\n<\/p>\n
Anthropic’in \u00dcst Y\u00f6neticisi Amodei de \u00e7evrim i\u00e7i kat\u0131ld\u0131\u011f\u0131 toplant\u0131da, bu ola\u011fan\u00fcst\u00fc ilerlemenin y\u00f6netilmesi gereken riskleri de beraberinde getirdi\u011fini anlatt\u0131.<\/p>\n
\u00a0<\/p>\n
Amodei, “\u0130ki ana risk kategorisi g\u00f6r\u00fcyorum. Birincisi, yapay zeka modellerinin k\u00f6t\u00fcye kullan\u0131lmas\u0131, \u00f6rne\u011fin biyoter\u00f6ristler taraf\u0131ndan biyolojik silah \u00fcretmek i\u00e7in kullan\u0131lmas\u0131. \u0130kinci risk ise kontrol kayb\u0131d\u0131r; model yeteneklerinin, geli\u015ftiricilerin kontrol edebilme kapasitesinin \u00f6tesinde h\u0131zlanmas\u0131d\u0131r.” diye konu\u015ftu.<\/p>\n
\u00a0<\/p>\n
Bu risklerin t\u00fcm insanl\u0131\u011f\u0131 ilgilendiren k\u00fcresel meseleler oldu\u011funu dile getiren Amodei, “K\u00f6t\u00fc y\u00f6netilmesi halinde yapay zekan\u0131n b\u00fct\u00fcn olarak insanl\u0131k i\u00e7in risk olu\u015fturabilece\u011fine inan\u0131yorum.” \u015feklinde konu\u015ftu.<\/p>\n
\u00a0<\/p>\n
Yapay zeka modellerinin risklere kar\u015f\u0131 test edilmesi i\u00e7in ortak standartlar\u0131n belirlenmesini isteyen Anthropic CEO’su Amodei, “Bunun bug\u00fcn d\u00fcnyan\u0131n kar\u015f\u0131 kar\u015f\u0131ya oldu\u011fu en \u00f6nemli k\u00fcresel g\u00fcvenlik sorunu oldu\u011funa inan\u0131yorum. Bu odadaki insanlar\u0131n t\u00fcm farkl\u0131l\u0131klar\u0131na ra\u011fmen, ayn\u0131 anda bize sunulan bu k\u00fcresel f\u0131rsat ve k\u00fcresel tehditle y\u00fczle\u015fmek i\u00e7in bu farkl\u0131l\u0131klar\u0131 bir kenara b\u0131rakmal\u0131y\u0131z. Hi\u00e7bir lider, hi\u00e7bir \u015firket ve hi\u00e7bir ulus bunu tek ba\u015f\u0131na ba\u015faramaz.” de\u011ferlendirmesini yapt\u0131.<\/p>\n
\u00a0<\/p>\n
Hugging Face \u00dcst Y\u00f6neticisi Clement Delangue ise otonom bir yapay zeka ajan\u0131 taraf\u0131ndan ger\u00e7ekle\u015ftirilen siber sald\u0131r\u0131y\u0131 d\u00fcnyaya a\u00e7\u0131k \u015fekilde ilk kez duyuran \u015firket olarak bu sald\u0131r\u0131dan \u00e7\u0131kard\u0131klar\u0131 dersleri anlatt\u0131. Yapay zekada siber g\u00fcvenlik tehditlerini daha iyi anlamak ve bu tehditleri azaltmak i\u00e7in k\u00fcresel \u00f6l\u00e7ekte daha g\u00fc\u00e7l\u00fc standartlara ihtiya\u00e7 oldu\u011funu s\u00f6yleyen Delangue, yapay zekan\u0131n kapasitesinin, kaynaklar\u0131n\u0131n ve kontrol\u00fcn\u00fcn birka\u00e7 ki\u015finin elinde yo\u011funla\u015fmas\u0131 yerine bunun da\u011f\u0131t\u0131lmas\u0131n\u0131n b\u00fcy\u00fck bir ihtiya\u00e7 haline geldi\u011fini vurgulad\u0131.<\/p>\n<\/div>\n