I-Artificial Intelligence (AI) ivame ukubonwa njengobuchwepheshe obushintshashintshayo, obukwazi ukuletha ukusebenza kahle, ukunemba, nokuvula amathuba amasha wamasu. Kodwa-ke, njengoba izinkampani zizuza ezinhlelweni ze-AI, kuvela inselelo ebucayi nevame ukunganakwa: ukulunga kwe-algorithmic. Ukuchema okufihliwe kulezi zinhlelo angeke kubeke engcupheni ukusebenza kahle kwezinqumo zebhizinisi kuphela kodwa futhi kudale imiphumela ebalulekile yezomthetho, yezimiso zokuziphatha, nezenhlalo.
Ukuba khona kokuchema kwe-algorithmic kungachazwa uhlobo lwe-AI ngokwayo, ikakhulukazi ekufundeni komshini. Amamodeli aqeqeshwa ngedatha yomlando, futhi uma le datha ibonisa ukucwasa noma ukuchema komphakathi, ama-algorithms ngokwemvelo agcina eqhubekisela phambili lokhu kuchema. Ngaphezu kokwenzelela olwazini, i-algorithm ngokwayo ingathula ukungalingani esikalini sezici ezenziwe, noma kudatha esetshenziswa njengommeleli—okungukuthi, idatha ethatha indawo yolwazi lwangempela kodwa engalungile kulokho kuhlaziya.
Isibonelo esiwuphawu salesi sigameko sitholakala ekusetshenzisweni kokubonwa kobuso, ikakhulukazi ezimweni ezibucayi njengokuphepha komphakathi. Amadolobha amaningana ase-Brazil amukele amasistimu azenzakalelayo okwandisa ukusebenza kahle kwezenzo zamaphoyisa, kodwa ukuhlaziya kubonisa ukuthi lawa ma-algorithms ngokuvamile enza amaphutha abalulekile, ikakhulukazi lapho kuhlonzwa abantu bezinhlanga ezithile, njengabantu Abansundu. Ucwaningo olwenziwe umcwaningi we-MIT uJoy Buolamwini lubonise ukuthi ama-algorithms okuhweba anamazinga amaphutha angaphezu kuka-30% kwabesifazane abamnyama, kanti kwabesilisa abamhlophe, izinga lehla kakhulu libe ngaphansi kuka-1%.
Umthetho waseBrazil: ukuqina okwengeziwe esikhathini esizayo
E-Brazil, ngaphezu kwe-General Data Protection Law (LGPD), Uhlaka Lwezomthetho Lwe-AI (Umthethosivivinywa No. 2338/2023) nalo luyaqhubeka, olusungula imihlahlandlela evamile yokuthuthukiswa nokusetshenziswa kwe-AI ezweni.
Nakuba ungakagunyazwa, lo mthethosivivinywa usuvele uveza amalungelo okufanele izinkampani ziwahloniphe, njengalokhu: ilungelo lokuthola ulwazi lwangaphambili (ukwazisa lapho umsebenzisi esebenzisana nesistimu ye-AI), ilungelo lencazelo yezinqumo ezizenzakalelayo, ilungelo lokubekela inselele izinqumo ze-algorithmic, kanye nelungelo lokungabandlululi ngenxa yokuchema kwe-algorithmic.
Lawa maphuzu azodinga ukuthi izinkampani zisebenzise ukubonisa izinto obala ezinhlelweni ezikhiqizayo ze-AI (isb., ukwenza kucace lapho umbhalo noma impendulo yenziwe ngomshini) kanye nezindlela zokuhlola ukuze zichaze ukuthi imodeli ifike kanjani ekuphumeni okunikeziwe.
Ukubusa Kwe-algorithmic: Isixazululo Sokuchema
Ezinkampanini, ukuchema kwe-algorithmic kudlulela ngale komkhakha wezimiso zokuziphatha futhi kuba yizinkinga ezibalulekile zamasu. Ama-algorithms achemile anamandla okuhlanekezela izinqumo ezibalulekile ezinqubweni zangaphakathi ezinjengokuqasha, ukunikezwa kwezikweletu, nokuhlaziywa kwemakethe. Isibonelo, i-algorithm yokuhlaziya ukusebenza kwegatsha elinganisela ngokweqile izifunda zasemadolobheni ngokulimala kwezifunda eziseduze (ngenxa yedatha engaphelele noma ukuchema) ingaholela ekutshalweni kwezimali okungaqondiswanga kahle. Ngakho-ke, ukuchema okucashile kubukela phansi ukusebenza kahle kwamasu aqhutshwa yidatha, okwenza abaphathi benze izinqumo ezisekelwe olwazini olungalungile ngokwengxenye.
Lokhu kuchema kungalungiswa, kodwa kuzoncika esakhiweni sokuphatha se-algorithmic, okugxile ezinhlobonhlobo zedatha esetshenzisiwe, ukucaca kwezinqubo, kanye nokufakwa kwamaqembu ahlukahlukene kanye nemikhakha eminingi ekuthuthukisweni kobuchwepheshe. Ngokutshala imali ezinhlobonhlobo emaqenjini ezobuchwepheshe, isibonelo, izinkampani zingakwazi ukuhlonza ngokushesha imithombo engase ibe khona yokuchema, ziqinisekise ukuthi kucatshangelwa imibono ehlukene nokuthi amaphutha atholwa kusenesikhathi.
Ngaphezu kwalokho, ukusetshenziswa kwamathuluzi okuqapha okuqhubekayo kubalulekile. Lawa masistimu asiza ukuthola ukuchema kwe-algorithmic ngesikhathi sangempela, anikeze amandla ukulungiswa okusheshayo futhi anciphise umthelela ongemuhle.
Ukwenza izinto obala kungomunye umkhuba obalulekile ekwehliseni ukuchema. Ama-algorithms akufanele asebenze njengamabhokisi amnyama, kodwa njengamasistimu acacile futhi achazekayo. Lapho izinkampani zikhetha ukubeka izinto obala, zithola ukwethenjwa ngamakhasimende, abatshalizimali, nabalawuli. Ukungafihli kusiza ukucwaningwa kwamabhuku kwangaphandle, kukhuthaza isiko lokuhlanganyela umthwalo wemfanelo ekuphathweni kwe-AI.
Ezinye izinhlelo zihlanganisa ukubambelela kuzinhlaka kanye nezitifiketi zokubusa kwe-AI okunesibopho. Lokhu kubandakanya ukudala amakomiti okuziphatha e-AI yangaphakathi, ukuchaza izinqubomgomo zebhizinisi zokusetshenziswa kwayo, nokwamukela izindinganiso zamazwe ngamazwe. Isibonelo, izinhlaka ezifana ne-ISO/IEC 42001 (ukuphathwa kobuhlakani bokwenziwa), ISO/IEC 27001 (ukuphepha kolwazi), kanye ne-ISO/IEC 27701 (ubumfihlo) bokulawula ukwakheka ezinqubweni zedatha ezisetshenziswa i-AI yokukhiqiza. Esinye isibonelo isethi yezinqubo ezinconyiwe yi-US National Institute of Standards and Technology (NIST) eqondisa ukuphathwa kwezingcuphe kwe-algorithmic, okuhlanganisa ukutholwa kokuchema, ukuhlolwa kwekhwalithi yedatha, nokuqapha imodeli okuqhubekayo.
Oxhumana nabo abakhethekile badlala indima yamasu kulesi simo. Ngobuchwepheshe bobuhlakani bokwenziwa obunesibopho, ukubusa kwe-algorithmic, nokuthobela imithetho, lezi zinkampani zisiza izinhlangano zingagcini nje ngokugwema ubungozi kodwa futhi ziguqule ukulingana kube inzuzo yokuncintisana. Umsebenzi walaba basebenzi bezokuxhumana usukela ekuhloleni ubungozi okunemininingwane kuye ekuthuthukisweni kwezinqubomgomo zangaphakathi kanye nokuqeqeshwa kwenkampani ngokuziphatha kwe-AI, okuqinisekisa ukuthi amathimba akulungele ukuhlonza nokunciphisa ukuchema okungaba khona kwe-algorithmic.
Ngakho-ke, ukunciphisa ukuchema kwe-algorithmic akuyona nje indlela yokuvimbela, kodwa indlela yamasu. Izinkampani ezigxile ekulungeni kwe-algorithmic zibonisa isibopho emphakathini, ziqinisa isithunzi sazo, futhi zizivikele ekujezisweni kwezomthetho nasezinkingeni zomphakathi. Ama-algorithms angachemile athambekele ekunikezeni ukuqonda okunembe kakhudlwana nokulinganiselayo, okwandisa ukusebenza kahle kwezinqumo zebhizinisi nokuqinisa isikhundla sokuncintisana sezinhlangano emakethe.
Ngu-Sylvio Sobreira Vieira, i-CEO kanye ne-Head Consulting kwa-SVX Consultoria
