Riaan 
“I-AI literacy isiqala ukuba yisizinda esisha sokufunda izindaba,” kusho u-Worland, wengeza ngokuthi inhlangano yakhe iyakha. izinsiza zokusiza abantu bazulazule ezimangalweni ezididayo nezingqubuzanayo mayelana ne-AI.
Kusukela ezinjinini ze-chess kuya ku-Google translate, ubuhlakani bokwenziwa bukhona ngandlela thize kusukela maphakathi nekhulu lama-20. Kodwa kulezi zinsuku, ubuchwepheshe bukhula ngokushesha kunokuba abantu abaningi bangenza umqondo wabo, ochwepheshe bemininingwane engamanga bayaxwayisa. Lokho kushiya abantu abavamile besengozini yezimangalo ezidukisayo mayelana nokuthi amathuluzi e-AI angenziwa ini nokuthi ubani onesibopho ngomthelela wawo.
Ngokufika kwe-ChatGPT, i i-chatbot ethuthukisiwe kusukela kunjiniyela i-OpenAI, abantu baqale ukusebenzisana ngokuqondile namamodeli ezilimi ezinkulu, uhlobo lwesistimu ye-AI evame ukusetshenziswa ukunika amandla impendulo ezenzakalelayo kuma-imeyili, ukuthuthukisa imiphumela yosesho noma okuqukethwe okusesilinganisweni ezinkundleni zokuxhumana. Ama-Chatbots avumela abantu ukuthi babuze imibuzo noma batshele isistimu ukuthi ibhale yonke into kusukela ezinkondlweni kuya ezinhlelweni. Njengezinjini zokukhiqiza izithombe njengoDall-E futhi uthole ukuthandwa, amabhizinisi ayaphithizela engeza amathuluzi e-AI kanti othisha bakhathazekile ngokuthi bazohlonza kanjani Izabelo ezigunyazwe yi-AI.
Izikhukhula zolwazi olusha kanye nokuqagela okuzungeze i-AI kuphakamisa ubungozi obuhlukahlukene. Izinkampani zingase zidlulele kulokho okungenziwa futhi kusetshenziselwe amamodeli azo e-AI. Abasekeli bangase baphushe izindaba zezinganekwane zesayensi ezidonsela ukunaka kokuningi izinsongo ezisheshayo. Futhi amamodeli ngokwawo angase avuselele ulwazi olungalungile. Ulwazi oluyisisekelo lokuthi amamodeli asebenza kanjani – kanye nezinganekwane ezivamile mayelana ne-AI – luzodingeka ukuze uzulazule inkathi ezayo.
“Kufanele sihlakaniphe ngalokho lobu buchwepheshe obungakwazi ukukwenza noma obungakwazi ukukwenza, ngoba siphila ezikhathini ezinzima lapho ulwazi, ngeshwa, lusetshenziswa khona,” kusho uClaire Wardle, umqondisi-jikelele we-Information Futures Lab eBrown University, ecwaninga imininingwane engamanga. nokusabalala kwayo.
Ziningi izindlela zokuhlanekezela i-AI, kodwa amanye amafulegi abomvu avela ngokuphindaphindiwe. Nazi ezinye izicupho ezivamile okufanele uzigweme, ngokusho I-AI kanye nochwepheshe bolwazi lokufunda nokubhala.
Ungavezi izimfanelo zomuntu
Kulula ukuveza izimfanelo zomuntu kubantu abangebona abantu. (Ngithengele ikati lami isitokisi seholide ukuze lingazizwa lishiywe ngaphandle.)
Lokho kuthambekela, okubizwa ngokuthi i-anthropomorphism, kubangela izinkinga ezingxoxweni ezimayelana ne-AI, kusho uMargaret Mitchell, umcwaningi wokufunda ngomshini nososayensi omkhulu wezimiso zokuziphatha enkampanini ye-AI iHugging Face, futhi sekunesikhathi kuqhubeka.
Ngo-1966, usosayensi wekhompiyutha we-MIT ogama lakhe lingu-Joseph Weizenbaum wasungula i-chatbot ebizwa ngokuthi i-ELIZA, eyaphendula imilayezo yabasebenzisi ngokulandela umbhalo noma ngokuphindaphinda imibuzo yabo. I-Weizenbaum ithole ukuthi abantu bafaka imizwa nenjongo ku-ELIZA ngisho noma bazi ukuthi imodeli isebenza kanjani.
Njengoba ama-chatbot amaningi alingisa abangani, abelaphi, izithandwa nabasizi, izinkulumo-mpikiswano mayelana lapho inethiwekhi yekhompiyutha efana nobuchopho “iqaphela” izophazamisa izinkinga ezicindezelayo, kusho uMitchell. Izinkampani zingagwema umthwalo wemfanelo we-AI eyinkinga ngokuphakamisa ukuthi uhlelo luhambe kabi. Abantu bangakha ubudlelwano obungenampilo nezinhlelo ezilingisa abantu. Izinhlangano zingavumela uhlelo lwe-AI indlela eyingozi yokwenza amaphutha uma zilubheka njengelinye nje “ilungu labasebenzi,” kusho u-Yacine Jernite, ofunda ngomshini kanye nohola womphakathi kwaHugging Face.
Izinhlelo ze-AI zobuntu ziphinde ziqinise ukwesaba kwethu, futhi abantu abesabayo basengozini enkulu yokukholelwa nokusabalalisa okungalungile imininingwane, kusho uWardle waseBrown University. Sibonga ababhali bezinganekwane zesayensi, ubuchopho bethu bugcwele izimo ezimbi kakhulu, waphawula. Izindaba ezifana nethi “Blade Runner” noma “I-Terminator” samanje ikusasa lapho amasistimu e-AI eqaphela futhi avule abadali bawo abangabantu. Njengoba abantu abaningi bejwayelene kakhulu namamuvi e-sci-fi kunamanuances wezinhlelo zokufunda ngomshini, sivame ukuvumela imicabango yethu igcwalise izikhala. Ngokuqaphela i-anthropomorphism uma kwenzeka, u-Wardle uthe, singaqapha izinganekwane ze-AI.
Ungayibheki i-AI njenge-monolith
I-AI akuyona into eyodwa enkulu — iqoqo lobuchwepheshe obuhlukahlukene obuthuthukiswe abacwaningi, izinkampani kanye nemiphakathi eku-inthanethi. Izitatimende ezishanelayo nge-AI zivame ukufingqa imibuzo ebalulekile, kusho uJernite. Iyiphi imodeli ye-AI esikhuluma ngayo? Ubani owayakha? Ubani othola izinzuzo futhi ubani okhokha izindleko?
Amasistimu e-AI angenza kuphela lokho abadali bawo abakuvumelayo, kusho uJernite, ngakho-ke kubalulekile ukubamba izinkampani ziphendule ngendlela amamodeli azo asebenza ngayo. Isibonelo, izinkampani zizoba nemithetho ehlukene, izinto ezibalulekile kanye nezindinganiso ezithinta indlela imikhiqizo yazo esebenza ngayo emhlabeni wangempela. I-AI ayiqondisi imicibisholo noma idale izinqubo zokuqasha ezichemile. Izinkampani zenza lezo zinto ngosizo lwamathuluzi e-AI, kusho uJernite noMitchell.
“Ezinye izinkampani zinesandla ekwethuleni [AI models] njengalezi zidalwa zemilingo noma izinhlelo zemilingo ezenza izinto ongakwazi nokuzichaza,” kusho uJernite. “Bancika kulokho ukukhuthaza ukuhlolwa ngokucophelela kwalezi zinto.”
Kubantu abasekhaya, lokho kusho ukuphakamisa ishiya lapho kungacaci ukuthi ulwazi lwesistimu luvelaphi noma ukuthi isistimu iyikhiphe kanjani impendulo yayo.
Okwamanje, imizamo yokulawula i-AI iyaqhubeka. Kusukela ngo-Ephreli 2022, cishe ingxenye eyodwa kwezintathu yezifundazwe zase-US wayehlongoze noma wamisa okungenani umthetho owodwa ukuze avikele abathengi ekulimaleni okuhlobene ne-AI noma ekufinyeleleni ngokweqile.
Uma umuntu ehlanganisa umusho ohambisanayo, ngokuvamile asihlabeki umxhwele. Kodwa uma i-chatbot ikwenza, ukuzethemba kwethu emandleni e-bot kungase kunyuke kakhulu.
Lokho kubizwa nge-automation bias, futhi kuvame ukusiholela ekutheni sithembele kakhulu ezinhlelweni ze-AI, kusho uMitchell. Singenza okuthile okuphakanyiswa yisistimu ngisho noma kungalungile, noma sehluleke ukwenza okuthile ngenxa yokuthi isistimu ayizange ikuncome. Ngokwesibonelo, a 1999 isifundo bathole ukuthi odokotela abasebenzisa uhlelo lwe-AI ukusiza ukuxilonga iziguli bazokuziba ukuhlola kwabo okulungile bevuna iziphakamiso ezingalungile zesistimu ngamaphesenti angu-6 esikhathi.
Ngamafuphi: Ngenxa yokuthi imodeli ye-AI ingenza okuthile akusho ukuthi ingakwenza ngokungaguquki nangendlela efanele.
Njengoba kulinga ukuthembela kumthombo owodwa, njenge-bot yenjini yokusesha enikeza izimpendulo ezigayekayo, lawa mamodeli awasho njalo imithombo yawo futhi wenza izifundo zomgunyathi. Sebenzisa amakhono afanayo okufunda nokubhala kwabezindaba ongawasebenzisa esihlokweni se-Wikipedia noma usesho lwe-Google, kusho i-Worland ye-News Literacy Project. Uma ubuza injini yokusesha ye-AI noma i-chatbot, hlola izimpendulo ezikhiqizwe nge-AI ngokuphikisana nazo eminye imithombo ethembekile, njengamaphephandaba, amawebhusayithi kahulumeni noma enyuvesi noma amajenali ezemfundo.
