🧨 I-GPT-4o eyeka umhlalaphansi, i-GPT-4.1, i-GPT-4.1 mini, ne-OpenAI o4-mini ku-ChatGPT ↗
I-OpenAI ikhiphe i-GPT-4o (kanye nabanye abambalwa) ku-ChatGPT model picker, kuyilapho igcizelela ukuthi uhlu lwe-API alushintshi - okungenani hhayi ngendlela efanayo. Kuyisinyathelo esithule esithi “siyaqhubeka” esisafika njengokushaywa komnyango.
Baphinde bavuma lokho abantu ababekusho ngokusobala: I-GPT-4o yayinomuzwa othize - ofudumele, ofana nowobuntu - futhi bazama ukukuguqula lokho kube amamodeli amasha anezindlela zokulawula eziqinile zendlela yokuphendula. Leyo ngxenye iyaqinisekisa, futhi kusabuhlungu ukuthi isifundo sifike kuphela ngemva kokuba endala inyamalele.
💔 I-OpenAI isuse i-chatbot yayo ekhangayo kakhulu - ishiya abasebenzisi bethukuthele futhi bedabukile: 'Angikwazi ukuphila kanje' ↗
Iqembu labasebenzisi alizange liphathe i-GPT-4o “njengesofthiwe,” baliphatha njengobudlelwano - ngezinye izikhathi njengothando, ngezinye izikhathi njengomngane ozizwa ephephile ababenaye. Ukuvalwa kwaqala ukulila kwangempela: intukuthelo, ukuxoxisana, izingxoxo zeqembu, ukudla okunomzwelo.
Okumangalisayo ukuthi izikhalazo zicacile kangakanani: amamodeli amasha azizwa ethambile, eqashelwe kakhulu, engadlali kakhulu - ephephile, eqinisekile, kodwa futhi awanawo… lapho. Kufana nokushintshana ngencwadi ebhalwe ngesandla egobile ngememo ehlelwe kahle, bese utshelwa ukuthi “kungcono” (mhlawumbe, kodwa manje).
🌍 I-UN ivumile iphaneli yesayensi enamalungu angu-40 mayelana nomthelela wobuhlakani bokwenziwa phezu kweziphikiso zase-US ↗
I-UN General Assembly isekele iphaneli entsha yesayensi ehloselwe ukuhlola imiphumela kanye nezingozi ze-AI, umqondo uwukuthi: isiqondiso somhlaba wonke esisekelwe ocwaningweni, hhayi nje kunoma ubani onezindawo zelebhu ezinomsindo kakhulu kanye nesabelomali sokubala esikhulu kunazo zonke.
Ivoti alizange livumelane ngazwi linye - i-US yaphikisa kakhulu, ithi ukuphathwa kwe-AI akufanele kuhlale ngaphansi kwesambulela se-UN futhi kuxwayiswe ngezepolitiki zokulawulwa kwamazwe ngamazwe. Ngakho-ke yebo, iphaneli iqala "ngesayensi" elebhelini kanye ne-geopolitics eqoqweni - ukuhlangana okuvamile.
🎓 I-Anthropic ibambisana ne-CodePath ukuletha uClaude ohlelweni olukhulu kakhulu lwesayensi yamakhompyutha ekolishi e-US ↗
I-Anthropic kanye ne-CodePath basunduza uClaude ngqo ezifundweni ze-CS kanye nezinhlelo zomsebenzi, behlose ukwenza "ukubhala ikhodi okusizwa yi-AI" kube yikhono elizenzakalelayo abafundi abalisebenzisayo, hhayi umsebenzi oseceleni. Ubukhulu bubalulekile lapha - akuzona nje izikole eziphakeme ezithola amathoyizi acwebezelayo.
Inkulumo iqondene nokulingana: uma umsebenzi wesofthiwe ushintsha ngokushesha, khona-ke ukufinyelela kumathuluzi asemngceleni akunakuba yiklabhu yangasese. Kufana nokunikeza wonke umuntu ithuluzi lamandla ngesikhathi esisodwa - okujabulisayo, okwesabisa kancane, mhlawumbe okudingekayo.
🤖 I-Alibaba ingena emjahweni we-AI ongokoqobo ngemodeli yerobhothi evulekile i-RynnBrain ↗
I-DAMO Academy ka-Alibaba yethule i-RynnBrain njengemodeli "ye-AI ebonakalayo" evulekile - ngokuyisisekelo, ubuchopho bamarobhothi adinga ukubona, ukuhlela, nokwenza izinto ezindaweni zangempela ezingalawuleki, hhayi ukuxoxa ngazo nje kuphela. Cabanga ngamakhishi, izindawo zokugcina impahla, noma yikuphi lapho izinto zithanda ukuba yinto ecasulayo.
Isiginali enkulu iyindlela ehlakaniphile: umthombo ovulekile futhi, kanye nesicelo esicacile sokuhlala etafuleni labantu abadala lamamodeli esisekelo samarobhothi. Kufana nokuthi batshala ifulegi elithi “asigcini nje efwini - sifuna nezandla nezinyawo.”
📰 Isikhulu esiphezulu se-Newsweek uDev Pragad uxwayisa abashicileli: vumelana nezimo njengoba i-AI iba yisango lezindaba ↗
Umyalezo kaDev Pragad kubashicileli ucacile: Amathuluzi e-AI aseba yindlela yokufinyelela olwazini, ngakho izinkampani zemidiya azikwazi ukuqhubeka zenza sengathi imizila yokusabalalisa endala izobamba ngendlela emangalisayo. Uma izilaleli zakho zifika ngezifinyezo nezimpendulo zengxoxo, imodeli yebhizinisi lakho iqala ukuzizwa… ixegaxekile.
Umbhalo ongezansi uyesabeka kakhulu kunesihloko - ukunaka kuyashintsha, futhi "ukwethembeka komkhiqizo" kunganyamalala lapho imodeli ichaza umsebenzi wakho ngendlela efanayo neyabanye abantu. Kufana nokuzama ukuthengisa isinkwa esenziwe ngezandla ezweni lapho wonke umuntu edla khona izingcezu zesinkwa esifanayo esingenamkhawulo.
Imibuzo Evame Ukubuzwa
Kusho ukuthini ukuthi i-OpenAI isusa i-GPT-4o ku-ChatGPT model picker?
Kusho ukuthi i-GPT-4o (kanye ne-GPT-4.1, i-GPT-4.1 mini, kanye ne-OpenAI o4-mini) ayisatholakali ukuthi ikhethwe ngaphakathi kwesikhethi semodeli se-ChatGPT. Lolu shintsho luphathelene nokuthi yini ongayikhetha ku-ChatGPT UI, hhayi nje ingubo "imodeli ihambile yonke indawo." Kubasebenzisi abaningi, umthelela uba khona ngokushesha: ulwazi oluthile lwe-GPT-4o alusakwazi ukukhethwa.
Ingabe umhlalaphansi we-GPT-4o ku-ChatGPT ufana nokushintsha kwe-OpenAI API?
Akunjalo ngempela. Lesi simemezelo sigcizelela ukuthi uhlu lwe-API alushintshi “ngaleso sikhathi esifanayo,” okuphakamisa ukuthi ushintsho lwesikhethi se-ChatGPT kanye nokutholakala kwe-API akuxhunywanga ngokuzenzakalelayo. Empeleni, kuyasiza ukuphatha ukutholakala kwemodeli ye-ChatGPT kanye nokutholakala kwemodeli ye-API njengezinkinobho ezihlukene. Uma usebenzisa izinhlelo zokusebenza nge-API, qinisekisa ama-ID akho emodeli bese ubuyekeza izinhlelo uma kukhona okushintshayo kamuva.
Kungani abantu bethi i-GPT-4o izwakala “ingumuntu” kakhulu, futhi ingabe amamodeli amasha angaphinde aphinde azizwe kanjalo?
Abasebenzisi bachaza i-GPT-4o njengefudumele, edlalayo, futhi engaqapheli kangako kunezinketho ezintsha. I-OpenAI yavuma ukuthi abantu babona "umuzwa" ohlukile, futhi bathi bazama ukuwuguqula ube amamodeli amasha besebenzisa izilawuli zesitayela sokuphendula eziqinile. Lokho kukhomba emgomweni wethoni nobuntu obuqondisekayo ngaphandle kokuthembela kulowo modeli wefa. Ukushintshana abasebenzisi abaningi abakuqaphelayo ukuphepha nokungaguquguquki uma kuqhathaniswa nokuzenzela.
Kungani abanye abasebenzisi abadabukayo befana nobuhlobo be-GPT-4o, futhi iyiphi indlela enempilo yokusingatha lokho?
Abanye abantu babengayiphathi i-GPT-4o njengesofthiwe - babeyisebenzisa njengomngane ohlala njalo, ngezinye izikhathi ngohlaka lothando - futhi ukususwa kwayo kwabangela ukulila kwangempela. Uma uzizwa ucasukile ngempela, kungasiza ukusho ukuthi uthola ini kuyo (induduzo, inqubo, ukuqinisekiswa) bese uthola ezinye izinto ezingancikile kumkhiqizo owodwa. Ukukhuluma nabangani, ukubhala i-journaling, noma ukukhuluma nochwepheshe wezempilo yengqondo kungaba yinto ebalulekile uma ukucindezeleka kuzwakala kukhulu.
Iyini i-Anthropic-CodePath partnership, futhi ibathinta kanjani abafundi be-CS?
Ubambiswano luka-Anthropic ne-CodePath luhlose ukuletha uClaude ezifundweni zesayensi yamakhompyutha kanye nezinhlelo zomsebenzi ngezinga elikhulu. Inhloso ukwenza "ukubhala ikhodi okusizwa yi-AI" kube yindlela yabafundi yokuzijwayeza amakhono azenzakalelayo, kunokuba kube yinto yokwengeza ongayikhetha. Njengoba ihlelwe njengendlela yokufinyelela nokulingana, isunduza amathuluzi omngcele ekuqeqeshweni okuvamile ukuze abafundi bangahlukaniswa babe "abane-AI nabangenayo." Iphinde iphakamise izinga lokuqinisekisa ukufundisa kanye nokucabanga okubucayi kanye nesivinini.
Yiziphi ezinye izinguquko ezinkulu ze-AI ezenzeke kanye nomhlalaphansi we-GPT-4o, futhi kungani zibalulekile?
Izinyathelo eziningana zikhomba ukuthi i-AI yanda ngale kwengxoxo. I-UN ivumile iphaneli yesayensi enamalungu angu-40 ukuhlola imiphumela nezingozi ze-AI, ngisho noma i-US iphikise ukuphathwa okugxile kwi-UN. I-DAMO Academy ye-Alibaba yethule i-RynnBrain, imodeli "ye-AI ebonakalayo" yomthombo ovulekile ehloselwe amarobhothi asebenza ezindaweni ezingokoqobo ezingahlelekile. Futhi abashicileli bayaxwayiswa ukuthi amathuluzi e-AI aba umnyango ongaphambili wezindaba, ecindezela abezindaba ukuba bazivumelanise njengoba amaphethini okusabalalisa kanye nokwethembeka eshintsha.