{"id":39996,"date":"2026-09-09T15:15:54","date_gmt":"2026-09-09T14:15:54","guid":{"rendered":"https:\/\/by-tv.co.uk\/?p=39996"},"modified":"2026-09-09T15:15:54","modified_gmt":"2026-09-09T14:15:54","slug":"anthropic-arastirmacisi-evan-hubinger-yapay-zekanin-tum-insanlari-yok-etme-ihtimali-%10dan-fazla","status":"publish","type":"post","link":"https:\/\/by-tv.co.uk\/en\/anthropic-arastirmacisi-evan-hubinger-yapay-zekanin-tum-insanlari-yok-etme-ihtimali-%2510dan-fazla\/","title":{"rendered":"Anthropic ara\u015ft\u0131rmac\u0131s\u0131 Evan Hubinger “Yapay zekan\u0131n ‘t\u00fcm insanlar\u0131 yok etme ihtimali %10’dan fazla”"},"content":{"rendered":"

Amerikan yapay zek\u00e2 \u015firketi Anthropic\u2019in g\u00fcvenlik ara\u015ft\u0131rmac\u0131lar\u0131ndan Evan Hubinger<\/strong>, yapay zek\u00e2 teknolojisinin h\u0131zl\u0131 geli\u015fimine ili\u015fkin dikkat \u00e7ekici bir uyar\u0131da bulundu. Hubinger, gelecek 10 y\u0131l i\u00e7inde yapay zek\u00e2n\u0131n \u201ct\u00fcm insanlar\u0131 \u00f6ld\u00fcrmesi\u201d ihtimalini y\u00fczde 10\u2019dan fazla<\/strong> g\u00f6rd\u00fc\u011f\u00fcn\u00fc a\u00e7\u0131klad\u0131.<\/p>\n

Hubinger, X \u00fczerinden yapt\u0131\u011f\u0131 de\u011ferlendirmede mevcut yapay zek\u00e2 modellerinin olu\u015fturdu\u011fu riskin d\u00fc\u015f\u00fck oldu\u011funu belirtti. Ancak teknolojinin h\u0131zla ilerlemesi ve kendi geli\u015fimini h\u0131zland\u0131rabilecek kapasiteye ula\u015fmas\u0131 halinde risklerin ciddi bi\u00e7imde artabilece\u011fini savundu.<\/p>\n

Hubinger, yapay zek\u00e2n\u0131n insanl\u0131\u011f\u0131n yok olmas\u0131na hangi mekanizmalar \u00fczerinden yol a\u00e7abilece\u011fine ili\u015fkin ayr\u0131nt\u0131l\u0131 bir senaryo payla\u015fmad\u0131.<\/p>

\"\"<\/a><\/div>\n

\u201cS\u00fcper zek\u00e2 i\u00e7in hen\u00fcz bir plan\u0131m\u0131z yok\u201d<\/h4>\n

Yapay zek\u00e2 sistemlerini insan de\u011ferleri ve hedefleriyle uyumlu tutmaya odaklanan \u201cAI alignment\u201d alan\u0131nda \u00e7al\u0131\u015fan Hubinger, s\u00fcper zek\u00e2 d\u00fczeyindeki sistemleri g\u00fcvenli bi\u00e7imde kontrol etmeye y\u00f6nelik sorunun hen\u00fcz \u00e7\u00f6z\u00fclmedi\u011fini<\/strong> ifade etti.<\/p>\n

Hubinger, Anthropic\u2019in g\u00fcvenlik konusunda elinden geleni yapmaya \u00e7al\u0131\u015ft\u0131\u011f\u0131na inand\u0131\u011f\u0131n\u0131 ancak s\u00fcper zek\u00e2 i\u00e7in uyum sorununu \u00e7\u00f6zecek bir plan\u0131n bulunmad\u0131\u011f\u0131n\u0131 ve bu alanda net ilerleme kaydedilmedi\u011fini s\u00f6yledi.<\/p>\n

Eski Anthropic ve OpenAI ara\u015ft\u0131rmac\u0131s\u0131 Jacob Coxon<\/strong> da yapay zek\u00e2 \u015firketlerinin yeterince sorumlu davranmad\u0131\u011f\u0131n\u0131 \u00f6ne s\u00fcrd\u00fc. Coxon, gelecekteki sistemlerin siber g\u00fcvenlikten bilimsel ara\u015ft\u0131rmalara kadar bir\u00e7ok alanda insan\u00fcst\u00fc yeteneklere ula\u015fabilece\u011fini savundu.<\/p>\n

Anthropic hakk\u0131nda AISI iddias\u0131<\/h4>\n

BBC’nin Financial Times\u2019dan aktard\u0131\u011f\u0131na g\u00f6re Anthropic\u2019in en yeni modelini Birle\u015fik Krall\u0131k Yapay Zek\u00e2 G\u00fcvenli\u011fi Enstit\u00fcs\u00fc\u2019ne (AISI) de\u011ferlendirme amac\u0131yla sunmad\u0131\u011f\u0131 \u00f6ne s\u00fcr\u00fcld\u00fc. Anthropic iddiaya ili\u015fkin yorum yapmad\u0131.<\/p>\n

Birle\u015fik Krall\u0131k h\u00fck\u00fcmeti de s\u00f6z konusu modelin AISI\u2019den gizlenip gizlenmedi\u011fine ili\u015fkin do\u011frudan a\u00e7\u0131klama yapmazken, Anthropic dahil sekt\u00f6r temsilcileriyle modellerin g\u00fcvenli\u011fini art\u0131rmak i\u00e7in \u00e7al\u0131\u015fmalar\u0131n s\u00fcrd\u00fc\u011f\u00fcn\u00fc bildirdi.<\/p>\n

Cambridge \u00dcniversitesi Makine \u00d6\u011frenimi Profes\u00f6r\u00fc Neil Lawrence<\/strong> ise BBC Radio 4\u2019\u00fcn Today program\u0131nda, ABD\u2019nin yapay zek\u00e2 alan\u0131n\u0131 giderek \u00c7in ile stratejik bir rekabet olarak de\u011ferlendirmesinin uluslararas\u0131 i\u015fbirli\u011fini etkileyebilece\u011fini s\u00f6yledi.<\/p>\n

\u015eirketler de riskleri de\u011ferlendiriyor<\/h4>\n

Anthropic\u2019in A\u011fustos 2026 tarihli g\u00fcvenlik de\u011ferlendirmesinde, geli\u015fmi\u015f modellerin kurumlar\u0131n hedefleriyle uyumsuz hale gelerek sistemleri istismar etmesi veya \u201cyapay zek\u00e2 kaynakl\u0131 felaket d\u00fczeyinde zararlara\u201d yol a\u00e7mas\u0131 riskinin d\u00fc\u015f\u00fck oldu\u011fu belirtildi. Ancak \u015firket, bu de\u011ferlendirmeye duydu\u011fu g\u00fcvenin \u00f6nceki d\u00f6neme g\u00f6re azald\u0131\u011f\u0131n\u0131 da kaydetti.<\/p>\n

Yapay zek\u00e2n\u0131n olu\u015fturabilece\u011fi varolu\u015fsal riskler konusunda teknoloji sekt\u00f6r\u00fcnde g\u00f6r\u00fc\u015f ayr\u0131l\u0131klar\u0131 s\u00fcr\u00fcyor. OpenAI, Google DeepMind ve Anthropic y\u00f6neticileri<\/strong> de 2023 y\u0131l\u0131nda ileri d\u00fczey yapay zek\u00e2dan kaynaklanabilecek a\u011f\u0131r risklerin k\u00fcresel \u00f6l\u00e7ekte ele al\u0131nmas\u0131 gerekti\u011fini a\u00e7\u0131klam\u0131\u015ft\u0131.<\/p>\n

\"\"<\/a><\/div>","protected":false},"excerpt":{"rendered":"

Amerikan yapay zek\u00e2 \u015firketi Anthropic\u2019in g\u00fcvenlik ara\u015ft\u0131rmac\u0131lar\u0131ndan Evan Hubinger, yapay zek\u00e2 teknolojisinin h\u0131zl\u0131 geli\u015fimine ili\u015fkin dikkat \u00e7ekici bir uyar\u0131da bulundu. Hubinger, gelecek 10 y\u0131l i\u00e7inde yapay zek\u00e2n\u0131n \u201ct\u00fcm insanlar\u0131 \u00f6ld\u00fcrmesi\u201d ihtimalini y\u00fczde 10\u2019dan fazla g\u00f6rd\u00fc\u011f\u00fcn\u00fc a\u00e7\u0131klad\u0131. Hubinger, X \u00fczerinden yapt\u0131\u011f\u0131 de\u011ferlendirmede mevcut yapay zek\u00e2 modellerinin olu\u015fturdu\u011fu riskin d\u00fc\u015f\u00fck oldu\u011funu belirtti. Ancak teknolojinin h\u0131zla ilerlemesi ve kendi geli\u015fimini h\u0131zland\u0131rabilecek kapasiteye ula\u015fmas\u0131 halinde risklerin ciddi bi\u00e7imde artabilece\u011fini savundu. Hubinger, yapay zek\u00e2n\u0131n insanl\u0131\u011f\u0131n yok olmas\u0131na hangi mekanizmalar \u00fczerinden yol a\u00e7abilece\u011fine ili\u015fkin ayr\u0131nt\u0131l\u0131 bir senaryo payla\u015fmad\u0131. \u201cS\u00fcper zek\u00e2 i\u00e7in hen\u00fcz bir plan\u0131m\u0131z yok\u201d Yapay zek\u00e2 sistemlerini insan de\u011ferleri ve hedefleriyle uyumlu tutmaya odaklanan \u201cAI […]<\/p>\n","protected":false},"author":11,"featured_media":33756,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ngg_post_thumbnail":0,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-39996","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-genel"],"_links":{"self":[{"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/posts\/39996","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/users\/11"}],"replies":[{"embeddable":true,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/comments?post=39996"}],"version-history":[{"count":1,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/posts\/39996\/revisions"}],"predecessor-version":[{"id":39997,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/posts\/39996\/revisions\/39997"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/media\/33756"}],"wp:attachment":[{"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/media?parent=39996"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/categories?post=39996"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/by-tv.co.uk\/en\/wp-json\/wp\/v2\/tags?post=39996"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}