Ubuyobozi bwa tekiniki

Ikarita y'Icyitegererezo na Datasheets ya Datasets

Ikarita yicyitegererezo ninyandiko ngufi, yubatswe isobanura icyerekezo cya AI igamije, uko yasuzumwe, uko ikora neza mumatsinda atandukanye, kandi aho itagomba gukoreshwa.

Umva iki gitaboAI-byavuzwe · ~3 min

Kuriyi page3 min soma
  1. Incamake
  2. Kwibira cyane
  3. Gushyira mu bikorwa Isi
  4. Ejo hazaza h'amakarita yicyitegererezo na Datasheets ya Datasets
  5. Kubishyira mubikorwa
  6. Komeza Ubushakashatsi
  7. Ibibazo bikunze kubazwa

Incamake

Urupapuro rwerekana imibare ikora kimwe kubijyanye namakuru yamahugurwa: impamvu amakuru yakusanyijwe, ibiyirimo, uko yakusanyirijwe hamwe kandi yanditseho, nuburyo agomba kubungabungwa. Izi nyandiko zifite akamaro kuko zemerera abaguzi, abakoresha, abagenzuzi n’abagenzuzi gusuzuma niba sisitemu ijyanye n’intego, kandi amategeko nk’amategeko y’ubumwe bw’ibihugu by’Uburayi arasaba ibyangombwa bisa.

Kwibira cyane

Ikarita y'icyitegererezo yatanzwe na Margaret Mitchell na bagenzi be mu mpapuro "Ikarita y'icyitegererezo yo gutanga raporo y'icyitegererezo", yerekanwe mu nama ya ACM FAT * mu mwaka wa 2019. Ibice byatanzwe ni ibisobanuro by'icyitegererezo, imikoreshereze yabigenewe, ibintu (nk'imiterere y'abaturage cyangwa ibidukikije), ibipimo, amakuru yo gusuzuma, amakuru y'amahugurwa, isesengura ry'imibare, gutekereza ku myitwarire, hamwe na caveats n'ibyifuzo. Igitekerezo cyingenzi nisuzumabumenyi ridatandukanijwe: gutanga raporo kubikorwa bitandukanye kubitsinda rito aho kuba nk'impuzandengo imwe ishobora guhisha kunanirwa. Datasheets ya Datasets, na Timnit Gebru na bagenzi be, yakwirakwijwe bwa mbere muri 2018 nyuma isohoka mu Itumanaho rya ACM. Itira inguzanyo za elegitoroniki. Abakora Dataset basubiza ibibazo bijyanye na motifation, ibihimbano, inzira yo gukusanya, gutunganya mbere na label, gukoresha, gukwirakwiza no kubungabunga. Imiterere ijyanye nayo ikubiyemo amakuru yamakuru yamakuru yamakarita namakarita yamakuru. Ikarita ya sisitemu, yakoreshejwe na Meta na OpenAI kuva mu 2022 kandi ikazwi cyane kuri GPT-4 mu 2023, yerekana sisitemu yose yoherejwe, harimo gupima umutekano na mituweli, aho kuba icyitegererezo gusa. Amabwiriza yahinduye byinshi mubikorwa mubisabwa. Itegeko ry’ubumwe bw’ibihugu by’Uburayi risaba ibyangombwa bya tekiniki kuri sisitemu zo mu bwoko bwa AI zifite ibyago byinshi, bigaragara ku mugereka wa IV. Abatanga icyitegererezo rusange cya AI bagomba kubika ibyangombwa bya tekiniki, gutanga amakuru kubatanga isoko yo hasi, no gutangaza incamake yibirimo byamahugurwa ukoresheje icyitegererezo cya Komisiyo. Igitekerezo gikocamye ni uko ikarita yicyitegererezo yerekana ko icyitegererezo gifite umutekano. Nukumenyekanisha, ntabwo ari icyemezo, kandi ikarita irashobora kuba ituzuye, itajyanye n'igihe cyangwa yanditse nko kwamamaza. Agaciro kayo gashingiye kumipaka no kwisuzumisha undi muntu ashobora kubyara.

Gushyira mu bikorwa Isi

Ibitaro bisuzuma ibyiciro byuruhu-ibisebe bisoma ikarita yicyitegererezo kandi ugasanga byavuzwe neza ukurikije imiterere yuruhu. Hanyuma ihitamo gukora ibyayo byemewe mbere yo gukoresha.

Ububiko bw'icyitegererezo bwa Hugging Face bwerekana ikarita y'icyitegererezo hamwe na YAML metadata yerekana uruhushya rwayo, amahugurwa yamakuru, indimi n'ibisubizo byerekana, kandi abantu barashobora gushungura ibisubizo by'ishakisha kuri iyo mirima.

Itsinda ry’ubushakashatsi risohora imibare yamakuru hamwe nurupapuro rwerekana uruhushya rwabavuga, imiterere yo gufata amajwi hamwe ninshusho zidahagarariwe.

Laboratwari ya AI irekura ikarita ya sisitemu yo kuganira isobanura ibisubizo bitukura-bitukura, kugabanya umutekano hamwe ningaruka zisigaye, birenze ibyo ikarita y'ibanze ikubiyemo.

Ejo hazaza h'amakarita yicyitegererezo na Datasheets ya Datasets

Inyandiko ziva mubikorwa byubushake bigana ku kubahiriza ibisabwa. Amategeko y’ubumwe bw’ibihugu by’Uburayi, amategeko agenga amasoko n’ibipimo nka ISO / IEC 42001 byose biteze inyandiko zubatswe zerekana uburyo sisitemu yubatswe kandi ikageragezwa. Imashini isomeka imashini nibisekuruza byikora birashoboka cyane, kugirango amakarita asuzumwe kandi agereranwe. Inzitizi zikomeje ni ugukomeza inyandiko neza nkuko moderi zivugururwa, guhishura bihagije kubyerekeye amahugurwa atagaragaje amabanga yubucuruzi cyangwa amakuru yihariye, no kwemeza ko abasomyi bakoresha inyandiko mugihe bafata icyemezo.

Kubishyira mubikorwa

Impamvu ari ngombwa

  • Igiciro na bije. Ibyemezo byubwubatsi bitwara imikorere nigiciro cyimikorere kumyaka.
  • Ibyemezo bisobanutse. Ubuhanga bwa tekinike bufasha amakipe guhitamo umurongo ukwiye, ntabwo ari shyashya gusa.
  • Kugenzura ubuziranenge. Guhitamo neza bya injeniyeri bigabanya ibintu byizewe mubikorwa.

Witondere

  • Gutezimbere igipimo kimwe gishobora guhisha intege nke za sisitemu.
  • Ibikorwa Remezo no kubungabunga akenshi usanga bidahabwa agaciro.
  • Icyuho cyumutekano no kwitegereza birashobora kwiyongera uko sisitemu igenda igorana.

Intambwe ikurikira

  1. Sobanura ubukererwe, ubuziranenge, nigiciro cyibiciro mbere yo kubishyira mubikorwa.
  2. Ibipimo byerekana umutwaro ufatika hamwe namakuru yimiterere.
  3. Gukurikirana ibikoresho kubikosa, drift, ningaruka zabakoresha.
  4. Tegura inzira yo gusubiza ibyabaye mbere yo gupima.

Komeza Ubushakashatsi

Ibaruwa y'amakuru y'ubuntu

Ibuka amakuru ya buri munsi ya AI

Inkuru eshatu zizewe za AI buri gitondo cy'iminsi y'akazi, zanditse mu rurimi rworoshye. Ubuntu iteka ryose, nta kwamamaza.

Imeri imwe buri munsi y'akazi. Kwiyimura mu buryo bumwe. Ntitwigera tugurisha cyangwa ngo dusangire aderesi yawe.

Isuzume

Take the Model Cards and Datasheets for Datasets quiz

Ibisubizo ako kanya kuri buri kibazo, n'icyemezo gishobora gusangirwa gifite ID yagenzurwa umaze gutsinda isomo.

Tangira ikibazo

Fasha uburezi bwa AI bw'ubuntu. AI Understanding ni umuryango utegamiye kuri leta wa 501(c)(3) — nta kwamamaza, nta kwishyura, iteka ryose. Tanga impano

Ibibazo bikunze kubazwa

Ikarita y'Icyitegererezo na Datasheets ya Datasets ni iki?

Ikarita yicyitegererezo ninyandiko ngufi, yubatswe isobanura icyerekezo cya AI igamije, uko yasuzumwe, uko ikora neza mumatsinda atandukanye, kandi aho itagomba gukoreshwa. Urupapuro rwerekana imibare ikora kimwe kubijyanye namakuru yamahugurwa: impamvu amakuru yakusanyijwe, ibiyirimo, uko yakusanyirijwe hamwe kandi yanditseho, nuburyo agomba kubungabungwa. Izi nyandiko zifite akamaro kuko zemerera abaguzi, abakoresha, abagenzuzi n’abagenzuzi gusuzuma niba sisitemu ijyanye n’intego, kandi amategeko nk’amategeko y’ubumwe bw’ibihugu by’Uburayi arasaba ibyangombwa bisa.

Niyihe ntego nyamukuru yikarita yicyitegererezo?

Ikarita yicyitegererezo iha abasomyi ibyo bakeneye kugirango bamenye niba icyitegererezo gihuye nikoreshwa runaka, harimo n’aho gikora nabi.

Ninde wayoboye impapuro zerekana amakarita yicyitegererezo?

Margaret Mitchell na bagenzi be berekanye "Ikarita y'Icyitegererezo cyo Gutanga Icyitegererezo" muri ACM FAT * muri 2019.

Isuzuma ritemewe ni iki?

Gutandukanya ibisubizo byitsinda ryerekana kunanirwa impuzandengo rusange yihisha.

Datasheets ya Datasets itira igitekerezo cyayo kuva murwego ki?

Ibikoresho bya elegitoronike bizana imibare isobanura ibiranga. Gebru na bagenzi be bakoresheje icyo gitekerezo kuri datasets.

Ni gute ikarita ya sisitemu itandukanye n'ikarita y'ibanze y'icyitegererezo?

Ikarita ya sisitemu, nka OpenAI ya GPT-4, isobanura ingaruka zoherejwe ku bicuruzwa, guhuriza hamwe umutuku no kurinda.