Ukuthatha Indlela Enengqondo ku-AI Yehlisa Kanjani Kumasethi Wedatha Achemile

Ama-Datasets akhethiwe kanye ne-AI Yokuziphatha

Izixazululo ezinikwe amandla yi-AI zidinga amasethi wedatha ukuze zisebenze kahle. Futhi ukudalwa kwalawo masethi wedatha kugcwele inkinga yokuchema engacacile ezingeni elihlelekile. Bonke abantu bahlushwa ukucwasa (bobabili bayazi futhi abazi). Ukuchema kungase kuthathe noma yiziphi izinhlobo zezinhlobo: ngokwendawo, ngokolimi, kwezenhlalo-mnotho, ngokobulili, nokucwasa ngokwebala. Futhi lokho kubandlulula okuhlelekile kubhakwa kudatha, okungaholela emikhiqizweni ye-AI eqhubekisela phambili futhi ikhulise ukuchema. Izinhlangano zidinga indlela enengqondo yokunciphisa ukuchema okungena eqoqweni ledatha.

Izibonelo Ezibonisa Inkinga Yokuchema

Isibonelo esisodwa esiphawulekayo salokhu kukhetha kwesethi yedatha okwazuza imibiko eminingi engemihle ngaleso sikhathi kwakuyisixazululo sokufunda kabusha esasivuna abantu besilisa kunabesifazane. Lokhu kungenxa yokuthi amasethi edatha ethuluzi lokuqasha athuthukiswe kusetshenziswa ukuqalisa kabusha kusukela eminyakeni eyishumi edlule lapho iningi labafake izicelo bekungamadoda. Idatha ibifanele futhi imiphumela ikhombise lokho kukhetha. 

Esinye isibonelo esibikwe kabanzi: Engqungqutheleni yaminyaka yonke yonjiniyela be-Google I/O, i-Google yabelane ngokubuka kuqala kwethuluzi losizo lesikhumba elinikwe amandla yi-AI elisiza abantu baqonde ukuthi kwenzakalani ngezinkinga eziphathelene nesikhumba sabo, izinwele, nezinzipho. Umsizi we-dermatology ugcizelela ukuthi i-AI ithuthuka kanjani ukuze isize ekunakekelweni kwezempilo - kodwa iphinde yagqamisa amandla okuthi ukwenzelela kungene ku-AI ngemuva kokugxekwa ngokuthi leli thuluzi alibanele kubantu bemibala.

Lapho i-Google imemezela ithuluzi, inkampani yaphawula:

Ukuqinisekisa ukuthi sakhela wonke umuntu, imodeli yethu ilandisa ngezinto ezifana neminyaka, ubulili, uhlanga, nezinhlobo zesikhumba - kusukela esikhumbeni esiphaphathekile esingashintshisi ukuya kwesikhumba esinsundu esingavamisile ukusha.

I-Google, Ukusebenzisa i-AI ukusiza ukuthola izimpendulo ezimweni ezivamile zesikhumba

Kepha i-athikili ku-Vice ithe i-Google yehlulekile ukusebenzisa isethi yedatha ebandakanyayo:

Ukufeza lo msebenzi, abacwaningi basebenzise idathasethi yokuqeqeshwa yezithombe ezingama-64,837 zeziguli eziyi-12,399 ezitholakala ezifundazweni ezimbili. Kodwa ezinkulungwaneni zezimo zesikhumba ezibonisiwe, amaphesenti angu-3.5 kuphela avela ezigulini ezinezinhlobo zesikhumba ze-Fitzpatrick V no-VI-lezo ezimele isikhumba esinsundu kanye nesikhumba esimnyama noma esimnyama, ngokulandelana. Amaphesenti angama-90 esizindalwazi akhiwe ngabantu abanesikhumba esimhlophe, isikhumba esimhlophe esimnyama, noma isikhumba esinsundu ngokukhanyayo, ngokusho kocwaningo. Njengomphumela wesampuli esichemile, odokotela besikhumba bathi uhlelo lokusebenza lungagcina ludlulile- noma lungaxilongi kahle abantu abamhlophe.

Vice, Uhlelo Lokusebenza Lwesikhumba Olusha lwe-Google Aludizayinelwe Abantu Abanesikhumba Esimnyama

I-Google iphendule ngokuthi izocwengisa ithuluzi ngaphambi kokulikhulula ngokusemthethweni:

Ithuluzi lethu le-AI-powered dermatology assist isiphetho seminyaka engaphezu kwemithathu yocwaningo. Njengoba umsebenzi wethu ufakwe ku-Nature Medicine, siye saqhubeka nokuthuthukisa kanye nokwenza ngcono ubuchwepheshe bethu ngokufaka amasethi edatha engeziwe afaka idatha enikelwe izinkulungwane zabantu, kanye nezigidi zezithombe zokukhathazeka zesikhumba ezikhethiwe.

I-Google, Ukusebenzisa i-AI ukusiza ukuthola izimpendulo ezimweni ezivamile zesikhumba

Ngendlela esingethemba ngayo ukuthi i-AI nezinhlelo zokufunda ngomshini zingalungisa lokhu kucwasana, iqiniso lihlala likhona: zinjengo smart njengoba amasethi abo edatha ehlanzekile. Ekuvuseleleni i-adage yohlelo oludala udoti ngaphakathi/udoti uphumile, Izixazululo ze-AI ziqine kuphela njengekhwalithi yamasethi azo edatha kusukela lapho zisuka khona. Ngaphandle kokulungiswa okuvela kohlelo, lawa masethi wedatha awanawo ulwazi lwangemuva lokuzilungisa - ngoba awanaso esinye isethenjwa.

Ukwakha amasethi edatha ngokuzibophezela kuwumongo wawo wonke ubuhlakani bokwenziwa bokuziphatha. Futhi abantu basemongweni wesixazululo. 

I-Mindful AI i-Ethical AI

Ukuchema akwenzeki endaweni engenalutho. Amasethi edatha angalungile noma achemile avela ekuthatheni indlela engalungile phakathi nesigaba sokuthuthukiswa. Indlela yokulwa namaphutha okuchema iwukusebenzisa indlela enesibopho, egxile kubantu abaningi embonini abayibiza nge-Mindful AI. I-Mindful AI inezingxenye ezintathu ezibalulekile:

1. I-AI Enengqondo Igxile Kubantu

Kusukela ekuqaleni kwephrojekthi ye-AI, ezigabeni zokuhlela, izidingo zabantu kufanele zibe maphakathi kuzo zonke izinqumo. Futhi lokho kusho bonke abantu - hhayi nje isethi encane. Yingakho abathuthukisi bedinga ukuthembela eqenjini elihlukahlukene labantu abasuselwa emhlabeni jikelele ukuqeqesha izinhlelo ze-AI ukuthi zibandakanyeke futhi zingabi namkhethe.

I-Crowdsourceing amasethi edatha evela eqenjini lomhlaba wonke, elihlukahlukene iqinisekisa ukuchema kuyahlonzwa futhi kuhlunga kusheshe. Labo bezinhlanga ezihlukene, amaqembu obudala, ubulili, amazinga emfundo, izizinda zezenhlalo-mnotho, nezindawo bangabona kalula amasethi edatha athanda isethi yamanani ngaphezu kwenye, ngaleyo ndlela kukhishwe ukuchema okungahlosiwe.

Bheka izinhlelo zokusebenza zezwi. Lapho kusetshenziswa indlela ye-AI enengqondo, nokusebenzisa amandla echibi lamathalente lomhlaba wonke, abathuthukisi bangakwazi ukulandisa ngezakhi zolimi njengezilimi zesigodi ezahlukene kanye nezimpawu zokuphimisela kumasethi wedatha.

Ukusungula uhlaka lokuklama olugxile kumuntu kusukela ekuqaleni kubalulekile. Kuhamba ibanga elide ekuqinisekiseni ukuthi idatha ekhiqiziwe, ekhethiwe, futhi ilebula ihlangabezana nokulindelwe ngabasebenzisi bokugcina. Kodwa futhi kubalulekile ukugcina abantu benolwazi kuwo wonke umjikelezo wokuphila wokuthuthukiswa komkhiqizo. 

Abantu abakulophu bangasiza futhi imishini ukuthi idale umuzwa ongcono we-AI kuzithameli ezithile. E-Pactera EDGE, amaqembu ethu ephrojekthi yedatha ye-AI, atholakala emhlabeni wonke, ayaqonda ukuthi amasiko nezimo ezihlukene zingaba nomthelela kanjani ekuqoqweni nasekugcinweni kwedatha ethembekile yokuqeqeshwa kwe-AI. Banamathuluzi adingekayo abawadingayo ukuze bahlabe umkhosi izinkinga, baziqaphe, futhi bazilungise ngaphambi kokuba isisombululo esisekelwe ku-AI sibe bukhoma.

I-Human-in-the-loop AI iphrojekthi "inethi yokuphepha" ehlanganisa amandla abantu - kanye nezizinda zabo ezihlukene ngamandla ekhompyutha ekhompyutha asheshayo. Lokhu kubambisana kwabantu ne-AI kudinga ukusungulwa kusukela ekuqaleni kwezinhlelo ukuze idatha echemile ingenzi isisekelo kuphrojekthi. 

2. I-AI Eqondayo Iyabhekela

Ukuba nesibopho wukuqinisekisa ukuthi izinhlelo ze-AI zokungachemi nokuthi zisekelwe ezimisweni zokuziphatha. Kumayelana nokukhumbula ukuthi idatha yenziwa kanjani, ngani, futhi kuphi, ukuthi ihlanganiswa kanjani izinhlelo ze-AI, nokuthi isetshenziswa kanjani ekwenzeni isinqumo, izinqumo ezingaba nemithelela yokuziphatha. Enye indlela ibhizinisi elingenza ngayo lokho ukusebenzisana nemiphakathi engameleli kahle ukuze ihlanganise wonke umuntu futhi ingachemi. Emkhakheni wezichasiselo zedatha, ucwaningo olusha lugqamisa ukuthi imodeli yemisebenzi eminingi enezichasiselo eziningi ephatha amalebula esichasiseli ngasinye njengomsebenzi ongaphansi ohlukene ingasiza kanjani ekwehliseni izinkinga ezingaba khona ezindleleni zeqiniso eliyisisekelo lapho ukungaboni ngaso linye kwabachasi kungase kube ngenxa yokumelela kancane futhi ingashaywa indiva ekuhlanganisweni kwezichasiselo zeqiniso elilodwa eliyisisekelo. 

3. Uthembekile

Ukuthembeka kuvela ekutheni ibhizinisi libe sobala futhi lichazeke ngendlela imodeli ye-AI eqeqeshwa ngayo, isebenza kanjani, nokuthi kungani incoma imiphumela. Ibhizinisi lidinga ubuchwepheshe nge-AI okwenziwa ngayo indawo ukuze lenze ukuthi amaklayenti alo akwazi ukwenza izinhlelo zawo ze-AI zifake futhi zenziwe ngezifiso, ukuhlonipha ama-nuances abucayi olimini lwasekhaya kanye nokuhlangenwe nakho komsebenzisi okungenza noma okuphule ukwethembeka kwesisombululo se-AI kusuka ezweni elilodwa kuya kwelilandelayo . Isibonelo, ibhizinisi kufanele lidizayine izinhlelo zalo zokusebenza zezimo ezenziwe zaba ngezakho nezasendaweni, kufaka phakathi izilimi, izingqimba, nezindlela zokukhuluma kuzinhlelo zokusebenza ezenzelwe izwi. Ngaleyo ndlela, uhlelo lokusebenza luletha ileveli efanayo yokuhlangenwe nakho kwezwi kuzo zonke izilimi, kusuka esiNgisini kuya ezilimini ezingamelwe kahle.

Ubulungisa Nokwehlukahlukana

Ekugcineni, i-AI enengqondo iqinisekisa ukuthi izixazululo zakhelwe phezu kwamasethi edatha afanelekile futhi ahlukahlukene lapho imiphumela nomthelela wemiphumela ethile igadwa futhi ihlolwe ngaphambi kokuba isisombululo siye emakethe. Ngokuqaphela kanye nokufaka abantu kuzo zonke izingxenye zokuthuthukiswa kwesixazululo, sisiza ukuqinisekisa ukuthi amamodeli e-AI ahlala ehlanzekile, echemile kancane, futhi eziphatha ngendlela engakhona ngayo.

UCABANGANI?

Le sayithi isebenzisa i-Akismet ukunciphisa ugaxekile. Funda ukuthi idatha yakho yokuphawula isetshenziswa kanjani.