OpenAI Hleba
I-Whisper isistimu ye-_AIU_PROTECTED_10__ yokubona inkulumo ezenzakalelayo yomthombo ovulekile ebhala futhi ihumushe umsindo okhulunywayo ngezilimi eziningi.
Uhlolojikelele
It matters because it brought robust, free, near-human transcription to anyone who can run the model.
I-Deep Dive
Ikhishwe ngoSepthemba 2022, i-Whisper yaqeqeshwa cishe amahora angu-680,000 omsindo wezilimi eziningi, wemisebenzi eminingi oqoqwe kuwebhu. Leyo dathasethi enkulu nenhlobonhlobo iyimfihlo yokuqina kwayo: iphatha ama-accents, umsindo wangemuva, kanye ne-jargon yobuchwepheshe kangcono kakhulu kunezinhlelo ezindala, ngaphandle kokudinga ukucutshungulwa kahle kwesizinda ngasinye esisha. I-Whisper ingabhala inkulumo ngolimi loqobo, ihumushe inkulumo isuka ezilimini eziningi iye esiNgisini, ihlonze ulimi olukhulunywayo, futhi yengeze izitembu zesikhathi. OpenAI ikhiphe izisindo zemodeli nekhodi ngokuvulekile, ngakho isebenza endaweni kukhompuyutha ephathekayo noma esikhungweni sedatha, okugqugquzele ukuqhuma kwamathuluzi omphakathi, ukwenziwa kabusha okusheshayo, nezinhlelo zokusebenza ezakhelwe phezu kwayo. Ukunemba kuyahlukahluka ngolimi nekhwalithi yomsindo, futhi njengawo wonke amasistimu anjalo ngezinye izikhathi 'ingabonisa' umbhalo.
I-Technical Insight
I-Whisper iyi-Transformer encoder-decoder eqeqeshelwe umsebenzi wokulandelana-kuya-kulandelana. Umsindo uguqulwa ube i-spectrogram ye-log-Mel, ukumela okubonakalayo okufana kwamafrikhwensi ngokuhamba kwesikhathi, okucutshungulwa isifaki khodi. Idekhoda ibe isibikezela amathokheni ombhalo, afakwe ngaphansi kwamathokheni akhethekile atshela imodeli ukuthi yimuphi umsebenzi okufanele wenziwe: ukuloba, ukuhumusha, thola ulimi, noma engeza izitembu zesikhathi. Ngenxa yokuthi ifunde kumsindo wewebhu onelebula elibuthakathaka kuyo yonke imisebenzi eminingi ngesikhathi esisodwa, imodeli eyodwa yenza ngokubanzi ngokubanzi esikhundleni sokushunwa ibhentshimakhi eyodwa encane.
I-Strategic Impact
Finyelela futhi ufinyelele
Ithuthukisa ukufinyeleleka ngokuloba, ukulandisa, nezixhumi ezibonakalayo zezwi.
Izindleko kanye nesabelomali
Amaqembu emidiya angathumela umsindo opholishiwe ngokushesha ngamabhajethi amancane.
Isivinini nesikali
Amasistimu abhekene nekhasimende angacubungula ukusebenzelana okukhulunyiwe ngesilinganiso esikhulu.
Ikusasa le- OpenAI Whisper
I-Whisper isiphenduke ibhulokhi yokwakha ezenzakalelayo yokulotshwa, futhi ithrendi ibheke ekwahlukeni okusheshayo, okuncane, nokwesikhathi sangempela okusebenza kumafoni nakumadivayisi asemaphethelweni. Lindela ukusekela kokusakaza-bukhoma okuqinile, ukuhlukaniswa okungcono kwezipikha, nokuhlanganiswa namamodeli amakhulu olimi ukuze kuhlanzwe, kufinyezwe, namagama-ncazo abukhoma. Izisindo ezivulekile zisho ukuthi umphakathi uqhubeka nokukuthuthukisa, kuyilapho i-OpenAI nabanye bephusha amamodeli enkulumo amasha. Ukwehliswa kombhalo okhohliwe, ikakhulukazi ekusetshenzisweni kwezokwelapha nezomthetho, kuhlala kuyinto ehamba phambili.
Ukuqaliswa Komhlaba Wangempela
Intatheli ibhala izingxoxiswano ezirekhodiwe ngokuzenzakalelayo esikhundleni sokuzibhala ngesandla
Inkundla ye-podcast ikhiqiza okulotshiweyo okuseshekayo namagama-ncazo aso sonke isiqephu
Ithuluzi lomhlangano likhiqiza amazwibela abukhoma kanye nerekhodi elibhaliwe lekholi yevidiyo
Umcwaningi uhumusha okurekhodiwe kwenkundla yolimi olukhulunywayo kumbhalo wesiNgisi ukuze kuhlaziywe
Izingozi & Guardrails
Ukusetshenziswa kabi kwezwi kanye nezingozi zokuzenza ongeyena ziyanda uma imvume ingekho.
Ukunemba kungase kwehle kuzo zonke izinhlobo zokuphimisela, izilimi zesigodi, noma izindawo ezinomsindo.
Umsindo wokwenziwa ungenziwa iphutha njengenkulumo eyiqiniso ngaphandle kokulebula okucacile.
Ukuqalisa Umhlahlandlela
Thola imvume esobala yokuthwebula izwi, ukuhlanganisa, nokusebenzisa kabusha.
Ikhwalithi yokuhlola kuzo zonke izipikha nezimo zangemuva.
Chaza ukuthi kunini lapho umuntu kufanele abuyekeze noma agunyaze okuphumayo.
Lebula umsindo wokwenziwa futhi ugcine amarekhodi atholakalayo ukuze aziphendulele.
Qhubeka Uhlole
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the OpenAI Whisper quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Umhlahlandlela olandelayo
Ukuqondanisa Kwegama Elinesitembu Sesikhathi
Imibuzo evame ukubuzwa
What is OpenAI Whisper?
I-Whisper isistimu ye-_AIU_PROTECTED_10__ yokubona inkulumo ezenzakalelayo yomthombo ovulekile ebhala futhi ihumushe umsindo okhulunywayo ngezilimi eziningi. Ibalulekile ngoba ilethe okulotshiweyo okuqinile, kwamahhala, okuseduze komuntu kunoma ubani ongasebenzisa imodeli.
Iyini injongo eyinhloko yokuhleba kwe-OpenAI?
I-Whisper iwuhlelo oluzenzakalelayo lokuqaphela inkulumo oluloba futhi luhumushe umsindo okhulunywayo ezilimini eziningi.
Cishe ungakanani umsindo u-Whisper aqeqeshwe kuwo?
U-Whisper waqeqeshwa cishe emahoreni angu-680,000 omsindo wezilimi eziningi, wemisebenzi eminingi oqoqwe kusuka kuwebhu, okuyiwona oqhuba ukuqina kwawo.
Ikuphi ukumelela komsindo inqubo yesifaki khodi sika-Whisper?
Umsindo uguqulwa ube i-spectrogram ye-log-Mel, ukumelwa kokuqukethwe kwemvamisa ngokuhamba kwesikhathi, okufundwa yi-Transformer encoder.
Iliphi ikhono i-Whisper ENGAYENZA ngokwemvelo?
I-Whisper iphatha ukuloba, ukuhumushela kusiNgisi, ukutholwa kolimi, nezitembu zesikhathi, kodwa akuyona ijeneretha yevidiyo.
Kungani i-Whisper ibangele igagasi lamathuluzi omphakathi nezinhlelo zokusebenza?
Ngenxa yokuthi i-OpenAI ivule izisindo nekhodi, onjiniyela bangakwazi ukusebenzisa i-Whisper endaweni futhi bakhe amathuluzi amaningi nokuphinda kusetshenziswe ngokushesha.