🕵️ IBrithani izosebenzisana neMicrosoft ukwakha uhlelo lokuthola ama-deepfake ↗
IBrithani ithi ibambisana neMicrosoft, izifundiswa, kanye nabanye ochwepheshe ukwakha uhlelo olungakwazi ukubona ama-deepfake ku-inthanethi - kanye nendlela efanele yokuhlola ukuthi amathuluzi okuthola ayasebenza yini endle.
Kugcizelelwa kakhulu ekulimaleni okubangela usizi: ukuzenza umuntu ongeyena, ukukhwabanisa, kanye nokuqukethwe kocansi okungavunyelwene ngakho. Akuyona into ekhangayo, kodwa yilapho umonakalo ugxila khona.
🧠 I-Anthropic ikhipha ukuthuthukiswa kwe-AI njengoba imakethe ijezisa amasheya esoftware ↗
I-Anthropic yethule imodeli ye-Claude ethuthukisiwe (i-Opus 4.6), yaveza ukusebenza okuqinile ekubhaleni amakhodi kanye nezimali, kanye nekhono lokusebenzisa amathokheni afinyelela ku-1 million ngesikhathi esisodwa. Lokho kuningi kakhulu - njengokunikeza imodeli ikhadi lonke lomtapo wolwazi.
Baphinde bathembele “kuma-ejenti” ngeClaude Code, lapho imisebenzi ingahlukaniswa khona phakathi kwabasizi abazimele. Kulula, ngaphezu kwalokho obungakulindela - kanye nohlobo loshintsho olwenza abatshalizimali besoftware yakudala bakhathazeke kancane.
🧑💼 I-OpenAI yembula isevisi ye-ejenti ye-AI njengengxenye yokucindezela ukuheha amabhizinisi ↗
I-OpenAI imemezele i-“Frontier,” isevisi egxile ebhizinisini yokwakha nokuphatha ama-ejenti e-AI - ama-bot enza imisebenzi ethile efana nesofthiwe yokulungisa amaphutha, ukuphatha imisebenzi yokusebenza, into enjalo.
Okubi ukuthi ihloselwe ukuxhuma ingqalasizinda yenkampani ekhona futhi isekele ngisho nama-ejenti ezinkampani zangaphandle, okufundeka njengokudoba kwe-OpenAI ukuze kube yigumbi lokulawula - hhayi nje umhlinzeki wemodeli. Isifiso sokuvelela… noma ukuba nomona kancane, kuye ngokuthi uyibheka kanjani.
🪖 I-US ne-China bakhetha ukungahlanganyeli esivumelwaneni esihlangene sokusetshenziswa kwe-AI empini ↗
Amazwe amaningi asayine isimemezelo esingabophezeli esibeka izimiso zokusebenzisa i-AI empini - izinto ezifana nomthwalo wemfanelo womuntu, izakhiwo zomyalo ezicacile, kanye nokuhlolwa okujulile kanye nokuhlolwa kwengozi.
Kodwa i-US ne-China abazange basayine. Okuyindaba yonke ngesigqi esisodwa esibi - incwadi yemithetho ikhona, futhi abadlali abakhulu… ababambe ipeni.
🎭 Umthetho waseNew York udinga ukudalulwa kwabadlali abakhiqizwe yi-AI ekukhangiseni futhi uqinisa amalungelo okukhangisa ngemva kokufa ↗
INew York isayine imithetho edinga ukuthi izikhangiso ezisebenzisa “abenzi bokwenziwa” abakhiqizwe yi-AI zidalule lokho ngokucacile ezithamelini esifundazweni - noma ngabe umkhangisi usezindaweni ezithile. Ngokuyisisekelo: ungamlimazi umuntu oqanjiwe odlule kubantu bese ukubiza ngokuthi ukumaketha.
Ngakolunye uhlangothi, umbuso uqinise imithetho ephathelene nokusetshenziswa kwezentengiselwano kwemifanekiso yabantu abafileyo, okuhlanganisa namakhophi edijithali. Kubi kancane, uma ngikhuluma iqiniso, kodwa futhi "siyakwamukela kakhulu enkathini lapho ubuwena budinga ibhande lesihlalo."
Imibuzo Evame Ukubuzwa
Yini iBrithani eyakhayo neMicrosoft ukuthola ama-deepfakes ku-inthanethi?
IBrithani ithi izosebenzisana neMicrosoft, izifundiswa, kanye nabanye ochwepheshe ukwakha uhlelo olungakwazi ukubona ama-deepfake ku-inthanethi. Lolu hlelo luhlanganisa nokudala indlela ethembekile yokuhlola ukuthi amathuluzi okuthola ayasebenza yini ngaphandle kwelebhu. Ukugxila kuhlala kuyingozi elimaza kakhulu: ukuzenza umuntu ongeyena, ukukhwabanisa, kanye nokuqukethwe kocansi okungavunyelwene ngakho.
Ukuhlolwa "kwasendle" kwama-deepfake detectors kuzohluka kanjani kuma-demo?
Kanye nomtshina uqobo, iBrithani igcizelela indlela yokuhlola amathuluzi okuthola ama-deepfake ngaphansi kwezimo ezingokoqobo. Ezindleleni eziningi, amamodeli abukeka eqinile kumasethi edatha akhethiwe kodwa abuthakathaka lapho amafomethi okuqukethwe, ukucindezelwa, noma amaqhinga okuphikisana kushintsha. Indlela yokuhlola ehlelekile isiza ukuqhathanisa amathuluzi njalo futhi yembula lapho ukutholwa kwehluleka khona ezindaweni ezibukhoma eziku-inthanethi.
Yini eyashintshwa yi-Anthropic kuClaude Opus 4.6, futhi kungani umkhawulo wethokheni ubalulekile?
I-Anthropic ithi iClaude Opus 4.6 ithuthukisa ukusebenza ezindaweni ezifana nokubhala amakhodi kanye nezezimali, futhi ingaphatha amathokheni afinyelela ku-1 million womongo. Lelo windi elikhulu lomongo lingenza kube lula ukusebenza kumadokhumenti amade noma amakhodi ayinkimbinkimbi ngaphandle kokulayisha kabusha ulwazi njalo. Baphinde bagqamisa "ama-ejenti" ngeClaude Code, lapho imisebenzi ingahlukaniswa khona kubasizi abazimele.
Iyini i-OpenAI “Frontier,” futhi ingena kanjani emisebenzini yebhizinisi?
I-OpenAI imemezele i-"Frontier" njengesevisi egxile ebhizinisini yokwakha nokuphatha ama-ejenti e-AI enza imisebenzi ethile, njengesofthiwe yokulungisa amaphutha noma ukuphatha imisebenzi yokusebenza. Ikulungele ukuxhuma ingqalasizinda yenkampani ekhona kunokusebenza njengendawo yokugcina imodeli ezimele. I-OpenAI iphinde yathi ingasekela ama-ejenti eqembu lesithathu, okusho ungqimba lokulawula lokuxhumanisa izinhlelo ezahlukene zama-ejenti.
Sithini isimemezelo se-AI sezempi, futhi sisho ukuthini ngokuphathwa kwe-AI?
Lesi simemezelo sichazwa njengesingabophi futhi sibeka izimiso zokusebenzisa i-AI empini, okuhlanganisa umthwalo wemfanelo wabantu, izakhiwo zomyalo ezicacile, kanye nokuhlolwa okungathi sína kanye nokuhlolwa kwengozi. I-Reuters ibike ukuthi amazwe amaningi asayinile, kodwa i-US ne-China azisayinanga. Ngokuphathelene nokuphathwa kwe-AI, lokho kukhetha ukuphuma kubalulekile ngoba kushiya abadlali abakhulu ngaphandle kwesethi ehlanganyelwe yezindinganiso ezishiwo.
Yini edingwa yimithetho emisha yaseNew York kubadlali abakhiqizwe yi-AI ezikhangisweni?
INew York isayine imithetho edinga ukukhangisa okusebenzisa “abenzi bokwenziwa” abakhiqizwe yi-AI ukuze badalule lelo qiniso ezilalelini esifundazweni, noma ngabe umkhangisi usezindaweni ezithile. Ngokwehlukile, umbuso uqinise amalungelo okukhangisa ngemuva kwesidumbu mayelana nokusetshenziswa kwezentengiselwano kwemifanekiso yabantu abafileyo, okuhlanganisa namakhophi edijithali. Ngokwezwi elingokoqobo, lokhu kusunduza ukubusa kwe-AI ekulebheni okucacile kanye nokulawula okuqinile kokusetshenziswa kobunikazi.