🧹 I-GPT-4o eyeka umhlalaphansi, i-GPT-4.1, i-GPT-4.1 mini, ne-OpenAI o4-mini ku-ChatGPT ↗
I-OpenAI ithi izokhipha amamodeli ambalwa amadala avela ku-ChatGPT maduze nje - okuhlanganisa i-GPT-4o, i-GPT-4.1, i-GPT-4.1 mini, kanye ne-o4-mini - ngenkathi ishiya ukufinyelela kwe-API kungashintshiwe okwamanje. Ngokuyisisekelo ukuhlanza kwentwasahlobo... ngaphandle kokuthi umshanelo uhloselwe izintandokazi ezimbalwa zabalandeli.
Ingxenye ebabayo kancane ukuthi bayavuma ukuthi i-GPT-4o “yayinomuzwa” owawuthandwa ngabantu, kodwa ukusetshenziswa kwayo sekushintshe kakhulu ezinkethweni ezintsha. Ngakho-ke, yebo, kuyindlela yakudala ethi “siyakuzwa, kodwa siyakwenza noma kunjalo” - eqotho, noma esebenza kahle nje.
💰 I-Amazon ixoxisana ngokutshala imali efinyelela ku-$50 billion ku-OpenAI, kusho umthombo ↗
Kubikwa ukuthi i-Amazon isezingxoxweni zakuqala mayelana nokutshala imali efinyelela ku-$50B ku-OpenAI. Lelo nani likhulu kangangokuthi cishe alisho lutho - njengalapho umdlalo wakho uthola amaphuzu olwazini lwesayensi.
Umbiko ofanayo uthi i-OpenAI ifuna ukukhushulwa okukhulu kakhulu, kanti nabanye abasekeli abasindayo nabo bayajikeleza. Futhi yebo, i-Amazon isivele inesabelo esikhulu ku-Anthropic, ngakho lokhu kufana nokuvikela amasu… noma ukubamba isiteringi ngezandla zombili, kuye ngokuthi uzizwa unamandla kangakanani.
🪖 Imithombo ithi iPentagon ingqubuzana ne-Anthropic mayelana nokusetshenziswa kwe-AI yezempi ↗
Kubikwa ukuthi i-Anthropic kanye nePentagon baphikisana ngalokho okuvunyelwe - ikakhulukazi izivikelo ezingavimba izinto ezifana nokuqondiswa kwezikhali ezizimele noma amacala okusetshenziswa kokuqapha kwasekhaya. Isimo sePentagon sizwakala sisondele kakhulu kokuthi "uma kusemthethweni, kuyasebenziseka," kuyilapho i-Anthropic izama ukugcina inqubomgomo igxile engxoxweni.
Kuhlobene nezingxoxo zenkontileka ezibiza u-$200M, okwenza ukungezwani kuzwakale kungokoqobo kakhulu. Lesi ngesinye sezikhathi lapho "izimiso zokuziphatha ze-AI" ziyeka ukuba yingxoxo yephaneli bese ziphenduka amaphepha, imigqa yesabelomali, kanye nemihlangano exakile.
🔎 Ukudideka kusayine isivumelwano samafu se-AI esingu-$750 million ne-Microsoft, ngokusho kweBloomberg News ↗
Kubikwa ukuthi i-Perplexity isayine isivumelwano seminyaka eminingi esingu-$750M sokusebenzisa i-Microsoft Azure, kanti i-Microsoft's Foundry iba yisikhungo esiyinhloko sokuthola nokusebenzisa amamodeli - okuhlanganisa nezinhlelo ezivela kumalebhu amaningi adumile. Kuyisinyathelo esingathi sína sokukhulisa, futhi sidonsela i-Perplexity ekujuleni kwe-orbit ye-Microsoft (okuyinto ethokomele... kuze kube yilapho ingekho).
Okuphawulekayo ukuthi bathi futhi i-AWS isalokhu ingumhlinzeki wengqalasizinda abayithandayo, ngakho-ke akuyona ingxabano ehlanzekile - kufana nokwengeza ukhiye wesibili wefulethi "uma kwenzeka." Futhi, kukhona ukungezwani kwezomthetho ne-Amazon ngemuva, okwenza lonke unxantathu wothando lwamafu uzizwe uthukiwe kancane.
📉 Abatshalizimali bajezisa ukusetshenziswa kwe-AI ye-Big Tech okuletha ukukhula okuhamba kancane ↗
Isimo sengqondo sikaWall Street: sebenzisa imali engenangqondo ku-AI, yebo - kodwa kuphela uma ukukhula kubonakala kushesha futhi kusobala. Umehluko wawubukhali: I-Microsoft yathinteka kakhulu ngemuva kokuba imiphumela ingathobisanga ukukhathazeka mayelana nomfutho wamafu, kuyilapho i-Meta iveze izimpawu zokuthi umshini wayo wezikhangiso oqhutshwa yi-AI uyaphumelela.
Akukhona ukuthi abatshalizimali bayakuzonda ukusetshenziswa kwe-AI ngokuzumayo… bayakuzonda ukusetshenziswa kwemali “kwethemba thina mfowethu”. Okungukuthi, noma ngabe kwenza amamephu omkhiqizo azwakale sengathi anikezwa amamaki ngumgcinimabhuku ongenasineke.
🕵️ Owayengunjiniyela we-Google ulahlwe yicala lokweba izimfihlo ze-AI zezinkampani zaseShayina ↗
Owayengunjiniyela we-Google, uLinwei Ding, ulahlwe yicala enkantolo yase-US ngokweba izimfihlo zokuhweba ezihlobene ne-AI, kanti abashushisi bathi lokhu kwakuhloselwe ukuzuzisa izinkampani ezimbili zaseShayina ayesebenza nazo. Lezi zinsolo zigxile emibhalweni eyimfihlo ehlobene nengqalasizinda esetshenziselwa ukuqeqesha amamodeli amakhulu e-AI - izibilini ezingathandeki kodwa ezibaluleke kakhulu kulo lonke umsebenzi.
Leli cala liyingxenye yemizamo ebanzi yase-US yokuqeda ukudluliselwa kobuchwepheshe okubucayi, futhi kuyisikhumbuzo sokuthi "umncintiswano we-AI" awugcini nje ngokubala kanye nokuboniswa komkhiqizo - futhi uyizinkokhelo zobunhloli kanye nemiphumela yezomthetho engathandeki kakhulu.
Imibuzo Evame Ukubuzwa
Kusho ukuthini ukuthi i-OpenAI isusa i-GPT-4o ku-ChatGPT, futhi imaphi amamodeli asuswayo?
I-OpenAI ithi maduze izosusa amamodeli amaningana amadala kusikhethi samamodeli se-ChatGPT, okuhlanganisa i-GPT-4o, i-GPT-4.1, i-GPT-4.1 mini, kanye ne-o4-mini. Lokhu kushintsha lokho ongakhetha kukho ngaphakathi kwe-ChatGPT, hhayi lokho okukhona kwezinye izindawo. I-OpenAI ihlela lesi sinyathelo ngokuthi "ukuhlanza kwentwasahlobo," okuqhutshwa ukushintshela kokusetshenziswa ezinkethweni ezintsha. Uma umsebenzi wakho uncike ekuziphatheni kwemodeli ethile, lindela ukulungiswa komsebenzi bese uhlela ngokufanele.
Ingabe ukufinyelela kwe-API ku-GPT-4o noma ku-GPT-4.1 kuzoshintsha ngemva kokususwa ku-ChatGPT?
Ngokusekelwe kulokho okushiwo lapha, ukufinyelela kwe-API akukashintshi okwamanje, noma lawo mamodeli esusiwe ku-ChatGPT. Ngamanye amazwi, uhlelo lokusebenza noma ukuhlanganiswa kungaqhubeka nokusebenza njengokujwayelekile ngenkathi isikhethi semodeli ye-ChatGPT sishintsha. Ezindleleni eziningi, indlela ephephile ukulandelela izimemezelo zokutholakala kwemodeli nokugwema ukucabanga ukuthi iminikelo ye-ChatGPT ne-API ihlala ihambisana kahle. Uma ukuvumelana kubalulekile, gcina ama-fallbacks elungile.
Kungani i-OpenAI inqume ukuyeka i-GPT-4o yize abasebenzisi beyithandile "imizwa" yayo?
I-OpenAI ivuma ngokusobala ukuthi i-GPT-4o "yayinomuzwa" abanye abantu abawuthandayo, kodwa ithi ukusetshenziswa kwayo kushintshele kakhulu kumamodeli amasha. Empeleni, imikhiqizo ivame ukunciphisa okuzenzakalelayo ukuze kuncishiswe ukudideka, ukusekela izindleko, kanye nokuqhekeka. Lokho akwenzi ukukhetha kube yinto eqanjiwe - kusho ukuthi bekungaholi ukuthambekela kokusetshenziswa. Uma uyithandile i-vibe, cishe kuzodingeka uhlole izinketho ezintsha kuze kube yilapho eyodwa ifanelana nomsebenzi wakho ngendlela efanayo.
Kungasho ukuthini ukutshalwa kwezimali okubikiwe kwe-Amazon okufika ku-$50B ku-OpenAI kumncintiswano we-AI?
Kubikwa ukuthi i-Amazon isezingxoxweni zakuqala zokutshala imali efinyelela ku-$50B ku-OpenAI, kanye nomzamo omkhulu wokuqoqa imali. Uma kwenzeka, kungasho ukuthi kuzoba nothango olukhulu lweqhinga - ikakhulukazi njengoba i-Amazon isivele inesabelo esikhulu ku-Anthropic. Kuye ngemigomo, ingase ibumbe ubudlelwano, ukukhetha ingqalasizinda, kanye nokuncintisana phakathi kwabahlinzeki be-AI abakhulu kanye nabahlinzeki bamafu. Njengoba izingxoxo zichazwe njengasekuqaleni, akukho okuxazululiwe.
Kungani iPentagon ne-Anthropic bephikisana ngezivikelo ze-AI zempi?
Imibiko ithi uMnyango Wezokuvikela wase-US kanye ne-Anthropic baxabana ngemikhawulo engavimba ukusetshenziswa njengokuhlasela izikhali ezizimele noma ukuqapha kwasekhaya. Isimo sePentagon sichazwa njengesisondele kakhulu “uma kusemthethweni, kuyasebenziseka,” kuyilapho i-Anthropic ifuna imingcele yenqubomgomo eqinile eyakhiwe kusukela ekuqaleni. Njengoba izingxoxo zenkontileka ziboshelwe ku-$200M, impikiswano iguqula ukuziphatha kube ulimi lwenkontileka. Yilapho izithiyo zokuvikela zibamba khona - noma zinyamalale buthule.
Yini eshintsha isivumelwano se-Perplexity sika-$750M Azure, futhi kungani sigcina i-AWS futhi?
Kubikwa ukuthi i-Perplexity isayine isivumelwano samafu seminyaka eminingi esingu-$750M esigxile ku-Azure, kanti i-Microsoft's Foundry ibekwe njengesikhungo esiyinhloko sokuthola nokusebenzisa amamodeli. Ngesikhathi esifanayo, kubikwa ukuthi isabiza i-Amazon Web Services ngokuthi umhlinzeki wayo wengqalasizinda okhethwayo, okukhomba ukuma kwamafu amaningi kunokuba kube ukushintshana okugcwele. Izisusa ezivamile zifaka phakathi ukuqina, ukuxoxisana ngegunya, kanye nokuguquguquka kwamandla. Ukungqubuzana kwezomthetho kwangemuva ne-Amazon kunezela umongo owengeziwe kuleyo ndlela yokuhlukanisa.
Kungani abatshalizimali bejezisa ukusetshenziswa kwe-AI ye-Big Tech yize i-AI iyinto eza kuqala?
Isihloko esichazwe siqondile: abatshalizimali bazobekezelela ukusetshenziswa kwemali enkulu ye-AI lapho ukukhula kuvela ngokucacile nangokushesha, kodwa abayithandi imigomo yesikhathi "yokusethemba". Esibonelweni esinikeziwe, iMicrosoft yathinteka lapho imiphumela ingakudambisi ukukhathazeka mayelana nomfutho wamafu, kuyilapho iMeta iphakama ngezibonakaliso ze-AI ezithuthukisa ukusebenza kwezikhangiso. Okufanele ukwenze ukuthi izimakethe zifuna ubufakazi - ukukhushulwa kwemali engenayo, izinzuzo zokusebenza kahle, noma indlela ecacile yokukhokha. Ukusebenzisa imali kuphela akuyona indaba; izinzuzo ezibonisiwe ziyizo.
Kwenzekeni ecaleni lezimfihlo zokuhweba lonjiniyela wangaphambili we-Google, futhi yini okufanele izinkampani ziyifunde kulo?
Unjiniyela we-Google uLinwei Ding ulahlwe yicala enkantolo yase-US ngokweba izimfihlo zokuhweba ezihlobene ne-AI, kanti abashushisi bathi lezi zinto zazihloselwe ukuzuzisa izinkampani ezimbili zaseShayina ayesebenza nazo. Kubikwa ukuthi imibhalo ihlobene nengqalasizinda yokuqeqesha amamodeli amakhulu e-AI - ubuchwepheshe obubucayi, obungabonakali. Leli cala lisezandleni ze-US zokunciphisa ukudluliselwa kobuchwepheshe okubucayi. Kumaqembu, liqinisa izilawuli zokufinyelela eziqinile, ukuqapha okuqinile, kanye nokulawula okuqondisiwe.