I-OpenAI ithathe elinye igxathu ekuhlanganisweni kwe ubuhlakani bokwenziwa obusebenzisa ikhompyutha nsuku zonke ngokuvuselela ngokuphelele i-Codex, i-ejenti yayo yekhodi yedeskithophu. Leli thuluzi, elalaziwa kakhulu ngosizo lwayo lokuhlela, manje seliba umsizi okwazi ukuphatha izinhlelo zokusebenza, ukukhumbula izintandokazi, nokuphatha amaphrojekthi cishe kalula.
Ngalesi sibuyekezo, inkampani eholwa nguSam Altman isondeza uhlelo lwayo lokudala i- "Uhlelo lokusebenza oluhle kakhulu" lwedeskithophu oluhlanganisa i-ChatGPT, i-Codex, kanye nesiphequluli sakho se-AtlasLesi sinyathelo sihlose ukwenza lula ulwazi lomsebenzisi, ukugxila izinsiza endaweni eyodwa yokusebenza, kanye nokubhekana nezimbangi ezifana ne-Anthropic, ikakhulukazi endaweni yokuthuthukiswa kwesofthiwe kanye nokwenza izinto ngokuzenzakalela okuthuthukisiwe.
I-Codex ifunda ukusebenzisa ikhompyutha njenganoma yimuphi omunye umsebenzisi
Enye yezinguquko eziphawuleka kakhulu ukuthi iCodex manje isingakwazi lawula ikhompyutha ngemuvaI-ejenti iyakwazi ukubona okusesikrinini, ihambise ikhesa layo, ichofoze, ithayiphe umbhalo bese isabela kulokho okwenzekayo ezinhlelweni zokusebenza ezifakiwe, ngaphandle kokuphazamisa lokho umsebenzisi akwenzayo ngegundane noma ikhibhodi.
Leli khono livumela ama-ejenti amaningi e-Codex ukuthi asebenze ndawonye ngesikhathi esifanayo eqenjini elifanayoLokhu kuhilela ukwenza izivivinyo, ukushintsha ukucushwa, noma ukwenza imisebenzi ephindaphindwayo ngenkathi umuntu eqhubeka nemisebenzi yakhe evamile. Lokhu kuvula umnyango wokwenza imisebenzi isebenze ngokuzenzakalela eyayidinga ukugadwa njalo, njengokuhlola uhlelo lokusebenza ngemva kokushintsha kwekhodi noma izenzo zesixhumi esibonakalayo eziphindaphindwayo.
I-OpenAI iphakamisa lo msebenzi njengendlela yokuthi umsizi asebenze ngayo amathuluzi angenayo i-API noma ukuhlanganiswa okuqondile, into evamile ezinhlelweni eziningi zedeskithophu. I-Codex, "ngokubona" isikrini nokulawula i-cursor, ingasebenzisana nalezi zinhlelo zokusebenza njengoba nje umsebenzisi ongumuntu ebengenza, yize okwamanje ngendlela elawulwayo futhi egxile kakhulu ekuphepheni.
Ngokusho kwenkampani, ukusetshenziswa kwekhompyutha kuyatholakala ekuqaleni i-macOS ngokusebenzisa uhlelo lokusebenza lwedeskithophu ye-Codex, exhunywe kwi-akhawunti ye-ChatGPT. Inguqulo ye-Windows izofika kamuva, uma ukusebenza sekuhloliwe futhi izici sezivumelaniswe nemvelo ye-Microsoft.
Endabeni yeYurophu ne-United Kingdom, i-OpenAI iqinisekisile ukuthi ukulawula ikhompyutha, ukwenza ngokwezifiso okuthuthukisiwe, iziphakamiso zomongo, kanye nenkumbulo Zizosebenza kamuva. Isizathu yisidingo sokulungisa lezi zici ngokwemithethonqubo yokuvikela idatha, okuyisidingo esidinga kakhulu e-European Union kanye nasemakethe yaseBrithani.
Isiphequluli esihlanganisiwe kanye nomsebenzi oqondile kumakhasi ewebhu
Lokhu kubuyekezwa kuqinisa nobudlelwano beCodex newebhu ngenxa ye- isiphequluli esihlanganiswe ngaphakathi kohlelo lokusebenza ngokwaloEsikhundleni sokuthola nje imiyalelo yombhalo ejwayelekile, umenzeli angathola imiyalo ngqo ezintweni ezithile zekhasi.
Ngokwesibonelo, ekuthuthukisweni kwe-frontend, umsebenzisi angakwazi Khetha ingxenye yewebhu bese uchaza ushintsho. Umsebenzisi ufuna ukwenza okuthile (njengokushintsha umbala, usayizi wombhalo, noma ukwakheka kwebhulokhi), futhi i-Codex iyaqonda imiyalelo ngaphakathi kwalowo mongo obonakalayo othize. Lokhu kugwema ukubhala izimpendulo ezinde ezichaza ukuthi iyiphi ingxenye yesayithi edinga ukuguqulwa.
Lesi siphequluli esifakiwe sivumela ne-Codex Hlaziya okuqukethwe, uqoqe ulwazi, bese ulusebenzisa njengomongo Kweminye imisebenzi, njengokubhala ikhodi, ukubuyekeza imininingwane yomkhiqizo, noma ukuqhathanisa imibhalo yobuchwepheshe, konke kwenziwa endaweni efanayo, ngaphandle kokushintsha phakathi kwamafasitela amaningi esistimu yokusebenza.
Ukukhiqiza nokuhlela izithombe nge-gpt-image-1.5
Enye insika yenguqulo entsha ye-Codex ukuhlanganiswa kwemodeli i-gpt-image-1.5 yokudala nokuguqula izithombe Kusukela ngaphakathi kohlelo lokusebenza lwedeskithophu ngokwalo, i-ejenti ingakha amashadi, imidwebo, ama-mockup, noma imidwebo ebonakalayo evela embhalweni, noma ngokuhlanganisa imiyalelo nezithombe-skrini kanye neziqeshana zekhodi.
Leli khono liqondiswe ngqo emaqenjini asebenza e umklamo womkhiqizo, izixhumi zomsebenzisi, noma ama-prototypes emidlalo yevidiyoIsibonelo, unjiniyela angathwebula isithombe-skrini sesixhumi esiphakathi esiqediwe, acele i-Codex ukuthi iphakamise ukwakhiwa kabusha kwemenyu noma uhlelo lombala olusha, bese ethola izithombe ezikhiqizwayo ezifanelana nalowo mongo, ngaphandle kokushintsha amathuluzi.
Njengoba konke kuhlanganiswe ndawonye ku-Codex, ama-ejenti angakwazi ukuhlanganisa imisebenzi yokubhala ikhodi nokuklama: bhala umqondo wokusebenza, ubuyekeze i-frontend efanele bese udala i-mockup ebonakalayo ukuze uyethule eqenjini, ngaphandle kokushiya umsebenzi noma ukuvula izinhlelo zokusebenza zokuhlela izithombe zangaphandle.
Imemori, umongo, kanye nokwenza umsebenzi wesikhathi eside ngokuzenzakalela
Ngaphandle kokulawulwa kwekhompyutha kanye nokukhiqizwa kwezithombe, inhloso enkulu ye-OpenAI ngale nguqulo ukuhlomisa i-Codex nge inkumbulo eqhubekayo kanye nokuphathwa komongo okubanzi kakhuluIthuluzi lingaphinda lisebenzise izingxoxo zangaphambilini, lilondoloze ulwazi olubalulekile, futhi liluthole uma kudingeka ukuze kuqhutshekwe nephrojekthi.
Isici sememori, esikhishwa ekubukeni kuqala, sivumela i-Codex khumbula izintandokazi zakho, ukulungiswa okuvamile, noma idatha okwakunzima ukuyiqoqaLokhu kufaka phakathi konke kusukela ku-stack ejwayelekile yethimba kuya endleleni amafolda ahlelwe ngayo, amafomethi ombiko asetshenziswa kakhulu, kanye nethegi evame ukusetshenziswa ku-GitHub yezinhlobo ezithile zezinkinga.
Ngaleyo mongo eqongelelwe, umenzeli angakwazi phakamisa imisebenzi ngendlela ephumelelayoLokhu kungasetshenziswa ukusikisela ukuqalisa kabusha iphrojekthi engakaqedwa, ukuncoma ukulungiswa kabusha okusalindile, noma ukwazisa abasebenzisi ngezicelo zokudonsa ezinde. Ezinye izinkampani sezivele zikusebenzisa ukulandelela izingxoxo ku-Slack, Gmail, noma ku-Notion futhi zisabela lapho kwenzeka izenzakalo ezithile.
I-Codex ingakwazi futhi hlela umsebenzi wesikhathi esizayo bese uzivuselela ngokuzimela ukuqedela imisebenzi yesikhathi eside. Isibonelo, ingaphatha umugqa wokubuyekezwa kwekhodi izinsuku noma amasonto, ibhekane namazwana futhi ibuyekeze amadokhumenti ngaphandle kokuthi unjiniyela aqhubeke nokuwaqapha.
Izengezo ezingaphezu kuka-90 kanye nokusekelwa okuthuthukisiwe konjiniyela
Ukuze kuqedwe ushintsho, i-OpenAI ingeze cishe Ama-plugins amasha angu-90 e-CodexLokhu kufaka phakathi ukuhlanganiswa ne-Microsoft Suite, i-GitLab Issues, i-Neon by Databricks, i-Render, kanye ne-Superpowers, kanye nokusekelwa kwamaseva e-MCP (Model Context Protocol). Umgomo ukunikeza i-ejenti imithombo eminingi yomongo kanye namakhono amakhulu okusebenza.
Ngenxa yalezi plugins, i-Codex ingakwazi qoqa ulwazi oluvela ezinsizeni ezahlukene bese ulusebenzisa Ukusuka endaweni yokusebenza eyodwa, ungabuyekeza izinkinga ku-GitLab, ubuze ngedatha endaweni ye-Databricks, ubuyekeze amadokhumenti e-Office, noma uxhumane nezinsizakalo zokuthunyelwa, ngaphandle kokuphoqa umsebenzisi ukuba agxume esuka epulatifomu eya kwenye.
Ezingeni lobuchwepheshe kuphela, isibuyekezo siqinisa imisebenzi ejwayelekile yonjiniyelaManje i-Codex ikusiza ukuthi ubuyekeze izicelo zokudonsa, uphathe amathebhu amaningi e-terminal ngesikhathi esisodwa, uxhumeke ezindaweni zokuthuthukiswa ezikude nge-SSH, futhi uvule amafayela anokubuka kuqala okucebile, kungakhathaliseki ukuthi angama-PDF, amaspredishithi, izethulo, noma amadokhumenti angezinye izindlela.
Konke lokhu kugxile kokuhlangenwe nakho okuhlangene, ngokusho kwe-OpenAI uqobo, okuvumela intuthuko esheshayo kuzo zonke izigaba zomjikelezo wokuphila kwesofthiweKusukela ekubhaleni ikhodi yokuqala kuya ekuhlolweni, ekusetshenzisweni, ekubhalweni nasekugcinweni kwayo, kanye nama-ejenti e-AI asebenza ngokubambisana ngaphakathi kwendawo efanayo yedeskithophu.
Ukutholakala, ukugxila konjiniyela, kanye nesimo eYurophu
Inguqulo entsha ye-Codex isivele isetshenziswa ku- uhlelo lokusebenza lwedeskithophu lwe-macOS Kulabo abangena ngemvume nge-akhawunti ye-ChatGPT. Akuzona zonke izici ezidinga uhlobo olufanayo lokubhalisela, futhi ezinye izici ezithuthukisiwe zingase zincike ohlelweni oluthengiwe, njengoba kunjalo nangamanye amathuluzi e-OpenAI.
UThibault Sottiaux, umqondisi weCodex, uchaze ukuthi lokhu kubuyekezwa Iqondiswe kakhulu kubathuthukisiNakuba inhloso iwukukwandisa lokhu kuze kufike kubantu abaningi kamuva. Umqondo uwukuthi amaqembu obuchwepheshe azoba ngawokuqala ukusizakala ngama-ejenti amasha kanye nemisebenzi ezenzakalelayo, futhi ngokuhamba kwesikhathi, adlulisele leyo modeli emisebenzini yehhovisi, ekuphathweni kwamadokhumenti, noma ekukhiqizeni komuntu siqu.
I-OpenAI igcizelela ukuthi ukusetshenziswa kwamandla ekhompyutha kanye nememori kusetshenziswa nge indlela yokuphepha nokuthobela imithethoLokhu kubaluleke kakhulu e-European Union nase-United Kingdom. Ngenxa yalesi sizathu, izici zokwenza ngokwezifiso ezithuthukisiwe, iziphakamiso zomongo, inkumbulo eqhubekayo, kanye nokulawulwa kwesistimu yokusebenza okuqondile kuzothatha isikhathi eside ukuthi kutholakale kulezi zimakethe.
Kunoma ikuphi, inkampani igcizelela ukuthi lokhu kuthuthuka kweCodex kuvula umnyango we- uhlelo lokusebenza oluphezulu lwedeskithophu olugxile ezenzakalelayo ezihlakaniphilelapho ama-ejenti e-AI engasizi nje kuphela ukubhala ikhodi, kodwa futhi ehlela imisebenzi, azivumelanise nesitayela somsebenzisi, futhi agcine amaphrojekthi ayinkimbinkimbi esebenza isikhathi eside. Kwabathuthukisi besofthiwe eYurophu, umyalezo ucacile: amakhono amasha ayeza, kodwa ngokulungiswa okudingekayo ukuze kuhambisane nohlaka lokulawula olukhona.