Ukubukeka kwe- ama-deepfakes ku-YouTube Kushintshe ngokuphelele isimo samavidiyo aku-inthanethi. Lokho okwakubonakala sengathi kuwukuhlola okulula kwezobuchwepheshe eminyakeni embalwa edlule sekuyinkinga enkulu: amavidiyo alingisa osaziwayo, osopolitiki, noma abadali bokuqukethwe ngokunemba okuxakile, abakwazi ukukhohlisa, ukuphatha kabi, noma ukubhubhisa idumela ngamahora ambalwa.
Ibhekene nale nselele, i-YouTube inqume ukuthatha isinyathelo esibalulekile bese isebenzisa isu elinamandla kakhulu ngokumelene nokuqukethwe okukhiqizwe yi-AI ezizenza abantu bangempela. Amathuluzi amasha, izinqubomgomo eziqinile, izinhlelo zokubika ezithile, kanye nokusebenzisana nemboni yezokuzijabulisa kanye nabantu ababalulekile emphakathini kuyingxenye yokuhlasela okuhlose ukunqanda ukusabalala kwalezi zinto ezingelona iqiniso.
Ayini ama-deepfake futhi kungani eyinkinga ku-YouTube?
Uma sikhuluma ngama-deepfake, sibhekisela kumavidiyo, izithombe, noma umsindo owenziwe noma oguquliwe yi- ubuhlakani bokwenziwa kokukhiqiza ukulingisa ubuso, izwi, noma ukushukuma komuntu wangempela. Ku-YouTube, lokhu kuhunyushwa kube iziqeshana lapho usaziwayo ebonakala esho noma enza into engakaze yenzeke, noma izingoma eziculwa ngezwi elihlanganisiwe lomculi ngaphandle kwemvume yakhe.
Lolu hlobo lokuqukethwe lunezinhloso eziningi ezingaba khona: kusukela ku- ukukhohlisa okukhulu kanye nolwazi olungelona iqiniso (isibonelo, ngesikhathi sokhetho), kusukela ekuzijabuliseni okulula okungenasibopho kuya ekuhlaselweni okuqondile kwedumela losaziwayo, osopolitiki, noma abantu abadumile emphakathini. Inkinga enkulu ukuthi, njengoba ubuchwepheshe buthuthuka, kuba nzima kakhulu ukuhlukanisa okuyiqiniso kulokho okukhiqizwe yi-AI.
Eminyakeni yamuva nje, i-YouTube igcwele lokho i-CEO yayo ekuchaze ngokuthi "Okuqukethwe okungadingekile" okukhiqizwe yi-AIPhakathi kwazo, izinganekwane ezijulile zosaziwayo zivelele ngamandla azo ayingozi: azigcini nje ngokukhathaza umsebenzisi ofuna amavidiyo asezingeni eliphezulu, kodwa futhi zibeka engcupheni ubusemthethweni, ilungelo lokuba nesithombe sakho kanye nokwethemba ipulatifomu.
Ochwepheshe sebekhuluma kakade ngengxabano enkulu ne- ubuwena bedijithali kanye nokuvikelwa kwabantu abadumile emphakathiniNoma ubani onokufinyelela kumathuluzi e-AI athuthukile angakha ividiyo ekholekayo elingisa umlingisi, umculi, noma usopolitiki bese eyilayisha ku-YouTube. Ngaphandle kohlelo lokuthola nokuphendula olusebenzayo, umonakalo wedumela ungaba mkhulu futhi kube nzima ukuwubuyisela emuva.
Impendulo ye-YouTube: Sebenzisa i-AI ukulwa ne-AI
Ngenxa yalesi simo, i-YouTube ikhethe ukusebenzisa umqondo wokuthi “Uma ungakwazi ukubahlula, bajoyineEsikhundleni sokumane siphendule emibikweni eyenziwa ngesandla, ipulatifomu ithuthukise izinhlelo ezisekelwe ekuhlakanipheni kokwenziwa ukuze kutholakale amavidiyo okwenziwa ahlukumeza isithombe noma izwi labantu bangempela.
Le nkampani imemezele ukwandiswa kwe- ubuchwepheshe bokuthola ukufanaLeli thuluzi lenzelwe ukuhlonza ngokuzenzakalelayo okuqukethwe okukhiqizwe yi-AI okulingisa ubuso noma izwi labantu abathile. Lobu buchwepheshe abubheki nje kuphela izinto ezingabonakali: buhlaziya amaphethini obuso, ama-timbres ezwi, nezinye izici ezihlukile ukuze kunqunywe ukuthi okuqukethwe kungaba yini i-deep fake.
Okuthakazelisayo ukuthi i-YouTube ihlanganisa lezi ntuthuko zobuchwepheshe ne- inqubomgomo yokuvikela eqinile kakhuluAkukhona nje ukuthola ividiyo, kodwa kumayelana nokunikeza labo abathintekile isiteshi esicacile nesisheshayo sokucela ukuthi isuswe, ikakhulukazi uma kukhona ukuzenzisa, ukukhohlisa noma ukusetshenziswa kabi kwesithombe sabo.
Lo mzamo uphendula ekucindezelweni okukhulayo okuvela kubasebenzisi, abaculi, amalebula okurekhoda, izinhlangano zamakhono, kanye neminye imikhakha esengozini enkulu yokuthola ulwazi olungelona iqiniso. Bonke bebelokhu benxusa i-YouTube ukuthi ithathe isinyathelo isikhathi esithile. izindlela ezisebenzayo zokuvimba ukuhlanganiswa kwamanga kanye nokusetshenziswa okungagunyaziwe kwesithombe noma izwi labo.
Kusukela ku-Content ID kuya ohlelweni lokuthola olufana nolwakho
Ukuze uqonde ukuthi lokhu kuhlasela okusha kusebenza kanjani, kuyasiza ukubheka emuva. I-YouTube ibilokhu ikusebenzisa lokhu iminyaka eminingi. Ubunikazi bokuqukethweLuhlelo oluskena amavidiyo alayishwe epulatifomu bese luwaqhathanisa nesizindalwazi esikhulu semisebenzi enelungelo lobunikazi (umculo, iziqeshana zevidiyo, njll.). Uma luthola ukufana, lusebenzisa ngokuzenzakalelayo imithetho yelungelo lobunikazi.
Ubuchwepheshe bokuthola ukufana buphefumulelwe ngqo yilowo modeli, kodwa ngendlela ehlukile. Esikhundleni sokulandelela izingoma noma iziqeshana ezine-copyright, buyasebenza sesha ukufana phakathi kwamavidiyo alayishiwe kanye "nomzila wedijithali" wabantu bangempelaLowo mfanekiso uhlanganisa izici zobuso babo, amaphethini okunyakaza, futhi, ezimweni eziningi, izici zezwi labo.
Njalo uma ividiyo entsha ilayishwa, ama-algorithm e-YouTube ahlaziya okuqukethwe kusetshenziswa amasu e- ukuqashelwa kobuso nomsindo kuhlanganiswe namamodeli e-AI akhiqizayo aqeqeshwe ukuthola ukuxhashazwa. Uma uhlelo luthola okuthile okusolisayo, lumaka ividiyo ukuze ibuyekezwe futhi luyihlobanise nephrofayili yedijithali yomuntu ongase alingiswa.
Empeleni, lokhu kuvumela izisulu ezingaba khona ze-deepfakes ukuthi zibe nedeshibhodi lapho zingathola khona okuqukethwe okusebenzisa isithombe noma izwi lazo ngaphandle kwemvume, futhi kusukela lapho, cela ukususwa kwayoKuyinguquko ecacile kusukela endleleni "yokusabela" esekelwe kuphela ezikhalweni ezenziwe ngesandla kuya kumodeli yokusabela esekelwe ekuskeni okuzenzakalelayo.
Ukufinyelela okubalulekile embonini yezokuzijabulisa
Enye yentuthuko entsha enkulu ukuthi i-YouTube inqume ukwandisa lobu buchwepheshe bokuthola obufanayo ku- imboni yezokuzijabulisaAma-ejensi ethalente, izinkampani zokumela, kanye nosaziwayo ababaphathile kakade banenketho yokusebenzisa leli thuluzi ukuvikela abaculi ababamele.
Le nkundla ibambisene eduze nezinhlangano ezinkulu kulo mkhakha, njenge I-CAA, i-UTA, i-WME kanye nokuphathwa okungenasihlokoukulungisa ama-algorithms ukuze avumelane nezimo zangempela. Umqondo uwukuthi ithuluzi ngeke lisebenze kuphela elabhorethri, kodwa lizosebenza kahle ekusetshenzisweni kwansuku zonke, ngazo zonke izinhlobo zokukhanya, izimfanelo zevidiyo, kanye nezimo lapho ama-deepfake avame ukuzungeza khona.
Imininingwane ebalulekile ukuthi abaculi nosaziwayo Akudingeki babe nesiteshi se-YouTube Ukuze bazuze kulokhu kuvikelwa, badinga nje ukubhaliswa ohlelweni ngabameleli babo noma ngokuqondile, ukuze ipulatifomu ikwazi ukudala nokugcina iphrofayili yabo yedijithali.
Le ndlela yandisa ububanzi: ayivikeli nje kuphela abadali abalayisha okuqukethwe nsuku zonke, kodwa futhi nosaziwayo abanezithombe ezingasetshenziswa ngisho noma bengenakho ukuba khona okusebenzayo ku-YouTube. Eqinisweni, ezimweni eziningi ama-deepfake athembela ezithombeni noma kumavidiyo asakazwa kwamanye amanethiwekhi, kodwa agcina esetshenziswa epulatifomu ye-Google.
Ukuvikelwa okwengeziwe kwezombusazwe, izintatheli kanye nezikhulu zomphakathi
Ithuluzi lokuthola elifana nelibukekayo aligcini nje ezweni lokuzijabulisa. I-YouTube yaqala ukulethula ngo-2024 ukuze lisetshenziswe iqembu elincane labasebenzisi abazwela kakhulufuthi ngokuhamba kwesikhathi iye yandisa ukumbozwa kwayo kwamanye amaphrofayili anobungozi obukhulu.
Ekuqaleni kuka-2026, inkampani yamemezela ukuthi uhlelo luzotholakala futhi izikhulu zikahulumeni, izintatheli, kanye nabantu abazongenela ukhetho kwezepolitikiLa maqembu ayizisulu ezibaluleke kakhulu zemikhankaso yolwazi olungelona iqiniso, izithombe ezishintshiwe, namavidiyo mbumbulu aklanyelwe ukuthonya umbono womphakathi.
Kulesi simo, uhlelo lwe Ukutholwa Kokufana Iba ithuluzi elibalulekile. Ubuchwepheshe buhlaziya ozimele namathrekhi omsindo ukuze kutholakale ukuthi izithombe zalaba bantu zisetshenziswa yini ku-montage eyenziwe nge-AI engadukisa, ihlasele ukwethembeka kwabo, noma iphazamise impikiswano yomphakathi.
Abantu ababhalisele lolu hlelo bathola izexwayiso lapho ipulatifomu ithola icala elingase libe khona lokuzenza ongeyena okusebenzisa i-AI. Kodwa-ke, Ukususwa kwevidiyo akuzenzekeli.Labo abathintekayo kumele bacele ngokusobala ukuthi isuswe, bese i-YouTube yenza ukuhlaziya okwengeziwe ukuhlola umongo bese inquma ukuthi izinqubomgomo zayo zephuliwe yini.
Ukubhaliswa nokuqinisekiswa ohlelweni lokuvikela
Ukuze bazuze kulezi zinyathelo ezithuthukisiwe, abantu abadumile abafisa ukuhlanganyela kulolu hlelo kumele badlule ku- inqubo yokubhalisa nokuqinisekisaI-YouTube idinga ukuthi umuntu afakazele ubuwena bakhe ngokusebenzisa imibhalo nezinto ezilalelwayo nezibonwayo.
Empeleni, abazuzi bacelwa ukuthi bandise imali engenayo yabo ngomuntu oyedwa Ividiyo yokuqinisekisa kanye nokugunyazwa okusemthethweni okuvumela isiqinisekiso sokuthi bangobani ngempela. Ngalolu lwazi, ipulatifomu yakha iphrofayili yedijithali ethembekile esetshenziswa ama-algorithms ukuqhathanisa nokuqukethwe okulayishwe abasebenzisi.
Lesi sinyathelo sibalulekile ukuvimbela ukusetshenziswa kabi kohlelo. Ngaphandle kokuqinisekiswa okuqinile, othile angazama ukubiza isithombe somunye umuntu njengesakhe, noma aphoqelele ukususwa kwamavidiyo asemthethweni ngokumangalela ukuzenza ongeyena. Yingakho i-YouTube ihlanganisa ukuqinisekiswa ngesandla nezinqubo zobuchwepheshe ezizenzakalelayo ngaphambi kokusebenzisa ukuvikelwa.
Uma sekuqinisekisiwe, abasebenzisi abafakiwe ohlelweni bangafinyelela amathuluzi athile okusesha nokubikalapho amavidiyo asolisayo ebalwe khona futhi kunikezwa izinketho ezicacile zokuqalisa izinqubo zokukhalaza nokususa okuqukethwe.
I-Creative AI vs. i-AI enonya: izinhlangothi ezimbili zalesi simo
Konke lokhu kumele kuqondwe esimweni lapho i-YouTube nayo iqhuba khona ukusetshenziswa kobuhlakani bokwenziwa ngokudala ngabadali. Ipulatifomu ngokwayo isiqalile ukunikeza amathuluzi okukhiqiza ama-avatar abukeka futhi azwakala njenge-YouTuber, izihlungi ezithuthukisiwe, kanye nezinketho zokuhlela ezihlakaniphile.
Inkinga ukuthi ubuchwepheshe obufanayo obukuvumela ukuthi udale isithombe esihle sesiteshi sakho bungasetshenziswa futhi ekwenzeni ukuzenzisa okungokoqobo kakhuluYingakho inkampani ihamba kahle kakhulu: ikhuthaza ukusungula izinto ezintsha, kodwa idinga ukuyeka ukusetshenziswa okuwela imingcele yezomthetho noma yokuziphatha.
Isibonelo okuxoxwe ngaso kakhulu kube yindaba yezingoma ezidumile ezizwakala sengathi zingabaculi abadumile, njengalezo ezicatshangelwayo Ingoma ethi Bad Bunny ikhiqizwe yi-AI ngokupheleleEqinisweni, lo mculi akazange aqophe leyo ngoma, kodwa ukufana kwezwi kwakukhulu kangangokuthi abasebenzisi abaningi bayithatha njengeyiqiniso futhi yasakazeka ezinkundleni zokuxhumana kanye namavidiyo.
Amacala anjengalawa abangele izingcingo ezixwayisa amalebula okuqopha, izinkampani zokuqopha, kanye nabaculi ngokwabo, asebekwenzile lokho manje. Izindlela zokucela ukususwa kwezingoma namavidiyo abalingisa izwi noma isithombe sabo ngaphandle kwemvume. Ukuvikelwa akusagcini nje ngelungelo lobunikazi lendabuko: ilungelo lokuzazi kanye nesithombe sakho manje seliqala ukusebenza.
Izibopho ezintsha zabadali ku-YouTube
Ngaphezu kwamathuluzi okuthola, i-YouTube ibuyekeze izinqubomgomo zayo ukuze ifune abadali ukuthi ukumaka ngokusobala okuqukethwe okukhiqizwe noma okushintshiwe yi-AIikakhulukazi lapho kukhulunywa ngezindaba ezibucayi njengokhetho, impilo yomphakathi, noma izinkinga zobuntu.
Kusukela manje kuqhubeke, amavidiyo asebenzisa kakhulu i-AI kumele afake phakathi amalebula noma izaziso ezicacile ukuze umbukeli azi ukuthi ezinye izingxenye zokuqukethwe zenziwe ngendlela yokwenziwa. Lokhu kufaka phakathi ukusetshenziswa kwamathuluzi asekelwe ku-AI e-YouTube kanye nezinhlelo zokusebenza zezinkampani zangaphandle.
Ipulatifomu ihlukanisa, isibonelo, phakathi kwamavidiyo abonisa umuntu esho noma enza into angakaze ayenze (icala elijwayelekile lokuzenzisa) kanye nokuphinda kwenziwe izehlakalo ezingakaze zenzeke. Kuzo zombili izimo, abadali bacelwa ukuthi babonise ngokusobala ukuthi okuqukethwe kukhiqizwe ngedijithali noma kushintshiwe.
Ukuze kuqiniswe lokhu kucaca, i-Google yayikade ifuna lokho ngaphambili Izikhangiso zezepolitiki ku-YouTube zizofaka izixwayiso lapho behlanganisa izinto ezidalwe ubuhlakani bokwenziwa. Manje, le filosofi ibuye ifinyelele kokuqukethwe kwezinto eziphilayo kweziteshi.
Isikrini Samaphupho kanye nenqolobane yokuqukethwe okukhiqizwe yi-AI
Kanye nezinqubomgomo zokumaka kanye nezinhlelo zokuthola ukubukeka okufanayo, i-YouTube yethule amathuluzi anjenge Isikrini Sephupho, eyenzelwe ngqo ukudala okuqukethwe ngamafomethi afana nama-Shorts kanye Ama-avatar e-AILesi sici sikhiqiza izizinda kanye nezinto ezibonakalayo sisebenzisa i-AI, eyenzelwe ukwenza amavidiyo amafushane akhange kakhulu.
Okubalulekile ngokombono wokulwa nama-deepfake ukuthi ipulatifomu idala uhlobo lwe- "i-cache" noma irekhodi lezinto ezikhiqizwe yi-AIUkuhlonza lezi zinsiza kwenza kube lula ukuhlola ukuthi ziyasetshenziswa kabusha noma zihlanganiswa ngendlela ekhohlisayo kwamanye amavidiyo.
Ngale ndlela, i-YouTube izama ukulinganisela izikali: ngakolunye uhlangothi, inikeza abadali amathuluzi anamandla okusungula izinto ezintsha; ngakolunye uhlangothi, uqapha ukusetshenziswa kwalezo zinsiza ukuvimbela ukuthi bangabhebhezeli igagasi "le-AI junk" elisongela ukuqeda ipulatifomu.
Le ndlela ivula futhi ithuba lokuthi inkampani isebenzise izinqubomgomo zokulawula ezizenzakalelayo ezithuthukisiwe esikhathini esizayo. iziqeshana ezikhiqizwe ngokuphelele yi-AI ezingangezi inani noma ezisetshenziselwa ukukhohlisa noma ukuqamba amanga kuphela.
Inqubo yokubika nokususa ama-deepfake ku-YouTube
Ngaphandle kokutholwa okuzenzakalelayo, i-YouTube yethule uhlelo lokusebenza inqubo ethile yezikhalazo kumavidiyo akhiqizwe yi-AI alingisa abantu bangempela. Le ndlela ayiyona eyabantu abadumile kuphela: ingasetshenziswa nabasebenzisi abavamile abathintekayo.
Uma kukhulunywa ngabaculi kanye namalebula abo, inqubomgomo entsha iyabavumela cela ukususwa kwama-deepfake alingisa izwi noma isithombe sakho ngaphandle kwemvume esemthethweni. Lolu ushintsho olukhulu kusukela ezigabeni zangaphambilini, lapho kwakugxilwe khona kakhulu ku-copyright.
Abasebenzisi abangaziwa nabo abakhishwa ngaphandle. Noma ubani othola ividiyo lapho isithombe noma izwi lakhe lisetshenziswa khona ngenhloso yokukhohlisa noma ukuhlukumeza angaqala inqubo yesikhalazo. Ipulatifomu izobuyekeza icala futhi, uma liwela ngaphansi kwezinqubomgomo zayo, sizoqhubeka nokususa okuqukethwe noma ukusebenzisa ezinye izinyathelo.
Ngakolunye uhlangothi, umphakathi ngokwawo ungadlala indima ebalulekile ngokubika okuqukethwe okusolisayo. Uma abasebenzisi abaningi bemaka ividiyo njenge-deepfake engaba yingozi, kungenzeka kakhulu ukuthi ifakwe ohlelweni. amasekethe okuhlola asheshisiwe futhi izinqumo zenziwa ngokushesha okukhulu.
Indlela yokuvimbela okuqukethwe kwakho okukhiqizwe yi-AI ukuthi kungasuswa
Ngokusobala, akuwona wonke okuqukethwe okukhiqizwe yi-AI okuhloselwe ukuzenza abanye noma ukukhohlisa izilaleli. Abadali abaningi bazama la mathuluzi ngendlela... ubuciko, ukuhlekisa noma ukufundisa ngaphandle kwenhloso embi. Kubo, i-YouTube ibeka nochungechunge lweziqondiso.
Umthetho wegolide wokunciphisa ingozi yokuhoxa yilona obalaUma ividiyo ifaka izigcawu ezingokoqobo ezikhiqizwe yi-AI, umbhali kufanele abonise lokhu ngokucacile: encazelweni, kumathegi, futhi, lapho kungenzeka, ngqo kuvidiyo ngokwayo ngezaziso noma amalebula.
Ipulatifomu incoma ukuthi umdali acacise ukuthi ividiyo iyaboniswa yini izehlakalo ezingakaze zenzekeNoma uma othile ebonakala ekhuluma noma enza izinto angazange azenze ngempela. Lo mongo usiza umbukeli ukuthi ahumushe lokho akubonayo futhi ubonise ukuthi akukho mzamo wokudlulisa i-montage njengedokhumenti yangempela.
Ngaphezu kwalokho, kubalulekile ukuthi labo abasebenzisa izwi noma isithombe sabantu besithathu, ngisho noma kukhulunywa ngokuhlekisa, babe ukugunyazwa okucacile lapho umphumela ungadidaniswa nokuqukethwe kwangempela noma ubangele umonakalo omkhulu edumeleni.
Uma umdali engayinaki le mihlahlandlela futhi esebenzisa kabi ukusetshenziswa kwe-AI ukudala ama-deepfake akhohlisayo, akazivezi nje kuphela ku- ukususwa kwevidiyokodwa futhi nasezinqumweni ezingathi sÃna kakhulu, njengokukhawulela ukusebenza kwesiteshi kanye nokumisa ukwenza imali.
Izijeziso kulabo abaphula izinqubomgomo ezintsha
Imiphumela yeziteshi ezingayihloniphi imithetho ye-YouTube mayelana ne-AI kanye nama-deepfake ingaba yingozi kakhulu. Uma kutholakala ukwephulwa okucacile, enye yezinyathelo zokuqala... ukususwa okuqondile kokuqukethwe okuthintekileikakhulukazi uma kuyizinto ezidukisayo ezindabeni ezibucayi noma ezihilela ukuzenzisa.
Uma ukwephulwa komthetho kuphindaphindwa, ipulatifomu ingakhuphukela ezijezisweni ezinzima kakhulu, njengokuthi ukumiswa kwezinkokhelo zemali ezikhiqizwa ngohlelo lozakwethu. Kwabadali abaningi, lokhu kusho ukulahlekelwa umthombo oyinhloko wemali engenayo.
Ezimweni ezimbi kakhulu neziphindaphindwayo, noma lapho isiteshi silayisha ngokuhlelekile ama-deepfake ayingozi noma "okuqukethwe okungenamsoco okudalwe yi-AI," i-YouTube ingase ithathe isinyathelo. vala isiteshi noma uvimbele ukudalwa kwamaphrofayili amasha kuxhunywe kumnikazi ofanayo. Umqondo uwukuthumela umlayezo ocacile wokuthi lolu hlobo lwemikhuba alumukelekile.
Ngalezi zijeziso, inkampani izama ukuvimba labo abasebenzisa i-AI ngokungafanele kanye nalabo abayisebenzisa ngokunganaki ngaphandle kokucabangela umthelela amavidiyo abo angaba nawo empilweni yangempela yabantu abathintekile.
Uma kuhlanganiswa, inhlanganisela yobuchwepheshe bokuthola okufana nokubukeka, izinqubomgomo zokucaca, amathuluzi okuveza ulwazi, kanye nezijeziso zomnotho kuveza isithombe lapho Ukusebenzisa kabi ama-deepfake kuya ngokuya kubiza kakhulu ngaphakathi kwepulatifomu.
Lonke lolu hambo lwe-YouTube lubonisa izinga impi yokulwa nama-deepfakes esebe yimbangela yempikiswano yedijithali. Ipulatifomu ifuna ukuhlala iyindawo lapho ubuhlakani kanye ne-AI kungahlala khona ndawonye, ​​kodwa ngokudweba imigqa ecacile ngokumelene nokuzenzisa, ukukhohlisa, kanye nolwazi olungelona iqiniso. Impumelelo yalezi zinyathelo ngeke incike kuphela kuma-algorithms ayo, kodwa futhi nasesibophweni sabadali kanye nomphakathi wabasebenzisi ohlonipha ubuqiniso kanye nokuzijabulisa.
