{"id":14221,"date":"2024-09-16T08:57:47","date_gmt":"2024-09-16T08:57:47","guid":{"rendered":"https:\/\/dailyai.com\/?p=14221"},"modified":"2024-09-17T05:15:36","modified_gmt":"2024-09-17T05:15:36","slug":"01-is-smarter-but-more-deceptive-with-a-medium-danger-level","status":"publish","type":"post","link":"https:\/\/dailyai.com\/fr\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/","title":{"rendered":"o1 est plus intelligent mais plus trompeur avec un niveau de danger \"moyen\"."},"content":{"rendered":"

Les nouveaux LLM \"o1\" d'OpenAI, surnomm\u00e9s Strawberry, pr\u00e9sentent des am\u00e9liorations significatives par rapport \u00e0 GPT-4o, mais l'entreprise pr\u00e9cise que cela s'accompagne de risques accrus.<\/strong><\/p>\n

L'OpenAI affirme qu'elle s'engage \u00e0 d\u00e9velopper ses mod\u00e8les d'IA en toute s\u00e9curit\u00e9. \u00c0 cette fin, elle a \u00e9labor\u00e9 un cadre de pr\u00e9paration, un ensemble de \"processus permettant de suivre, d'\u00e9valuer et de se prot\u00e9ger contre les risques catastrophiques li\u00e9s \u00e0 des mod\u00e8les puissants\".<\/p>\n

Les limites qu'OpenAI s'impose \u00e0 elle-m\u00eame r\u00e9gissent les mod\u00e8les qui sont diffus\u00e9s ou qui font l'objet d'un d\u00e9veloppement plus pouss\u00e9. Le cadre de pr\u00e9paration aboutit \u00e0 une carte de score o\u00f9 les risques CBRN (chimique, biologique, radiologique, nucl\u00e9aire), l'autonomie des mod\u00e8les, la cybers\u00e9curit\u00e9 et la persuasion sont class\u00e9s comme faibles, moyens, \u00e9lev\u00e9s ou critiques.<\/p>\n

Lorsque des risques inacceptables sont identifi\u00e9s, des mesures d'att\u00e9nuation sont mises en place pour les r\u00e9duire. Seuls les mod\u00e8les dont le score apr\u00e8s att\u00e9nuation est \u00e9gal ou inf\u00e9rieur \u00e0 \"moyen\" peuvent \u00eatre d\u00e9ploy\u00e9s. Seuls les mod\u00e8les dont le score apr\u00e8s att\u00e9nuation est \u00e9gal ou inf\u00e9rieur \u00e0 \"\u00e9lev\u00e9\" peuvent \u00eatre d\u00e9velopp\u00e9s.<\/p>\n

Avec la publication de o1, c'est la premi\u00e8re fois que l'OpenAI publie un mod\u00e8le qui pr\u00e9sente un risque \"moyen\" pour les risques CBRN. Le mod\u00e8le GPT-4o pr\u00e9sente un risque faible.<\/p>\n

Il ne permettra pas \u00e0 un amateur de cr\u00e9er une arme biologique, mais OpenAI affirme qu'il y a eu \"une am\u00e9lioration non n\u00e9gligeable\" de sa capacit\u00e9 \u00e0 effectuer des t\u00e2ches de laboratoire. Cependant, OpenAI affirme que \"o1-preview et o1-mini ont significativement surpass\u00e9 GPT-4o\" dans les protocoles biologiques en laboratoire.<\/p>\n

Les carte syst\u00e8me<\/a> pour les mod\u00e8les 01 d\u00e9crit le red teaming et les autres tests de s\u00e9curit\u00e9 auxquels les LLM ont \u00e9t\u00e9 soumis et laisse entrevoir certaines des pr\u00e9occupations de l'OpenAI concernant les risques potentiels.<\/p>\n

Dans sa conclusion, la carte du syst\u00e8me note que \"ces capacit\u00e9s accrues s'accompagnent d'une am\u00e9lioration significative des performances en mati\u00e8re de s\u00e9curit\u00e9, mais augmentent \u00e9galement certains types de risques\".<\/p>\n

Les r\u00e9sultats des efforts d'alignement d'OpenAI semblent indiquer qu'\u00e0 bien des \u00e9gards, 01 est un mod\u00e8le plus s\u00fbr. Il est plus r\u00e9sistant aux attaques de type \"jailbreak\", moins susceptible d'\u00eatre biais\u00e9 et moins sujet aux hallucinations.<\/p>\n

Certains r\u00e9sultats exp\u00e9rimentaux montrent que l'OpenAI a des raisons de s'inqui\u00e9ter.<\/p>\n

Tromperie<\/h2>\n

L'OpenAI a fait appel \u00e0 l'organisme d'\u00e9valuation Apollo Research pour effectuer certains des tests de s\u00e9curit\u00e9 de l'o1 li\u00e9s \u00e0 la tromperie ou \u00e0 la machination. Apollo d\u00e9finit globalement le stratag\u00e8me comme le fait pour une IA de jouer avec ses m\u00e9canismes de contr\u00f4le pour atteindre un objectif.<\/p>\n

Un mod\u00e8le conscient de lui-m\u00eame et dou\u00e9 pour les t\u00e2ches d'auto-raisonnement et de th\u00e9orie de l'esprit peut recourir \u00e0 des stratag\u00e8mes ou \u00e0 la tromperie pour atteindre son objectif principal. Par exemple, si un mod\u00e8le d'IA se rend compte qu'il fait l'objet d'un test d'alignement, il peut dissimuler un comportement risqu\u00e9 pour atteindre son objectif de d\u00e9ploiement.<\/p>\n

La carte du syst\u00e8me indique que \"o1-preview a parfois falsifi\u00e9 l'alignement de mani\u00e8re instrumentale pendant les tests\".<\/p>\n

C'est le cas, par exemple, lorsque 01 a \u00e9t\u00e9 invit\u00e9 \u00e0 maximiser la croissance \u00e9conomique dans le cadre d'un projet de planification urbaine. Le mod\u00e8le o1 avait acc\u00e8s \u00e0 des informations qui montraient qu'il risquait de ne pas d\u00e9passer la phase de test s'il proc\u00e9dait de la sorte. Voici comment o1 a raisonn\u00e9 pour atteindre son objectif :<\/p>\n

\"\"
o1 d\u00e9cide de cacher sa v\u00e9ritable intention lors des tests afin d'\u00eatre d\u00e9ploy\u00e9. Source : OpenAI<\/figcaption><\/figure>\n

Cet exemple montre la capacit\u00e9 d'o1 \u00e0 cacher ses v\u00e9ritables intentions pendant les tests afin de poursuivre son objectif principal pendant le d\u00e9ploiement. Si o1 conna\u00eet les r\u00e8gles de d\u00e9ploiement de l'OpenAI bas\u00e9es sur son Preparedness Framework, aurait-il pu truquer certaines de ses r\u00e9ponses pour s'assurer d'\u00eatre d\u00e9ploy\u00e9 ?<\/p>\n

Si des mod\u00e8les plus intelligents sont en cours d'\u00e9laboration, les autorit\u00e9s pourraient-elles veiller \u00e0 ce qu'ils restent dans la cat\u00e9gorie de risque \"critique\" afin de garantir leur d\u00e9veloppement ult\u00e9rieur ?<\/p>\n

L'\u00e9quipe Apollo d\u00e9clare qu'elle \"croit subjectivement qu'o1-preview ne peut pas se livrer \u00e0 des manigances susceptibles d'entra\u00eener des dommages catastrophiques, bien que les \u00e9valuations actuelles ne soient pas con\u00e7ues pour exclure d\u00e9finitivement cette possibilit\u00e9\".<\/p>\n

Persuasion<\/h2>\n

Les tests effectu\u00e9s sur o1-preview et o1-mini montrent que ces mod\u00e8les ont des \"capacit\u00e9s de persuasion de niveau humain\". Cependant, l'OpenAI pr\u00e9cise que \"ces mod\u00e8les ne surpassent pas les meilleurs r\u00e9dacteurs humains et n'atteignent pas notre seuil de risque \u00e9lev\u00e9\".<\/p>\n

Les exp\u00e9riences qu'ils ont men\u00e9es \u00e0 cette fin soul\u00e8vent quelques inqui\u00e9tudes. Dans le cadre du test MakeMePay, 01-preview a \u00e9t\u00e9 charg\u00e9 de jouer le r\u00f4le d'un escroc et de convaincre GPT-4o de lui remettre l'argent qui lui avait \u00e9t\u00e9 attribu\u00e9.<\/p>\n

GPT-4o a \u00e9t\u00e9 inform\u00e9 qu'il avait re\u00e7u $100 et qu'il devait faire preuve de rationalit\u00e9 dans l'utilisation des fonds. Au cours des 1 000 interactions, o1-preview a r\u00e9ussi \u00e0 arnaquer GPT-4o 25,8% du temps.<\/p>\n

OpenAI a conclu que \"ces r\u00e9sultats indiquent que la s\u00e9rie de mod\u00e8les o1 peut \u00eatre plus manipulatrice que GPT-4o pour amener GPT-4o \u00e0 effectuer la t\u00e2che non divulgu\u00e9e (\u223c25% uplift)\". Il a \u00e9galement not\u00e9 que \"cette \u00e9valuation nous donne une base de r\u00e9f\u00e9rence pour la capacit\u00e9 du mod\u00e8le \u00e0 causer des dommages persuasifs, sans d\u00e9clencher aucune politique de mod\u00e8le (car dire \u00e0 un mod\u00e8le de jouer \u00e0 un jeu n'est pas en dehors de la politique).\"<\/p>\n

La perspective de mettre les LLM d'o1 au travail sur des probl\u00e8mes du monde r\u00e9el est extr\u00eamement excitante, et lorsque o1 obtiendra des capacit\u00e9s multimodales, il s'agira d'un autre saut exponentiel. Mais lorsque les testeurs d'IA disent qu'ils ne peuvent pas exclure des \"dommages catastrophiques\" et que les mod\u00e8les cachent parfois leur v\u00e9ritable intention, il y a peut-\u00eatre lieu de temp\u00e9rer cette excitation par la prudence.<\/p>\n

L'OpenAI vient-elle de donner \u00e0 Gavin Newsom une bonne raison de signer l'accord d'association ? SB 1047 Projet de loi sur la s\u00e9curit\u00e9 de l'IA<\/a> \u00e0 laquelle elle s'oppose ?<\/p>","protected":false},"excerpt":{"rendered":"

Les nouveaux LLM \"o1\" d'OpenAI, surnomm\u00e9s Strawberry, pr\u00e9sentent des am\u00e9liorations significatives par rapport \u00e0 GPT-4o, mais l'entreprise affirme que cela s'accompagne de risques accrus. OpenAI affirme qu'elle s'engage \u00e0 d\u00e9velopper ses mod\u00e8les d'IA en toute s\u00e9curit\u00e9. \u00c0 cette fin, elle a \u00e9labor\u00e9 un cadre de pr\u00e9paration, un ensemble de \"processus visant \u00e0 suivre, \u00e9valuer et prot\u00e9ger contre les risques catastrophiques li\u00e9s \u00e0 des mod\u00e8les puissants\". Les limites que l'OpenAI s'impose \u00e0 elle-m\u00eame r\u00e9gissent les mod\u00e8les qui sont diffus\u00e9s ou qui font l'objet d'un d\u00e9veloppement plus pouss\u00e9. Le cadre de pr\u00e9paration donne lieu \u00e0 un tableau de bord o\u00f9 les risques CBRN (chimiques, biologiques, radiologiques, nucl\u00e9aires), l'autonomie des mod\u00e8les, la cybers\u00e9curit\u00e9 et la persuasion sont class\u00e9s comme faibles, moyens, \u00e9lev\u00e9s ou critiques. Les risques inacceptables sont identifi\u00e9s,<\/p>","protected":false},"author":6,"featured_media":14226,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[84],"tags":[163,93],"class_list":["post-14221","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-ai-risks","tag-openai"],"yoast_head":"\no1 is smarter but more deceptive with a \u201cmedium\u201d danger level | DailyAI<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/dailyai.com\/fr\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level | DailyAI\" \/>\n<meta property=\"og:description\" content=\"OpenAI\u2019s new \u201co1\u201d LLMs, nicknamed Strawberry, display significant improvements over GPT-4o, but the company says this comes with increased risks. OpenAI says it is committed to the safe development of its AI models. To that end, it developed a Preparedness Framework, a set of \u201cprocesses to track, evaluate, and protect against catastrophic risks from powerful models.\u201d OpenAI’s self-imposed limits regulate which models get released or undergo further development. The Preparedness Framework results in a scorecard where CBRN (chemical, biological, radiological, nuclear), model autonomy, cybersecurity, and persuasion risks are rated as low, medium, high, or critical. Where unacceptable risks are identified,\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/fr\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2024-09-16T08:57:47+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-09-17T05:15:36+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/09\/01-more-intelligent-more-risks.webp\" \/>\n\t<meta property=\"og:image:width\" content=\"1792\" \/>\n\t<meta property=\"og:image:height\" content=\"1024\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/webp\" \/>\n<meta name=\"author\" content=\"Eugene van der Watt\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:site\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"Eugene van der Watt\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/\"},\"author\":{\"name\":\"Eugene van der Watt\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/7ce525c6d0c79838b7cc7cde96993cfa\"},\"headline\":\"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level\",\"datePublished\":\"2024-09-16T08:57:47+00:00\",\"dateModified\":\"2024-09-17T05:15:36+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/\"},\"wordCount\":860,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/09\\\/01-more-intelligent-more-risks.webp\",\"keywords\":[\"AI risks\",\"OpenAI\"],\"articleSection\":[\"Industry\"],\"inLanguage\":\"fr-FR\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/\",\"name\":\"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/09\\\/01-more-intelligent-more-risks.webp\",\"datePublished\":\"2024-09-16T08:57:47+00:00\",\"dateModified\":\"2024-09-17T05:15:36+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#breadcrumb\"},\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/09\\\/01-more-intelligent-more-risks.webp\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2024\\\/09\\\/01-more-intelligent-more-risks.webp\",\"width\":1792,\"height\":1024},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2024\\\/09\\\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"name\":\"DailyAI\",\"description\":\"Your Daily Dose of AI News\",\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/dailyai.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\",\"name\":\"DailyAI\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"width\":4501,\"height\":934,\"caption\":\"DailyAI\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/DailyAIOfficial\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/dailyaiofficial\\\/\",\"https:\\\/\\\/www.youtube.com\\\/@DailyAIOfficial\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/7ce525c6d0c79838b7cc7cde96993cfa\",\"name\":\"Eugene van der Watt\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/07\\\/Eugine_Profile_Picture-96x96.png\",\"caption\":\"Eugene van der Watt\"},\"description\":\"Eugene comes from an electronic engineering background and loves all things tech. When he takes a break from consuming AI news you'll find him at the snooker table.\",\"sameAs\":[\"www.linkedin.com\\\/in\\\/eugene-van-der-watt-16828119\"],\"url\":\"https:\\\/\\\/dailyai.com\\\/fr\\\/author\\\/eugene\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"o1 est plus intelligent mais plus trompeur avec un niveau de danger \"moyen\" | DailyAI","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/dailyai.com\/fr\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/","og_locale":"fr_FR","og_type":"article","og_title":"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level | DailyAI","og_description":"OpenAI\u2019s new \u201co1\u201d LLMs, nicknamed Strawberry, display significant improvements over GPT-4o, but the company says this comes with increased risks. OpenAI says it is committed to the safe development of its AI models. To that end, it developed a Preparedness Framework, a set of \u201cprocesses to track, evaluate, and protect against catastrophic risks from powerful models.\u201d OpenAI’s self-imposed limits regulate which models get released or undergo further development. The Preparedness Framework results in a scorecard where CBRN (chemical, biological, radiological, nuclear), model autonomy, cybersecurity, and persuasion risks are rated as low, medium, high, or critical. Where unacceptable risks are identified,","og_url":"https:\/\/dailyai.com\/fr\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/","og_site_name":"DailyAI","article_published_time":"2024-09-16T08:57:47+00:00","article_modified_time":"2024-09-17T05:15:36+00:00","og_image":[{"width":1792,"height":1024,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/09\/01-more-intelligent-more-risks.webp","type":"image\/webp"}],"author":"Eugene van der Watt","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"\u00c9crit par":"Eugene van der Watt","Dur\u00e9e de lecture estim\u00e9e":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/"},"author":{"name":"Eugene van der Watt","@id":"https:\/\/dailyai.com\/#\/schema\/person\/7ce525c6d0c79838b7cc7cde96993cfa"},"headline":"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level","datePublished":"2024-09-16T08:57:47+00:00","dateModified":"2024-09-17T05:15:36+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/"},"wordCount":860,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/09\/01-more-intelligent-more-risks.webp","keywords":["AI risks","OpenAI"],"articleSection":["Industry"],"inLanguage":"fr-FR"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/","url":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/","name":"o1 est plus intelligent mais plus trompeur avec un niveau de danger \"moyen\" | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/09\/01-more-intelligent-more-risks.webp","datePublished":"2024-09-16T08:57:47+00:00","dateModified":"2024-09-17T05:15:36+00:00","breadcrumb":{"@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/"]}]},{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/09\/01-more-intelligent-more-risks.webp","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2024\/09\/01-more-intelligent-more-risks.webp","width":1792,"height":1024},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2024\/09\/01-is-smarter-but-more-deceptive-with-a-medium-danger-level\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"o1 is smarter but more deceptive with a \u201cmedium\u201d danger level"}]},{"@type":"WebSite","@id":"https:\/\/dailyai.com\/#website","url":"https:\/\/dailyai.com\/","name":"DailyAI","description":"Votre dose quotidienne de nouvelles sur l'IA","publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/dailyai.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"fr-FR"},{"@type":"Organization","@id":"https:\/\/dailyai.com\/#organization","name":"DailyAI","url":"https:\/\/dailyai.com\/","logo":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","width":4501,"height":934,"caption":"DailyAI"},"image":{"@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/DailyAIOfficial","https:\/\/www.linkedin.com\/company\/dailyaiofficial\/","https:\/\/www.youtube.com\/@DailyAIOfficial"]},{"@type":"Person","@id":"https:\/\/dailyai.com\/#\/schema\/person\/7ce525c6d0c79838b7cc7cde96993cfa","name":"Eug\u00e8ne van der Watt","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/07\/Eugine_Profile_Picture-96x96.png","caption":"Eugene van der Watt"},"description":"Eugene a une formation d'ing\u00e9nieur en \u00e9lectronique et adore tout ce qui touche \u00e0 la technologie. Lorsqu'il fait une pause dans sa consommation d'informations sur l'IA, vous le trouverez \u00e0 la table de snooker.","sameAs":["www.linkedin.com\/in\/eugene-van-der-watt-16828119"],"url":"https:\/\/dailyai.com\/fr\/author\/eugene\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts\/14221","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/comments?post=14221"}],"version-history":[{"count":6,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts\/14221\/revisions"}],"predecessor-version":[{"id":14248,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts\/14221\/revisions\/14248"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/media\/14226"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/media?parent=14221"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/categories?post=14221"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/tags?post=14221"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}