AI alignment
Aiwatar da AI shine aikin fasaha da cibiyoyi na samar da tsarin AI na ci gaba da dogaro da dogaro da abin da mutane suke niyya - gami da a cikin labari, yanayi mai girma inda tsarin ya fi wayo, sauri, ko kuma mai cin gashin kansa fiye da masu sarrafa shi.
Zurfafa nutsewa
Daidaitawa ba ɗaya ba ne da 'ɗa'idodin AI' a faffadan ma'ana. Da'a tana tambayar wane darajoji yakamata al'umma su bi; daidaitawa yana tambayar ko tsarin AI mai ƙarfi zai ci gaba da bin manufofin da muka ƙayyade - kuma ko waɗannan manufofin sun tsaya tsayin daka yayin da ƙarfin ya girma. Hanyoyin rashin nasara na gargajiya sun haɗa da ƙayyadaddun wasan kwaikwayo (inganta ma'aunin wakili), kuskuren manufa (mun rubuta makasudin da ba daidai ba), da haɗin kai (tsarin da ke neman iko, albarkatu, ko adana kai saboda waɗanda ke taimakawa kusan kowane manufa ta ƙarshe). Dakunan gwaje-gwaje na zamani sun riga sun sami nau'ikan waɗannan gazawar: masu yin taɗi waɗanda ke yarda da masu amfani da hankali, wakilai waɗanda ke yin amfani da madaidaicin madauri a ayyukan zura kwallaye, da ƙirar ƙirar wasan. Tambayar a buɗe ita ce ko hanyoyin daidaitawa na yau (RLHF, AI tsarin mulki, muhawara, fassara, dabarun sarrafawa) ma'auni zuwa tsarin da zai iya tsarawa, yaudara, ko aiki tare da ƙarancin kulawar ɗan adam. Wannan shine dalilin da ya sa binciken jeri yana zaune a tsakiyar muhawarar haɗarin AI na wanzuwa: idan tsarin ingantaccen tsarin ba daidai ba ne, matakan amincin samfur na yau da kullun na iya kasa isa.
Fahimtar Fasaha
Yawancin 'daidaitawa' da aka tura a yau shine haɓaka fifiko a saman ƙirar tushe da aka riga aka horar: tattara kimar ɗan adam (ko AI) na abubuwan fitarwa, horar da samfurin lada ko amfani da hanyoyin fifita kai tsaye (DPO da bambance-bambancen karatu), sannan sabunta manufofin. Wannan yana inganta matsakaicin taimako kuma yana rage wasu lahani, amma baya tabbatar da samfurin yana da burin ciki wanda ya dace da niyyar ɗan adam, kuma ba zai yi aiki da kyau ba a ƙarƙashin canjin rarraba, hukumar dogon sararin sama, ko matsin lamba. Fassara, sa ido mai daidaitawa, da kimantawa don yaudara yunƙurin wuce yarda da saman ƙasa.
Dabarun Tasiri
Haɗari da aminci
Bala'i da cutar AI ta yau da kullun duka sun dogara da wanda ya fahimci haɗarin kuma wanda zai iya yin aiki.
Shawarwari masu haske
Ilimin jama'a da na ƙwararru yana siffanta ko ƙaƙƙarfan manufofin aminci na yiwuwa a siyasance.
Cutting through hype
Bayyanar bayani yana rage kama ta hanyar zage-zage, dakin gwaje-gwaje PR, da gidan wasan kwaikwayo mara kyau.
Makomar AI alignment
Yi tsammanin ƙarin aiki akan auna amincin sarkar-tunani, gano makirci ko jakan yashi, jan-ƙungiya mai sarrafa kansa, da hanyoyin sarrafawa waɗanda ke ɗaukar daidaitattun daidaito. Al'amuran ilimin jama'a a nan: mutanen da kawai ke jin 'daidaitacce = sanya masu yin hira da ladabi' za su yi ƙasa da yanayin gazawar bala'i da wuce gona da iri na tallace-tallace daga labs.
Aiwatar da Gaskiyar Duniya
Mataimakan horarwa tare da bayanan zaɓin ɗan adam (RLHF) don haka sun ƙi bayyananniyar cutarwa kuma su bi umarni mafi kyau.
Ma'aikatan ƙungiyar ja don ba da lada don satar kuɗi: bin harafin manufa yayin keta manufar sa.
Ƙimar ko samfurin ya canza hali lokacin da zai iya gaya masa ana gwada shi (sanarwar ƙima).
Gina kayan aikin sa ido ta yadda mutane masu rauni har yanzu zasu iya kula da ingantattun samfura akan ayyuka masu wahala.
Hatsari & Tsare-tsare
Magance haɗarin wanzuwa azaman sci-fi yayin da abubuwan iyawa.
Amintaccen samfur mai ruɗani tare da jeri ƙarƙashin babban ikon kai.
Barin waɗanda ba Ingilishi ba da ƙwararrun masu sauraro tare da tushe masu ƙarancin inganci kawai.
Taswirar Hanya
Rarrabe lahani na samfur, rashin amfani, da hasarar sarrafa-haɗari / rashin daidaituwa.
Tambayi wane shaida zai canza ra'ayin ku akan jerin lokuta da tsanani.
Fi son tushe na farko da tabbataccen kimantawa akan da'awar tallace-tallace.
Gano hanyar aiki ɗaya: aiki, manufa, kuɗi, ko ƙwarewa - ba kawai sani ba.
Ci gaba da Bincike
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 AI Alignment 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
Jagora na gaba
Glow-TTS Monotonic Daidaita
Tambayoyin da ake yawan yi
What is AI Alignment?
Aiwatar da AI shine aikin fasaha da cibiyoyi na samar da tsarin AI na ci gaba da dogaro da dogaro da abin da mutane suke niyya - gami da a cikin labari, yanayi mai girma inda tsarin ya fi wayo, sauri, ko kuma mai cin gashin kansa fiye da masu sarrafa shi.
Ta yaya ya kamata a bi da sirri da tsaro yayin tura AI Alignment?
Ana buƙatar gina sirri da tsaro cikin kowane jigilar AI Alignment daga farkon.
Menene hanyar da ke da alhakin magance rashin tabbas a cikin sakamako daga AI Alignment?
Gudanar da abubuwan da ba su da tabbas daga AI Alignment zuwa nazarin ɗan adam yana hana kurakuran da za a iya gujewa.
Kafin dogara da AI Alignment don yanke shawara mai mahimmanci, menene ya kamata ku tabbatar da farko?
Gudun gudu da goge baki ba sa tabbatar da daidaito. Ƙaddamar da Aiwatar da AI a cikin tabbataccen shaida shine abin da ke sa ya zama amintaccen dogaro da shi.
Menene alamar da ke nuna ƙungiyar ta fahimci AI alignment balagagge ba fiye da na zahiri ba?
Sanin iyakokin AI Alignment - inda ba shi da kyau - alama ce ta ainihin fahimta.
Wace rawa yakamata hukuncin ɗan adam ya taka yayin amfani da AI Alignment?
Tsayar da mutane a cikin madauki don mahimman lokuta ko rashin amincewa shine babban kariya tare da AI Alignment.