(Isipho: Google)
Enkosi kubukrelekrele bokuzenzela, ukuhamba phesheya akukaze kube lula.
Inkqubo yoToliko lukaGoogle ivumela abasebenzisi ukuba baguqulele umbhalo kwangoko. Kwi-app, vele ukhombe ikhamera yakho kwisicatshulwa ofuna ukusiguqulela kwaye uza kuyibona iguqukela kulwimi lwakho olufunayo bukhoma, kanye phambi kwamehlo akho- akukho nxibelelwano lwe-Intanethi okanye idatha yeselfowuni efunekayo. Olu phawu luncedo belukhona kangangexesha elithile, kodwa lwalusebenza kuphela kwiilwimi ezisixhenxe. Ngoku , enkosi ekufundeni ngomatshini, uGoogle uphucule usetyenziso ukuze aguqulele kwangoko iilwimi ezingama-27.
Ke kwixesha elizayo xa usePrague kwaye awukwazi ukufunda imenyu, sikubuyile, u-Otavio Good, injineli yesoftware kuGoogle, wabhala kuphando lwenkampani Ibhlog .
UGoogle uphinde wasebenzisa i-AI ukusika iimpazamo zabo zokuqonda intetho.
Ukusukela namhlanje, ukongeza kuguqulelo phakathi kwesiNgesi, isiFrentshi, isiJamani, isiTaliyane, isiPhuthukezi, isiRashiya kunye neSpanish, ezi lwimi zingama-20 zilandelayo zinokuguqulelwa ngexesha lokwenyani: isiBulgaria, isiCatalan, isiCroatia, isiCzech, isiDanish, isiDatshi, isiFilipino, IsiFinnish, isiHungary, isi-Indonesia, isiLithuania, isiNorway, isiPolish, isiRomania, isiSlovak, iSweden, isiTurkey nesiUkraine. Kwaye ukuba ukhetha ukuthatha umfanekiso endaweni yokubukela isicatshulwa siguqulela bukhoma, zizonke iilwimi ezingama-37 ziyaxhaswa.
Ke uGoogle wakwazi njani ukunyusa inani leelwimi ezikhoyo? Baqala ukufumana iLens Lens, ngaphambili isicelo soguqulelo lwenyani, kunye nokusetyenziswa komatshini wokufunda kunye nenethiwekhi ye-neural eguqukayo ukuphucula amandla eapp. Inkqubela phambili yokwamkelwa komfanekiso yayiyeyona nto iphambili.
Kwiminyaka emihlanu edlulileyo, ukuba unike ikhompyuter umfanekiso wekati okanye inja, ibinengxaki yokuxela ukuba yeyiphi na. Ndiyabulela kuthungelwano lwe-neural eguqukayo, kungekuphela nje ukuba iikhompyuter zinokuthi zixelele umahluko phakathi kweekati kunye nezinja, zinokukwazi ukuqonda iintlobo ezahlukeneyo zezinja, utshilo uMnu. Ewe zilungile ngaphezulu kokulungileyo ubugcisa betrippy -Ukuba uguqula imenyu yangaphandle okanye utyikitye nohlobo lwamva lwe-Google Translate app, ngoku usebenzisa i-neural net net.
Inyathelo nenyathelo
Ekuqaleni , Guqulela kufuneka ukhulise ukungqubana okungasemva kwaye ufumane isicatshulwa. Xa ibeka iibhloko zepixels zombala ofanayo, igqiba ekubeni ngoonobumba. Kwaye xa ezo blobs zisondelelene, ziyaqonda ukuba ngumgca oqhubekayo ofundwayo.
Okulandelayo, usetyenziso kufuneka luqonde ukuba ngubani unobumba ngamnye. Kulapho kungena khona ukufunda okunzulu.
Sisebenzisa inethiwekhi ye-neural eguqukayo, ukuyiqeqesha ngoonobumba nakwizinto ezingezizo iileta ukuze ikwazi ukufunda ukuba jonga njani iileta ezahlukeneyo, ifunda iposti yebhlog.
Abaphandi kuye kwafuneka baqeqeshe isoftware bengasebenzisi iileta ezibonakala zicocekile, kodwa nazo zingcolile. Iileta ezivela kwilizwe lokwenyani zonakaliswe kukubonakaliswa, ukungcola, ii-smudges, kunye nazo zonke iintlobo zobungqindilili, utshilo uMnu. Ke siyile umenzi weleta yethu ukwenza zonke iintlobo zobumdaka obungeyonyani ukuze silingise ngokuqinisekileyo ingxolo yelizwe lokwenyani- ukubonakalisa okungeyonyani, ubuqhetseba obungeyonyani, ubuqhetseba obungeyonyani ngeenxa zonke. 
Ezinye zeeleta ezimdaka ezisetyenziselwa uqeqesho. (Ifoto: Google)
Inkqubo ye- Isithathu Inyathelo likhangela oonobumba abaziwayo kwisichazi-magama ukufumana inguqulelo. Kwaye kwimizamo eyongeziweyo yokuchaneka, ii-lookups zesichazi-magama ziyaqikelelwa xa ngaba i-S ayifundwanga kakuhle njenge-5.
Okokugqibela, isicatshulwa esiguqulelweyo senziwe ngaphezulu kwendalo ngohlobo olufanayo.
Singakwenza oku kuba sele sifumene kwaye sifunde iileta emfanekisweni, ke siyazi ngqo ukuba ziphi. Singajonga imibala ejikeleze iileta kwaye sisebenzise ukucima oonobumba bokuqala. Kwaye emva koko sinokuzoba ukuguqulela ngaphezulu sisebenzisa umbala wokuqala wangaphambili, iposti yebhlog iyafundwa.
Ukuze usebenze ngokusemandleni akho kwaye uvumele onke la manyathelo ukuba agqitywe ngexesha lokwenyani ngaphandle kwe-Intanethi okanye uqhagamshelo lwedatha, iqela likaGoogle liphuhlise intambo encinci kakhulu ye-neural enesibophelelo esiphakamileyo kuxinano lolwazi olunokuphatha. Kuba babesenza eyabo idatha yoqeqesho, kwakubalulekile ukubandakanya idatha elungileyo kodwa akukho nto yongezelelweyo ke inethiwekhi ye-neural ayisebenzisi kakhulu ubuninzi bayo bolwazi kwizinto ezingabalulekanga. Umzekelo ingayindlela ekufunwa ngayo ukuqonda unobumba onenani elincinci lokujikeleza, kodwa hayi kakhulu.
Ekugqibeleni, abasebenzisi bashiye iilwimi ezingama-20 ngaphezulu kodwa isantya esifanayo.
BONA KWAKHO: Iqela le-AI leGoogle elisinikeze ukuthothisa kuPhando lwabo lokuFunda ngomatshini
