Riaan Ubuchwepheshe be-Artificial intelligence (AI) bebuhamba phambili ekusunguleni izinto ezintsha ngeminyaka yawo-2020. Isofthiwe ekhiqizayo efana ne-ChatGPT ne-Midjourney icindezela imingcele yokudala, ngokungcono noma kokubi.
Ngakolunye uhlangothi, lawa mathuluzi anikeza umhlaba wamandla emikhakheni efana nokukopisha, ukwakheka kwezithombe, nocwaningo. Ngakolunye uhlangothi, zingaba umthombo wezinkinga ezimbalwa futhi, kusukela ekusabalaliseni ulwazi olunganembile kuye ekwephuleni ubumfihlo bomsebenzisi.
Njengoba lobu buchwepheshe budlanga kakhulu, kunesidingo esicindezelayo sokuchaza imingcele lapho amathuluzi e-AI avunyelwe ukusebenza khona. Nakhu ukubuka izindlela ezihlukene lapho abalawuli bomhlaba wonke babhekana khona nesivinini esisheshayo sokusungula izinto ezintsha kanye nokubeka ama- guardrails.
Izinkinga zobumfihlo bedatha
Amasistimu e-AI athembele kudatha ukuze athuthukise ukuphuma kwawo. Isibonelo, i-OpenAI – abadali be-ChatGPT – baqeqesha amathuluzi abo e-AI olwazini olubanzi olutholakala ku-inthanethi. Lokhu kungafaka idatha yomuntu siqu etholwe ngaphandle kwemvume yabantu. I-ChatGPT iphinde isindise konke ukwaziswa kwayo futhi le datha ingase isetshenziselwe ukuqeqeshwa okwengeziwe.
Ngabantu abangaphezu kwezigidi eziyi-100 abasebenzisa inkundla, i-OpenAI inomthwalo wemfanelo omkhulu wokuvikela idatha. Ngenyanga edlule, inkampani yenza iphutha futhi inkinga yavumela abasebenzisi ukuthi babuke umlando wengxoxo yabanye abantu ku-ChatGPT. Uma kuyekwa kungagadiwe, ukugada okunjalo kungaba nemiphumela ekhathazayo.
Kunengqondo ukuthi lawa mathuluzi e-AI akumele atholakale emphakathini kuze kube yilapho ukuhlolwa okuphelele kungaqinisekisa ukuthi noma iyiphi idatha yomsebenzisi eqoqiwe isezandleni eziphephile.
Kulandela ukwephulwa kwedatha okukhulunywe ngakho ngenhla, i-Italy ibe izwe lokuqala laseYurophu ukuvala i-ChatGPT. I-Italian Data Protection Authority iveze ukukhathazeka ngokuqoqwa kwedatha yomuntu siqu, kanye nokuntuleka kokuqinisekiswa kweminyaka yobudala – okungase kuvumele izingane ukuthi zifinyelele okuqukethwe okungafanele.
Amanye amazwe ase-Europe – wonke alandela I-General Data Protection Regulation (GDPR), enikeza abasebenzisi ukulawula okukhulu ngosuku lwabo lomuntu siqu – angase alandele okufanayo. Iziphathimandla zaseJalimane nase-Ireland kubikwa ukuthi zixoxisana ne-Italy mayelana nomnyakazo wazo wokuvala i-ChatGPT.
E-US, abalawuli badingida imithetho emisha yokuphatha isoftware ekhiqizayo ye-AI. Umnyango wezohwebo wezwe ubheke ukusungula izinqubomgomo ezisekela ukucwaningwa kwamabhuku kanye nokuhlolwa kwengozi okungahlola ukuthi uhlelo lwe-AI lungathenjwa yini.
I-China iphinde yathatha izinyathelo zayo zokuqala, yagunyaza ukuthi isoftware ekhiqizayo ye-AI iphase ukuhlolwa kokuphepha kukahulumeni ngaphambi kokuba yethulwe emphakathini.
Ukwakha i-AI yokuziphatha
Njengoba amasistimu e-AI esondela ekuhlakanipheni okufana nomuntu, ngeke esakwazi umgomo ngokuphelele. Uma bechaza izithombe noma bebhala izinto zokudala, bayaphoqeleka ukuthi baveze ukuchema kwabo. Lokhu kuvela njengomphumela wedatha lapho i-AI iqeqeshwa khona.
Esikhathini esedlule, i-ChatGPT itholakale ikhuluma amazwi alumelayo lapho icelwa ukuba isho amahlaya ngezihloko ezithile. I-OpenAI isebenze ukulungisa izimo ezinjalo futhi ngokwenza kanjalo, yakha ikhampasi ye-software yabo yokuziphatha.
E-Singapore, i-Personal Data Protection Commission (PDPC) kanye ne-Infocomm Media Development Authority (IMDA) bathuthukise ithuluzi lokuhlola ukulunga kwamathuluzi e-AI ngokwezimiso zokuziphatha ze-AI ezamukelwa emhlabeni wonke. Njengamanje, ukusetshenziswa kwalokhu kungokuzithandela futhi izinkampani ziyamenywa ukuthi ziveze obala mayelana nokuthunyelwa kwazo kwe-AI.
I-UK iphinde itshale izigidi zamadola ekuthuthukiseni izinhlelo zokusebenza ze-AI zokuziphatha, ikakhulukazi emkhakheni wezezimali.
Ngasekupheleni kuka-2022, i-White House yashicilela ipulani ye-AI Bill of Rights e-US. Phakathi kwamaphuzu amahlanu abalulekile ukuvikela ekubandlululweni kwe-algorithmic.
Lokhu kusho ukuthi abaklami nabathuthukisi badinga ukuthatha izinyathelo zokuqinisekisa ukuthi amasistimu abo e-AI awahloniphi abantu ngokusekelwe ezicini ezifana nobuhlanga, ubuzwe, noma ubulili. Njengoba kunje, leli pulani alisebenziseki ngokomthetho, kodwa lingaphakamisa uphawu lwezinqubomgomo ezizayo.
I-European Commission ibisebenza ngohlaka olungokomthetho lokuqala ngqa lwe-AI, oluhlukanisa izinhlelo ngokwesisekelo sezinga labo lobungozi. Njengoba kunje, umthetho we-AI osalungiswa uzobeka amathuluzi e-AI akhiqizayo kusethi yezinyathelo ezithathwa njengezifanele izinhlelo ‘ezinobungozi obukhulu’.
Ukwehlisa ijubane ukuqamba okusha
Yize isidingo sokulawulwa kwe-AI sisobala, indlela okwenziwa ngayo ingase iphazamise inqubo yokusungula izinto ezintsha.
Isibonelo, ukuhlukanisa amathuluzi e-AI akhiqizayo ‘njengengcuphe ephezulu’ kungabafaka ezindleleni zokuthobela ezibizayo. Ngenxa yalokho, bangase bangakwazi ukunikeza izinkonzo zabo mahhala. Izinkokhelo zinganciphisa isisekelo sabasebenzisi futhi lokhu kunganciphisa idatha etholakalayo yokuqeqesha i-AI.
Empeleni, imithethonqubo ingaba nomthelela ongemuhle izinga amathuluzi anjalo e-AI angafunda futhi athuthuke ngalo.
Uma sekushiwo, ukukhula kancane akusho ukuthi kuyinsalela. Kunganikeza izinkampani isikhathi esengeziwe sokuqinisekisa ukuthi amathuluzi azo aphephile ukuwasebenzisa futhi nedatha yomsebenzisi ivikelekile. Kungaba namathuba amaningi okukhipha iziphazamisi ngaphambi kokuthola ukutholwa ngobuningi – ngaleso sikhathi, amaphutha amancane kakhulu ekhodi angaba nomthelela omubi kakhulu.
Njengoba kuvame ukuba njalo, abalawuli bomhlaba wonke bayekwa ukuze bathole ibhalansi efanele phakathi kokusungula izinto ezintsha nokuphepha.
Ikhredithi Yesithombe Esifakiwe: I-Analytics Insight
