Ukufunda Okungaphakathi Kokuqukethwe
Ukufunda ngaphakathi kokuqukethwe yikhono elimangalisayo lamamodeli amakhulu olimi lokuthatha umsebenzi omusha ezibonelweni ezimbalwa ezibekwe ekwazisweni, ngaphandle kokuqeqeshwa kabusha.
Uhlolojikelele
It is the reason you can 'teach' a model on the fly just by showing it what you want.
I-Deep Dive
Imvamisa, ukufundisa inethiwekhi ye-neural umsebenzi omusha kusho ukubuyekeza izisindo zayo ngokuqeqeshwa. Ukufunda okungaphakathi kokuqukethwe kuhlukile: ubhala izibonelo ezimbalwa ngqo ekwazisweni ('umongo'), futhi imodeli ithatha iphethini bese iyisebenzisa kokokufaka okusha. Akukho okushintshile ngaphakathi kwemodeli; izibonelo ziqondisa nje isibikezelo sethokheni elandelayo. Uzozwa 'i-zero-shot' (imiyalo kuphela), 'ishothi eyodwa' (isibonelo esisodwa), kanye 'nokudubula okumbalwa' (izibonelo ezimbalwa). Lokhu kuziphatha kwaduma yi-GPT-3 ngo-2020 futhi kwaba yikhono elivelayo: amamodeli amancane awakwazi ukukwenza, kodwa esikhathini esidlule cishe esikalini sepharamitha eyizigidi eziyizinkulungwane eziyi-100, ukunemba kokwaziswa okumbalwa kukhuphuka kakhulu. Imodeli ifunde ngempumelelo ukubona nokuqhubeka namaphethini phakathi nokuqeqeshwa kwangaphambili, ngakho-ke ingaphinda isebenzise lelo khono ngesikhathi sokunquma.
I-Technical Insight
Ucwaningo lokutolika lulandele ingxenye enkulu yaleli khono 'kumakhanda wokungeniswa' - amasekhethi okunaka avela ngesikhathi sokuqeqeshwa futhi enze ukufanisa isiqalo esingaqondakali: aphinde ahlole lapho kuvela khona ithokheni efanayo, bese ukopisha okulandelayo. Ngakho-ke uma umyalo wakho ubonisa 'i-apula -> isithelo, isanqante -> imifino', imodeli ifana nesakhiwo futhi ibikezela ilebula elilungile lento elandelayo. Okubaluleke kakhulu, awekho ama-gradients agelezayo futhi azikho izisindo ezibuyekezwa lapho kucatshangwa khona. Izibonelo zivele zimise kabusha ukwenza kusebenze okuphakelayo ukusabalalisa kwamathuba wethokheni elandelayo.
I-Strategic Impact
Isivinini nesikali
Ukugeleza komsebenzi wolimi kungahamba ngokushesha ngaphandle kokudela ukuvumelana.
Finyelela futhi ufinyelele
Yandisa ukufinyelela kuzo zonke izilimi nezitayela zokuxhumana.
Izinqumo ezicacile
Amaqembu angachitha isikhathi esiningi ekwahluleleni kuyilapho i-automation isingatha impinda.
Ikusasa Lokufunda Okungaphakathi Kokuqukethwe
Ukwandisa amafasitela womongo (manje amakhulu ezinkulungwane zamathokheni) kuphusha ukufunda okungaphakathi kokuqukethwe kuqondise ezimisweni 'ezithwebula kaningi', lapho izibonelo eziningi noma amakhulu zingancintisana nokulungisa eminye imisebenzi, ngaphandle kwezindleko zokuqeqesha. Lindela ukuhlanganiswa okuqinile nokubuyiswa, ukuze izibonelo ezifanele zilandwe ngokuzenzakalelayo, kanye nethiyori engcono yokuthi ukufunda okungaphakathi kokuqukethwe kwehluleka nini noma kuphazamiseka. Izohlala iyindlela esheshayo, eshibhile yokujwayela imodeli, ehambisanayo - hhayi esikhundleni - ukulungisa kahle imisebenzi ezinzile, enevolumu ephezulu.
Ukuqaliswa Komhlaba Wangempela
Ukunikeza i-chatbot izibonelo ezintathu zamathikithi okusekela nezigaba zawo, bese uwenza uhlukanise ithikithi elisha ngendlela efanayo
Ibonisa imodeli emibili ngaphambi/ngemuva kwamapheya ombhalo ongcolile ofomethiwe kabusha waba i-JSON ehlanzekile ukuze iguqule okunye
Ukunamathisela izincazelo ezimbalwa zesampula zomkhiqizo ngethoni yomkhiqizo wakho ukuze ezintsha zifane nesitayela
Ukubonisa inkinga yezibalo ezikhohlisayo kusetshenzwe isinyathelo ngesinyathelo ukuze imodeli ixazulule izinkinga ezifanayo ngefomethi efanayo yokucabanga.
Izingozi & Guardrails
Amaqiniso akhonjiwe angafaka ngokuthula imibiko, ukugeleza kosekelo, noma imiphumela yocwaningo.
Ukuzwela okusheshayo kungadala imiphumela engahambisani kuzo zonke izicelo ezifanayo.
Idatha yombhalo ebucayi ingase idalulwe uma izilawuli zokufinyelela zibuthakathaka.
Ukuqalisa Umhlahlandlela
Chaza ifomethi yokuphumayo, ithoni, namazinga wekhwalithi ngaphambi kokukhishwa.
Izimpendulo eziyisisekelo ngemithombo ethembekile noma nini lapho ukunemba kubalulekile.
Gcina indawo yokuhlola isibuyekezo somuntu ukuze uthole imiphumela ephezulu.
Landela amaphethini okuhluleka futhi uqeqeshe kabusha imiyalo noma ukuhamba komsebenzi njalo.
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 In-Context Learning 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
IWindows yokuqukethwe
Imibuzo evame ukubuzwa
What is In-Context Learning?
Ukufunda ngaphakathi kokuqukethwe yikhono elimangalisayo lamamodeli amakhulu olimi lokuthatha umsebenzi omusha ezibonelweni ezimbalwa ezibekwe ekwazisweni, ngaphandle kokuqeqeshwa kabusha. Kuyisizathu sokuthi 'ufundise' imodeli empukaneni ngokuyibonisa nje lokho okufunayo.
Yiziphi izinguquko ezibalulekile ngaphakathi kwemodeli ngesikhathi sokufunda okungaphakathi kokuqukethwe?
Ukufunda ngaphakathi kokuqukethwe kwenzeka ngokuphelele ngokuqonda. Izibonelo ekwazisweni zilolonga ukwenza kusebenze kwemodeli nokuqagela kwethokheni elandelayo, kodwa azikho izisindo ezibuyekeziwe.
Kungani ukufunda okungaphakathi kokuqukethwe kuchazwa njengekhono 'elivelayo'?
Ikhono libuthakathaka noma alikho kumamodeli amancane futhi likhuphuka ngokukhuphuka uma amamodeli efika esikalini esikhulu kakhulu, yingakho kubizwa ngokuthi abaqhamukayo.
Iyiphi indlela yangaphakathi ehlotshaniswa kakhulu nokufunda okungaphakathi kokuqukethwe kuma-transformer?
Umsebenzi wokutolika uhlonze amakhanda okungeniswa - amasekhethi okunakwa athola lapho ithokheni efanayo ivela khona futhi akopishe okwalandela - njengomshayeli oyinhloko wokufunda ngaphakathi kokuqukethwe.