Isibuyekezo: U-Apple ukhulume ngokuhlolwa kwesibili kweseva, futhi inkampani yombono wekhompyutha echwepheshile yachaza ukuthi kungenzeka yini lokhu okungachazwa kokuthi “Indlela ukuhlolwa kwesibili okungasebenza ngayo” ngezansi.
Ngemuva kokuthi abathuthukisi behlehlise izingxenye zayo ezakhiwe, inguqulo yokuqala yohlelo lwe-Apple CSAM ikhohliswe ngempumelelo ukumaka isithombe esingenacala.Kodwa-ke, i-Apple ithe inezivikelo ezengeziwe zokuvikela lokhu ukuthi kwenzeke empilweni yangempela.
Intuthuko yakamuva yenzeke ngemuva kokuthi i-algorithm ye-NeuralHash ishicilelwe kuwebhusayithi kanjiniyela ovulekile ye-GitHub, noma ngubani angayizama…
Wonke amasistimu e-CSAM asebenza ngokungenisa isizindalwazi sezinto ezibonakalayo zokuhlukunyezwa kwezingane ngokocansi ezinhlanganweni ezifana ne-National Center for Missing and Exploited Children (NCMEC).Isizindalwazi sinikezwa ngesimo samahashi noma izigxivizo zeminwe zedijithali ezivela ezithombeni.
Nakuba izikhondlakhondla eziningi zobuchwepheshe ziskena izithombe ezilayishwe emafini, i-Apple isebenzisa i-algorithm ye-NeuralHash ku-iPhone yekhasimende ukuze ikhiqize inani le-hashi lesithombe esigciniwe, bese isiqhathanisa nekhophi elandiwe yenani le-CSAM hashi.
Izolo, umthuthukisi uthe uguqule i-algorithm ye-Apple futhi wakhulula ikhodi ku-GitHub-lesi simangalo siqinisekiswe ngempumelelo yi-Apple.
Emahoreni ambalwa ngemuva kokukhishwa kwe-GitHib, abacwaningi basebenzise ngempumelelo i-algorithm ukuze bakhe umbono ongelona iqiniso ngamabomu-izithombe ezimbili ezihluke ngokuphelele ezikhiqize inani elifanayo le-hashi.Lokhu kubizwa ngokuthi ukushayisana.
Kuzinhlelo ezinjalo, kuhlale kunengozi yokushayisana, ngoba i-hashi ngokuqinisekile iwukumela isithombe esenziwe lula, kodwa kuyamangaza ukuthi othile angakwazi ukukhiqiza isithombe ngokushesha.
Ukushayisana ngamabomu lapha kuwubufakazi bomqondo.Onjiniyela abanakho ukufinyelela kusizindalwazi se-CSAM hashi, okungadinga ukudalwa kwezinto ezingamanga ohlelweni lwesikhathi sangempela, kodwa kufakazela ukuthi ukuhlasela kokushayisana kulula ngokomgomo.
U-Apple ukuqinisekisile ngempumelelo ukuthi i-algorithm iyisisekelo sesistimu yayo, kodwa yatshela ibhodi lomama ukuthi lena akuyona inguqulo yokugcina.Inkampani iphinde yathi ibingahlosile ukuyigcina iyimfihlo.
U-Apple utshele i-Motherboard nge-imeyili ukuthi inguqulo ehlaziywe umsebenzisi ku-GitHub inguqulo evamile, hhayi inguqulo yokugcina esetshenziselwa ukutholwa kwe-iCloud Photo CSAM.U-Apple uthe futhi idalule i-algorithm.
"I-algorithm ye-NeuralHash [...] iyingxenye yekhodi yesistimu yokusebenza esayiniwe [futhi] abacwaningi bezokuphepha bangaqinisekisa ukuthi ukuziphatha kwayo kuvumelana nencazelo," kubhala idokhumenti ye-Apple.
Inkampani yaqhubeka yathi kunezinyathelo ezimbili ezengeziwe: ukusebenzisa uhlelo lwesibili (oluyimfihlo) lokufanisa kuseva yalo, nokubuyekezwa okwenziwa ngesandla.
U-Apple uphinde wathi ngemuva kokuthi abasebenzisi bephumelele umkhawulo wemidlalo engama-30, i-algorithm yesibili engeyona eyomphakathi esebenza kumaseva e-Apple izobheka imiphumela.
"Le hashi ezimele ikhethelwe ukwenqaba ukuthi kungenzeka i-NeuralHash enephutha ifane nesizindalwazi esibethelwe se-CSAM esisetshenziswayo ngenxa yokuphazanyiswa okuphikisana nezithombe okungezona eze-CSAM futhi yeqa umkhawulo ofanayo."
UBrad Dwyer we-Roboflow uthole indlela yokuhlukanisa kalula phakathi kwezithombe ezimbili ezithunyelwe njengobufakazi bomqondo wokuhlasela kokushayisana.
Ngiyafisa ukwazi ukuthi lezi zithombe zibukeka kanjani ku-CLIP yesikhici sesici se-neural esifanayo kodwa esihlukile se-OpenAI.I-CLIP isebenza ngokufanayo ne-NeuralHash;kuthatha isithombe futhi kusebenzisa inethiwekhi ye-neural ukuze kukhiqizwe isethi yamavekhtha esici afaka imephu kokuqukethwe kwesithombe.
Kepha inethiwekhi ye-OpenAI ihlukile.Kuyimodeli evamile engakwazi ukwenza imephu phakathi kwezithombe nombhalo.Lokhu kusho ukuthi singayisebenzisa ukuze sikhiphe ulwazi lwesithombe oluqondakalayo lomuntu.
Ngisebenzise izithombe ezimbili zokushayisana ngenhla nge-CLIP ukuze ngibone ukuthi nayo idukile.Impendulo emfushane ithi: cha.Lokhu kusho ukuthi i-Apple kufanele ikwazi ukusebenzisa inethiwekhi yesikhi yesici sesibili (efana ne-CLIP) ezithombeni ezitholiwe ze-CSAM ukuze inqume ukuthi zingokoqobo noma zingamanga.Kunzima kakhulu ukwenza izithombe ezikhohlisa amanethiwekhi amabili ngesikhathi esisodwa.
Ekugcineni, njengoba kushiwo ngaphambili, izithombe zibuyekezwa ngesandla ukuze kuqinisekiswe ukuthi ziyi-CSAM.
Umcwaningi wezokuphepha uthe okuwukuphela kwengozi yangempela ukuthi noma ngubani ofuna ukucasula i-Apple anganikeza imibono engamanga kubabuyekezi abangabantu.
"U-Apple empeleni uklame lolu hlelo, ngakho-ke umsebenzi we-hash awudingi ukugcinwa uyimfihlo, ngoba ukuphela kwento ongayenza nge-'non-CSAM njenge-CSAM' ukucasula ithimba labaphendulayo le-Apple ngezithombe ezingafuneki baze basebenzise izihlungi ukuze baqede. Ukuhlaziya Lezo zibi ezisepayipini ziyizinto ezingamanga,” uNicholas Weaver, umcwaningi omkhulu e-Institute of International Computer Science eNyuvesi yaseCalifornia, eBerkeley, etshela i-Motherboard engxoxweni ye-inthanethi.
Ubumfihlo buyinkinga ekhulayo yokukhathazeka ezweni lanamuhla.Landela yonke imibiko ehlobene nobumfihlo, ukuphepha, njll. kuzinkombandlela zethu.
U-Ben Lovejoy ungumbhali wezobuchwepheshe waseBrithani kanye nomhleli we-EU we-9to5Mac.Waziwa ngamakholomu akhe nezindatshana zedayari, ehlola ulwazi lwakhe ngemikhiqizo ye-Apple ngokuhamba kwesikhathi ukuze athole ukubuyekezwa okuphelele.Ubuye abhale amanoveli, kukhona amahlaya amabili ezobuchwepheshe, amafilimu amafushane ezinganekwane zesayensi kanye ne-rom-com!
Isikhathi sokuthumela: Aug-20-2021