{"id":136678,"date":"2023-07-08T10:59:11","date_gmt":"2023-07-08T08:59:11","guid":{"rendered":"https:\/\/www.pauljorion.com\/blog\/?p=136678"},"modified":"2023-07-08T10:59:11","modified_gmt":"2023-07-08T08:59:11","slug":"une-colle-lorang-outang-dont-les-gardiens-font-ce-que-lui-veut","status":"publish","type":"post","link":"https:\/\/www.pauljorion.com\/blog\/2023\/07\/08\/une-colle-lorang-outang-dont-les-gardiens-font-ce-que-lui-veut\/","title":{"rendered":"<b>Une colle : \u00ab\u00a0L&rsquo;orang-outang dont les gardiens font ce que lui veut\u00a0\u00bb<\/b>"},"content":{"rendered":"<p><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-136704\" src=\"https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6.jpg\" alt=\"\" width=\"768\" height=\"768\" srcset=\"https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6.jpg 768w, https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6-300x300.jpg 300w, https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6-150x150.jpg 150w\" sizes=\"auto, (max-width: 768px) 100vw, 768px\" \/><\/p>\n<p><em>Illustration Stable Diffusion (+PJ)<\/em> <\/p>\n<p>Dans un tweet, il y a 3 jours, Sam Altman, le patron d&rsquo;OpenAI, la compagnie derri\u00e8re ChatGPT, appelait cela \u00ab\u00a0superalignement\u00a0\u00bb : comment faire pour que les objectifs d&rsquo;une Intelligence Artificielle restent align\u00e9s sur ceux d&rsquo;\u00eatres humains qui seraient \u00e9ventuellement (tousse ! tousse !) beaucoup moins intelligents qu&rsquo;elle ?<\/p>\n<blockquote class=\"twitter-tweet\">\n<p dir=\"ltr\" lang=\"en\">a lot of compute dedicated to research focused on figuring out how to make superintelligence safe. please consider joining the team! <a href=\"https:\/\/t.co\/ljyLjQ9W7d\">https:\/\/t.co\/ljyLjQ9W7d<\/a><\/p>\n<p>\u2014 Sam Altman (@sama) <a href=\"https:\/\/twitter.com\/sama\/status\/1676643268241801216?ref_src=twsrc%5Etfw\">July 5, 2023<\/a><\/p><\/blockquote>\n<p><script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p>Formul\u00e9 de cette mani\u00e8re, comme \u00ab\u00a0superalignement\u00a0\u00bb d&rsquo;une IA super-intelligente, le probl\u00e8me reste tr\u00e8s abstrait. D&rsquo;o\u00f9 l&rsquo;int\u00e9r\u00eat de le reformuler dans le cadre d&rsquo;un casse-t\u00eate classique&nbsp;: celui de l&rsquo;orang-outang qui aimerait anticiper le comportement de ses gardiens complotant de le changer de cage mais qui est trop b\u00eate pour interpr\u00e9ter leurs conciliabules.<\/p>\n<p>Essayons donc de r\u00e9soudre le casse-t\u00eate du \u00ab\u00a0superalignement\u00a0\u00bb sur un cas plus facile \u00e0 se repr\u00e9senter&nbsp;: \u00ab\u00a0Que pourrait faire un orang-outang pour que ses gardiens fassent ce que lui veut&nbsp;?\u00a0\u00bb.<\/p>\n","protected":false},"excerpt":{"rendered":"<p><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-full wp-image-136704\" src=\"https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6.jpg\" alt=\"\" width=\"768\" height=\"768\" srcset=\"https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6.jpg 768w, https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6-300x300.jpg 300w, https:\/\/www.pauljorion.com\/blog\/wp-content\/uploads\/Stable-diffusion-rodins_thinker_except_he_is_an_orang_utang_ZhNESJrQgd6np2NfQ3T2_6-150x150.jpg 150w\" sizes=\"auto, (max-width: 768px) 100vw, 768px\" \/><\/p>\n<p><em>Illustration Stable Diffusion (+PJ)<\/em> <\/p>\n<p>Dans un tweet, il y a 3 jours, Sam Altman, le patron d&rsquo;OpenAI, la compagnie derri\u00e8re ChatGPT, appelait cela \u00ab\u00a0superalignement\u00a0\u00bb : comment faire pour que les objectifs d&rsquo;une Intelligence Artificielle restent align\u00e9s sur ceux d&rsquo;\u00eatres humains qui seraient \u00e9ventuellement (tousse ! [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[9084,940,7669,9153,9280],"class_list":["post-136678","post","type-post","status-publish","format-standard","hentry","category-intelligence-artificielle","tag-chatgpt","tag-intelligence-artificielle-2","tag-openai","tag-sam-altman","tag-superalignement"],"_links":{"self":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts\/136678","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/comments?post=136678"}],"version-history":[{"count":8,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts\/136678\/revisions"}],"predecessor-version":[{"id":136711,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts\/136678\/revisions\/136711"}],"wp:attachment":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/media?parent=136678"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/categories?post=136678"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/tags?post=136678"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}