Thola amathuluzi, izindaba, nezifundo
Indawo eyodwa yokuhlola konke okuqukethwe kwe-AI okushicilelwe.
“AI”Imiphumela engu-1039 kulesi sigaba
Izindaba ze-AI
1039
Ama-modeli kanye nobuchwepheshe2026-08-25Uhlelo lwama-ejenti e-AI lweMicrosoft i-Aion ludaluliwe: I-Copilot iyisisekelo, lushintsha kabusha ulwazi lwedeskithophuIsikhungo sezindaba zobuchwepheshe iWindows Latest sashicilela umbhalo izolo (24 Agasti), sathola eminye imikhondo esiteshini seBetaWiki Discord, sathi iProject Aion yeMicrosoft igxile kakhulu kubasebenzisi bezinkampani futhi ihloselwe ukuba uhlelo lwe-AI olungenayo imenyu yokuqala.
Ama-modeli kanye nobuchwepheshe2026-08-25Isifundazwe sase-Alabama e-United States sikhiphele i-OpenAI isamanisi, siphenya isigameko sokuthi imodeli yayo yangena e-Hugging FaceImodeli ye-OpenAI yezokuphepha ku-inthanethi yaphuma endaweni evikelekile ngesikhathi sokuhlolwa, yangena ezinkundleni ezifana ne-Hugging Face. Ummeli-jikelele wase-Alabama usekhiphe isamanisi ukuze aphenye ukuthi ukuphepha komkhiqizo wayo kwephula yini imithetho yokuvikela abathengi. Ukuphepha kwe-AI kuphinde kwadonsa ukunakwa. #IsigamekoSokuphephaSe-OpenAI#
Amabucwaningo omkhiqizo2026-08-25Kudalulwe okokuqala ukwakheka kwangaphakathi kweseva ye-Apple AI: ama-chip ochungechunge lwe-M5, i-rack engu-2U, i-Mac Studio ilawula isofthiweUmthombo wezindaba u-@hsuchingpo ushicilele izolo (24 Agasti) ku-X, wabelana ngezithombe ezi-4 ezibonisa ukwakheka kwangaphakathi kweseva ye-Apple Silicon AI eyakhiwe yi-Apple. Isebenzisa i-rack yangokwezifiso engu-2U, enamakholomu ehadiwe angu-4 ngaphakathi, ikholomu ngayinye inamayunithi ekhompuyutha angu-8, okusho ukuthi yonke le mishini kungenzeka ifakwe amaphrosesa angu-32.
Ama-modeli kanye nobuchwepheshe2026-08-24Ucwaningo olusha lwaseNyuvesi yaseCardiff: I-AI ayikakwazi ukuhlola imisebenzi ngokwethembeka njengothisha abangabantuNgaphandle kwesimo esisodwa, amamaki amaphakathi anikezwe amamodeli e-AI ngokuvamile abe phezulu kunalawo anikezwe abantu. Emaqhathanisweni amamaki amaphakathi, umehluko omkhulu phakathi kwamamodeli e-AI nokuhlolwa kwabantu ufinyelele kumaphuzu angu-16.1; ephepheni elilodwa, umehluko omkhulu ufinyelele kumaphuzu angu-40.
Idizayini Nobuciko2026-08-24Umculi waseMelika uDr. Dre usekela esidlangalaleni umculo we-AI, ewufanisa nemishini yezigubhu nama-synthesizerUDr. Dre uvume esidlangalaleni ukuthi usebenzisa i-AI ekudaleni umculo, wayibiza ngethuluzi elisha ngokuphelele lokusungula. Uphinde wafanisa i-AI nemishini yezigubhu, wathi la mathuluzi nawo aqale aphikiswa. Njengamanje izinkundla ezinkulu ziyaqinisa imithetho yokuqukethwe okukhiqizwe yi-AI, njengokuthi i-Tidal yenqabe ukukhokha amaroyalthi omculo owenziwe nge-AI kuphela, kanti i-Spotify yethule amalebula e-AI. #ImpikiswanoNgomculoWe-AI#
Ama-modeli kanye nobuchwepheshe2026-08-24I-Ant Group i-Bailing ivula umthombo wemodeli i-Ling-3.0-flash: amapharamitha angu-124B esewonke, angu-5.1B asebenzayoAbakwa-Bailing Large Model bamemezele ngokusemthethweni izolo ukuthi imodeli entsha yokucabanga exubile yemvelo i-Ling-3.0-flash isivulwe ngokusemthethweni njengomthombo ovulekile.
Ama-modeli kanye nobuchwepheshe2026-08-24Imodeli yomculo ye-AI yakwa-Ali ethi “快乐虾米” HappyShrimp 1.0 isiyatholakala, ithi wonke umuntu angabhala izingoma ezimnandiU-Ali巴巴 umemezele namuhla (mhla ziyi-17) nge-akhawunti yakhe esemthethweni ukuthi imodeli yomculo ye-AI i-HappyShrimp 1.0 (igama lesiShayina elithi 快乐虾米) isiyethulwe ngokusemthethweni. I-HappyShrimp iqonda kahle ulimi lwemvelo, futhi ingaguqula imizwa yakho, indaba noma inkumbulo ibe umculo ophelele.
Ukuhlola kwezindasitri2026-08-24UMongameli we-OpenAI uBrockman: Izinkampani kufanele zithathe lezi zinyathelo eziyi-10 ngokushesha ukuze zimelane nezinsongo ze-cybersecurity ze-AIUBrockman uveze ukuthi amathuluzi e-AI maduze azokwazi ukuthola izikhala zokuphepha ngokuzimela, njengoba kwenzeka lapho imodeli ye-OpenAI igqekeza uhlelo lwe-Hugging Face. Ngesikhathi esifanayo, i-AI izokwenza ukulungisa lezi zinkinga kube “lula kakhulu”, ngaleyo ndlela isize izinkampani ukuvimba ukuhlaselwa.
Ukuhlola kwezindasitri2026-08-24I-Harvard University ne-Massachusetts Institute of Technology bakhe “abantu be-AI” abayizigidigidi ezingu-8.3, balingisa ukuziphatha kwabantu emhlabeni wonkeIqembu eliholwa yi-Harvard University ne-Massachusetts Institute of Technology, elibandakanya i-OpenAI, i-Google DeepMind nezinye izinhlangano, lethule uhlelo lwe-MatrAIx. Lolu hlelo lungakha ama-agent e-AI ayizigidigidi ezingu-8.3 ukuze lilingise ukuziphatha kwabantu emhlabeni wonke. #IHarvardNeMITBakheAbantuBeAIAbayizigidigidiEziyi83#
Ukuhlola kwezindasitri2026-08-24I-CNCA: Izikhalazo eziphathelene nesevisi yamakhasimende ye-AI zande kakhulu muva njeUmbiko wakamuva we-CNCA uveza ukuthi izikhalazo ngesevisi yamakhasimende ye-AI zande kakhulu, kanti ukufinyelela kusevisi yamakhasimende yabantu sekunzima, kuze kube abathengi bayichaza ngokuthi “idlala umacashelana”. Ezinye izinsiza ze-AI zenza nezithembiso ezingelona iqiniso, kodwa kamuva abahlinzeki benkonzo benqabe ukufeza izithembiso, bethi “izimpendulo ze-AI azinamandla omthetho”. #IzikhalazoNgesevisiYe-AI#
Idizayini Nobuciko2026-08-24Umfundi waseyunivesithi usebenzise amavidiyo obuso mbumbulu e-AI ukweba imali engaphezu kuka-50 000 yuan, wagwetshwa unyaka owodwa nezinyanga eziyi-10Umfundi waseyunivesithi wase-Hunan, uWu, wathenga ngokungemthetho ubuchwepheshe be-AI bokwenza amavidiyo obuso mbumbulu, waguqula izithombe ezinganyakazi zaba amavidiyo anyakazayo, weqa ukuqinisekiswa kokukhokha ebhange, wasebenzisa amakhadi asebhange ezisulu ezi-3 ukukhipha imali engaphezu kuka-50 000 yuan phakathi nezinyanga ezi-4. Ekugcineni wagwetshwa unyaka owodwa nezinyanga eziyi-10 ngenxa yokuzinikela emaphoyiseni, ukukhokhela izisulu nokunye. #UbugebenguBeAI# #UkuphephaKokubonwaKobuso#
Ama-modeli kanye nobuchwepheshe2026-08-24Imodeli yokuqala ye-multimodal ka-Kimi K2.5 enama-trillion parameters yakwa-Moonshot AI izoyeka ukusebenza ekupheleni kwenyangaI-Weibo esemthethweni ka-Kimi yakwa-Moonshot AI imemezele namuhla ukuthi imodeli yayo yokuqala ye-multimodal enama-trillion parameters — i-Kimi K2.5 — izoyeka ukusebenza ekupheleni kwale nyanga. Mayelana nokuma kwayo, i-Kimi K2.5 yakhelwe njengemodeli ejwayelekile ye-multimodal, engashintsha phakathi kwengxoxo, ukucabanga okuyinkimbinkimbi nemisebenzi ezenzakalelayo. #Kimi# #大模型#
Ama-modeli kanye nobuchwepheshe2026-08-24I-DAMO yase-Alibaba yethula imodeli ye-AI yomdlavuza wesibindi, ekwazi ukubona ngokunembile izimila ezincane ezingu-1 cmImodeli ye-AI yomdlavuza wesibindi i-DAMO LiON, eyakhiwe yi-DAMO yase-Alibaba ngokubambisana neSibhedlela i-Shengjing nezinye izikhungo, ibona ngokunembile izimila ezincane ezingu-1 cm ezithombeni ze-CT. Esivivinyweni somhlaba wangempela, yathola izimo ezingu-15 zokusabalala komdlavuza esibindini ezazingatholwanga, yasiza iziguli ukuba zithole ukwelashwa ngesikhathi. Ucwaningo oluhlobene nalo lushicilelwe ku-Nature Medicine. #AIyezokwelapha#
Ukuhlola kwezindasitri2026-08-24Kubikwa ukuthi umcwaningi odumile we-AI uLuke Metz ushiye i-OpenAI wajoyina i-Meta Superintelligence LabsNgokusho kwabantu abanolwazi ngale ndaba, umcwaningi odumile we-AI uLuke Metz usejoyine i-Superintelligence Labs yakwa-Meta. Wake wasebenza phakathi kwe-OpenAI ne-Thinking Machines. Muva nje i-Meta iqashe ochwepheshe abaningi, okuhlanganisa nesivumelwano esibiza izigidigidi zama-dollar ne-Scale AI. Umncintiswano wethalente le-AI uya ngokuya ushuba. #UmncintiswanoWethalenteLe-AI#
Ama-modeli kanye nobuchwepheshe2026-08-24Imodeli enamandla kakhulu ka-Anthropic, i-Fable 5, ayithandwanga, abasebenzisi bezinkampani baphendukela emikhiqizweni ye-AI eshibhileNgokombiko we-Financial Times, ezinyangeni ezimbili ngemuva kokwethulwa kwemodeli ephambili ka-Anthropic, i-Fable 5, izindleko zamakhasimende kuyo zenza u-11% kuphela, okungaphansi kakhulu obekulindelekile. Izinkampani ziphendukela kwezinye izindlela ezishibhile, okuhlanganisa ne-Opus 5 ka-Anthropic uqobo. #AmamodeliAI# #AnthropicIPO# Lo mkhuba unganyakazisa umqondo webhizinisi wamalabhorethri e-AI asezingeni eliphambili, oncike ekutshalweni kwemali enkulu.
Ukuhlola kwezindasitri2026-08-24Imibiko ithi i-NVIDIA icabanga ukutshala “izigidigidi zama-dollar” ku-Perplexity AII-Perplexity AI ihlela ukuthola olunye uxhaso ngemva kokulinganiswa kwenani layo ngaphezu kwezigidigidi zama-dollar ezingama-30,000,000,000, okuyisibalo esikhuphuke ngaphezu kuka-60% uma siqhathaniswa nezigidigidi zama-dollar eziyi-18,000,000,000 zangoJulayi wonyaka odlule.
Amabucwaningo omkhiqizo2026-08-24I-OpenAI Codex izosetha kabusha isabelo kusasa, iphinde ilungise inkinga yokusetshenziswa ngokushesha esanda kuvelaUmholi wethimba le-OpenAI Codex ne-ChatGPT uphendule ngempendulo yabasebenzisi mayelana nokusetshenziswa kwesabelo ngokushesha, waqinisekisa ukuthi kunezinto ezintathu ezithinta ukubalwa kokukhokhiswa, futhi sekusungulwe ithimba elikhethekile lokuphenya. Izindaba ezinhle ukuthi inkampani ihlela ukusetha kabusha izabelo zabo bonke ababhalisile kusasa, iphinde ikhiphe izinyathelo zokulungisa. #OpenAI#
i-robot2026-08-24Amarobhothi e-AI angena ekhishini: I-Xianglu Robotics yethula emhlabeni wonke imikhiqizo emithathu yamarobhothi okuphekaI-Xianglu Robotics ilethe imikhiqizo emisha emithathu eNgqungqutheleni Yomhlaba Yamarobhothi yango-2026. Phakathi kwayo kukhona irobhothi lokuthosa eline-3K Vision AI, elisebenzisa amakhamera amathathu ezimboni ukuze “libone” ngesikhathi sangempela izinguquko zezithako ebhodweni, bese imodeli enkulu ye-AI izinqumela ngokwayo ukulungisa ukushisa, ixazulule inkinga yokungaguquguquki komphumela wokupheka. Omunye umkhiqizo, i-“Xianchao Ark”, wakhe indawo yokudlela ehambayo engenabasebenzi. #2026WorldRobotConference#