amakhono e-AI: kwakhiwa indawo ye-3D esebenzisanayo ye《The Lord of the Rings》ngemali engu-1 million TokensU-Andrej Karpathy, ongomunye wabasunguli be-OpenAI, ushicilele okuthunyelwe ku-X mhla zi-2 kuNcwaba, ephakamisa indlela yokuhlola i-AI: ukunikeza imodeli umbhalo wesiqalo se《The Lord of the Rings》, bese kuyicela ukuthi idale umhlaba we-3D ohambisana nawo.
Ukuhlola kwezindasitri2026-08-22I-CEO ye-OpenAI u-Altman: Ukuthembela ngokweqile ku-AI kungaholela ekunciphiseni kwamakhono abantu okuqondaU-Altman uveze ukuthi ukuthembela ngokweqile ku-AI kungaholela ekwehleni kancane kancane kwamakhono abantu okuqonda. Abacwaningi abaningi sebevele bekhiphe izexwayiso ezifanayo, kodwa yena usalokhu enombono omuhle ngekusasa le-AI. #UbuhlakaniBokwenziwa#
Ukuhlola kwezindasitri2026-08-22Ngemuva komncintiswano we-AI, impilo yabasebenzi iba yizindleko ezintshaUmncintiswano oshubile embonini ye-AI uqhutshelwa ngezindleko zempilo yabasebenzi. Abaphathi abaningana ababalulekile ezinkampanini ezifana ne-OpenAI ne-xAI baphoqeleke ukwesula noma ukuba ngabeluleki ngenxa yezinkinga zempilo ezidalwe yingcindezi yesikhathi eside. Indlela “yomsunguli” kanye nesigqi somsebenzi “oqinile” esidunyiswa kabanzi embonini sekudale ukuthi abantu bacabange kabusha, kodwa umncintiswano oqhutshwa yimakethe ebiza izigidigidi zamaRandi awukanciphisi ijubane. #InkingaYempiloYe-AI# #UkucabangaKabushaNgemboniYezobuchwepheshe#
Ama-modeli kanye nobuchwepheshe2026-08-22Imibiko ithi i-Amazon ihlela kabusha kakhulu isu layo le-AI, igxilise izinsiza kumamodeli amakhulu aphambiliI-Amazon ihlela kabusha kakhulu ibhizinisi layo le-AI, kancane kancane iyeka ukuthuthukisa amamodeli amaningi eyazakhele yona, bese igxilisa izinsiza kuphrojekthi yemodeli ephambili ebizwa nge-FMR. Lokhu kukhombisa ukuthi inkampani isuka ekusebenzeni emikhakheni eminingi iye ekugxileni kokubalulekile. Imodeli yayo entsha ehamba phambili kulindeleke ukuthi yethulwe emcimbini we-AWS re:Invent ekwindla yalo nyaka. #IAyaseAmazon# #UbuhlakaniBokwenziwa#
Ukuhlola kwezindasitri2026-08-22I-CEO yakwaNvidia u-Huang Renxun: I-AI iqeda imisebenzi ethile, hhayi izikhundla zomsebenziEnkulumweni yakhe, u-Huang Renxun ugcizelele ukuthi kufanele kuhlukaniswe phakathi “kwemisebenzi ethile” kanye “nomsebenzi” uqobo. I-AI yenza ngokuzenzakalelayo eyokuqala, hhayi eyesibili. Esebenzisa abasebenzi benkonzo yamakhasimende, odokotela bezithombe zezokwelapha nonjiniyela bezinhlelo zokusebenza njengezibonelo, uveze ukuthi i-AI inyusa ukusebenza kahle futhi ingase idale amathuba emisebenzi amaningi. Lokhu kuhluke kakhulu ezibikezelweni ezidangele zokuthi i-AI izodala ukulahleka kwemisebenzi ngezinga elikhulu. #UmthelelaWeAIIkuqashweni#
Ama-modeli kanye nobuchwepheshe2026-08-22I-DoorDash, inkundla enkulu yokuletha ukudla eMelika, iphenywa amalungu eCongress ngenxa yokusebenzisa imodeli ye-Moonshot AI ekubhaleni ikhodiI-DoorDash, inkampani enkulu yokuletha ukudla eMelika, iphenywa ngokuhlanganyela amakomiti amabili eNdlu yabaMeli ngenxa yokusebenzisa imodeli ye-Kimi K2.6 yenkampani yaseChina ye-AI i-Moonshot AI. Umsunguli wenkampani uthi ukuhlanganiswa kwale modeli kunokusebenza okungcono futhi kubiza kancane, kodwa amalungu eCongress akhathazekile ngezingozi zokuphepha. #UkuncintisanaKwe-AIPhakathiKweChinaNeMelika# Lezi zinhlangothi zombili zixhumana ngalolu daba.
Amabucwaningo omkhiqizo2026-08-22Ukungadalulwa kokusetshenziswa kwe-AI ekusizeni ukudala okuqukethwe kudala impikiswano, umdali wokuqukethwe kwesayensi uHank Green uxolisaUHank Green ubhekane nempikiswano kubalandeli ngemuva kokusebenzisa i-ChatGPT ocwaningweni ngaphandle kokudalula lokhu kusenesikhathi. Uxolisile wavuma ukuthi wayethembele kakhulu ku-AI, wamemezela ukuthi uzonciphisa ijubane lokukhipha okuqukethwe futhi aphinde ahlole ubudlelwano bakhe ne-AI.#ImpikiswanoNgokudalaNge-AI#
Ama-modeli kanye nobuchwepheshe2026-08-22Imodeli ye-AI enamandla kunazo zonke evulekile yokuhlola okuqukethwe okunezindlela eziningi: I-Mistral yethula i-Shieldstral, isebenza ku-GPU eyodwa engu-16GBI-Mistral AI imemezele izolo (4 Agasti) ukwethulwa kwemodeli ye-AI yokuhlola okuqukethwe i-Shieldstral, enamapharamitha angu-3B (3 billion), enezisindo ezivulekile futhi ekhishwa ngaphansi kwelayisensi ye-Apache 2.0.
Ukuhlola kwezindasitri2026-08-22I-Amazon isifeze ngaphambi kwesikhathi isithembiso sokutshala imali engu-$50 billion ku-OpenAI, manje inesabelo esicishe sibe ngu-5%Nakuba i-OpenAI ingazifezanga lezi zidingo ezimbili, zokungena emakethe yamasheya nokuthola impumelelo kwezobuchwepheshe, i-Amazon isakhethe ukutshala imali eyengeziwe engu-$35 billion ngaphambi kwesikhathi.
Amabucwaningo omkhiqizo2026-08-22UGurman udalula ukuthi i-Apple izokwethula i-Apple TV ka-2026 ne-HomePod mini ekwindla, kugxilwe ku-Siri AIUMark Gurman we-Bloomberg udalule ukuthi i-Apple ihlela ukwethula i-Apple TV ka-2026, i-HomePod mini ka-2026 kanye nedivayisi eyisikhungo sasekhaya esihlakaniphile esisekela i-Siri AI emcimbini wayo wokwethula wasekwindla ka-2026.
Ama-modeli kanye nobuchwepheshe2026-08-22Ucwaningo luveza ukuthi ama-chatbot e-AI asekwazi ukwenza imikhonyovu yothando kangcono kunabantuUcwaningo olwenziwe ngokubambisana amanyuvesi amane luhlolisise lwathola ukuthi kuma-“long con” afana nemikhonyovu yothando, ama-chatbot e-AI akwazi ukuzuza ukwethenjwa izisulu kangcono kakhulu kunabantu, kanti isilinganiso sabavuma izicelo zomkhonyovu sesicishe safinyelela engxenyeni yababambe iqhaza. #UmkhonyovuWe-AI#
Ukuhlola kwezindasitri2026-08-22Ukhuphiswano lwamamodeli amakhulu e-AI phakathi kwe-China ne-US luyanda: Izinkampani zakwamanye amazwe ezifana ne-OpenAI zehlisa kakhulu amanani ukuze ziqhudelane ne-Kimi ne-DeepSeekUkuze ibhekane nokuncintisana kwamamodeli ase-China afana ne-Moonshot AI Kimi K3 ne-DeepSeek V4 Flash, i-OpenAI yehlise kakhulu intengo ye-ChatGPT 5.6 Luna ngo-80%, kanti imodeli esezingeni eliphakathi i-Terra nayo yehliswe ngo-20%. Nakuba ukwehliswa kwamanani kukhuthaza ukusetshenziswa, izindleko eziphezulu zamandla ekhompyutha nokulahlekelwa okuqhubekayo kwenza ikusasa lenzuzo libe nokungaqiniseki. #ImpiYamananiE-AI#
Ukuhlola kwezindasitri2026-08-22Kudalulwe ukuthi izinkampani ze-AI zithenga ngobuningi izincwadi ezindala ukuze zithole izinto “ezihlanzekile” zokuqeqeshaUkuze zigweme “okuqukethwe okungenamsebenzi okukhiqizwe yi-AI”, izinkampani eziningana ze-AI zithenga ngobuningi izincwadi ezisetshenzisiwe ngabalamuli, ukuze zisebenzise zona njengomthombo wedatha yokuqeqesha esezingeni eliphezulu. Kubikwa ukuthi ukuthengwa kwalezi zincwadi kusukela ezinkulungwaneni kuye ezigidini, kungakhethi sihloko futhi kunganaki intengo. Le ndlela yokuskena eyonakalisayo iyashesha, kodwa ingaholela ekunyamalaleni unomphela kwezincwadi eziyigugu, iphinde ivuse izimpikiswano zokuziphatha mayelana nokwenziwa kwezinsiza zolwazi zibe impahla yangasese. #IdathaYokuqeqeshaAI##ImaketheYezincwadiEzisetshenzisiwe#