{"id":6203,"date":"2023-10-07T17:15:19","date_gmt":"2023-10-07T17:15:19","guid":{"rendered":"https:\/\/dailyai.com\/?p=6203"},"modified":"2023-10-07T22:49:19","modified_gmt":"2023-10-07T22:49:19","slug":"can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior","status":"publish","type":"post","link":"https:\/\/dailyai.com\/fr\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","title":{"rendered":"L'\"IA constitutionnelle\" peut-elle r\u00e9soudre le probl\u00e8me du comportement probl\u00e9matique de l'IA ?"},"content":{"rendered":"

Alors que les mod\u00e8les d'IA continuent de s'int\u00e9grer dans notre vie quotidienne, les inqui\u00e9tudes concernant les limites et la fiabilit\u00e9 de ce que l'on appelle les \"garde-fous\" se multiplient.<\/b><\/p>\n

Les mod\u00e8les d'IA omnipr\u00e9sente tels que GPT-3.5\/4\/4V et autres sont dot\u00e9s de garde-fous et de mesures de s\u00e9curit\u00e9 int\u00e9gr\u00e9s qui les emp\u00eachent de produire des r\u00e9sultats illicites, contraires \u00e0 l'\u00e9thique ou ind\u00e9sirables. <\/span><\/p>\n

Cependant, ces dispositifs de s\u00e9curit\u00e9 sont loin d'\u00eatre imperm\u00e9ables et certains mod\u00e8les prouvent qu'ils peuvent se d\u00e9tacher de leurs garde-fous ou d\u00e9railler, pour ainsi dire. <\/span><\/p>\n

Le probl\u00e8me r\u00e9side en partie dans le fait que les garde-fous ne suivent pas le rythme de la complexit\u00e9 et de la diversit\u00e9 des mod\u00e8les.\u00a0<\/span><\/p>\n

Ces derni\u00e8res semaines, OpenAI, soutenue par Microsoft, a r\u00e9v\u00e9l\u00e9 des am\u00e9liorations majeures dans ChatGPT, lui permettant d'interagir en utilisant uniquement la voix et de r\u00e9pondre \u00e0 des requ\u00eates par le biais d'images et de textes. Cette version multimodale de GPT-4, capable de traiter des images, a \u00e9t\u00e9 baptis\u00e9e \"GPT-4V\".<\/span><\/p>\n

Parall\u00e8lement, Meta a annonc\u00e9 le d\u00e9ploiement d'un assistant d'IA, de plusieurs personnalit\u00e9s chatbot c\u00e9l\u00e9brit\u00e9s<\/a> pour les utilisateurs de WhatsApp et d'Instagram, ainsi qu'une multitude d'autres fonctionnalit\u00e9s d'IA discr\u00e8tes comme les autocollants AI.\u00a0<\/span><\/p>\n

Les gens ont rapidement manipul\u00e9 les <\/span>Autocollants AI \u00e0 g\u00e9n\u00e9rer<\/span><\/a> des images comiques et choquantes de type bande dessin\u00e9e, comme Karl Marx nu ou Mario avec un fusil d'assaut.\u00a0<\/span><\/p>\n

Alors que la course \u00e0 la commercialisation de l'IA s'intensifie, les garanties con\u00e7ues pour contr\u00f4ler les comportements de l'IA - et l'emp\u00eacher de g\u00e9n\u00e9rer des contenus nuisibles, de la d\u00e9sinformation ou de contribuer \u00e0 des activit\u00e9s illicites - s'av\u00e8rent plus fragiles.\u00a0<\/span><\/p>\n

L'IA constitutionnelle est-elle la solution ?<\/span><\/h2>\n

Pour lutter contre ce ph\u00e9nom\u00e8ne, les entreprises de d\u00e9veloppement de l'IA s'efforcent de cr\u00e9er des \"constitutions de l'IA\", un ensemble de principes et de valeurs fondamentaux auxquels les mod\u00e8les d'IA doivent adh\u00e9rer. La startup Anthropique<\/a> a \u00e9t\u00e9 l'un des premiers \u00e0 pr\u00f4ner l'\"IA constitutionnelle\" dans une <\/span>2022 papier<\/span><\/a>.<\/span><\/p>\n

Google DeepMind a \u00e9galement \u00e9tabli des r\u00e8gles constitutionnelles pour son chatbot. Moineau en 2022<\/a> d'entretenir des conversations \"utiles, correctes et inoffensives\".\u00a0<\/span><\/p>\n

Les constitutions de l'IA d'Anthropic tirent leurs principes de diverses sources, dont la D\u00e9claration des droits de l'homme des Nations unies et les conditions d'utilisation d'Apple. Le mod\u00e8le est dot\u00e9 de principes moraux fondamentaux qui orientent le comportement de la base vers le sommet plut\u00f4t que d'imposer des garde-fous du haut vers le bas.\u00a0<\/span><\/p>\n

Au lieu de former laborieusement l'IA \u00e0 l'aide d'innombrables exemples de bien ou de mal fournis par l'homme, cette approche int\u00e8gre un ensemble de r\u00e8gles ou de principes - une \"constitution\" - auxquels l'IA se conforme.<\/p>\n

Dans un premier temps, l'IA est confront\u00e9e \u00e0 une situation, puis elle est invit\u00e9e \u00e0 critiquer sa r\u00e9ponse et, enfin, \u00e0 affiner son comportement en fonction de la solution r\u00e9vis\u00e9e.<\/p>\n

Ensuite, le syst\u00e8me passe \u00e0 la phase d'apprentissage par renforcement. Ici, il \u00e9value la qualit\u00e9 de ses propres r\u00e9ponses, en distinguant la meilleure. Au fil du temps, cette auto-\u00e9valuation affine son comportement.<\/p>\n

La particularit\u00e9 est que l'IA utilise sa boucle de r\u00e9troaction pour d\u00e9terminer la r\u00e9compense selon une m\u00e9thode appel\u00e9e \"RL \u00e0 partir de la r\u00e9troaction de l'IA\" (RLAIF). Lorsqu'elle est confront\u00e9e \u00e0 des requ\u00eates potentiellement nuisibles ou trompeuses, l'IA ne se contente pas d'esquiver ou de refuser. Au contraire, elle aborde la question de front, en expliquant pourquoi une telle demande peut \u00eatre probl\u00e9matique.<\/p>\n

Il s'agit d'un pas en avant dans la cr\u00e9ation de machines qui non seulement calculent, mais aussi \"pensent\" de mani\u00e8re structur\u00e9e.<\/p>\n

Dario Amodei, PDG et cofondateur d'Anthropic, a soulign\u00e9 la difficult\u00e9 de comprendre le fonctionnement interne des mod\u00e8les d'IA. Il sugg\u00e8re qu'une constitution rendrait les r\u00e8gles transparentes et explicites, afin que tous les utilisateurs sachent \u00e0 quoi s'attendre.\u00a0<\/span><\/p>\n

Il est important de noter qu'il offre \u00e9galement un moyen de responsabiliser le mod\u00e8le s'il n'adh\u00e8re pas aux principes \u00e9nonc\u00e9s.<\/span><\/p>\n

Malgr\u00e9 ces efforts, les constitutions d'IA ne sont pas exemptes de d\u00e9fauts, et les mod\u00e8les de d\u00e9veloppeurs tels qu'Anthropic se sont r\u00e9v\u00e9l\u00e9s vuln\u00e9rables \u00e0 l'infection par le virus de la grippe aviaire. <\/span>jailbreaks<\/span><\/a> comme beaucoup d'autres.\u00a0<\/span><\/p>\n

Il n'existe pas de m\u00e9thode universellement accept\u00e9e pour former des mod\u00e8les d'IA s\u00fbrs et \u00e9thiques.<\/span><\/h2>\n

Historiquement, les mod\u00e8les d'IA ont \u00e9t\u00e9 affin\u00e9s \u00e0 l'aide d'une m\u00e9thode appel\u00e9e apprentissage par renforcement par retour d'information humain (RLHF), dans laquelle les r\u00e9ponses de l'IA sont class\u00e9es comme \"bonnes\" ou \"mauvaises\" par de grandes \u00e9quipes d'\u00e9valuateurs humains.\u00a0<\/span><\/p>\n

Bien qu'efficace dans une certaine mesure, cette m\u00e9thode a \u00e9t\u00e9 critiqu\u00e9e pour son manque de pr\u00e9cision et de sp\u00e9cificit\u00e9. Pour garantir l'\u00e9thique et la s\u00e9curit\u00e9 de l'IA, les entreprises explorent d\u00e9sormais d'autres solutions.<\/span><\/p>\n

Par exemple, OpenAI a adopt\u00e9 l'approche \"red-teaming\", en engageant des experts de diff\u00e9rentes disciplines pour tester et identifier les faiblesses de ses mod\u00e8les.<\/span><\/p>\n

Le syst\u00e8me d'OpenAI fonctionne par it\u00e9rations : le mod\u00e8le d'IA produit des r\u00e9sultats, des examinateurs humains \u00e9valuent et corrigent ces r\u00e9sultats sur la base de lignes directrices sp\u00e9cifiques, et le mod\u00e8le apprend \u00e0 partir de ce retour d'information. Les donn\u00e9es d'entra\u00eenement de ces \u00e9valuateurs sont essentielles pour le calibrage \u00e9thique du mod\u00e8le.<\/p>\n

ChatGPT opte souvent pour une r\u00e9ponse conservatrice lorsqu'il est confront\u00e9 \u00e0 des sujets controvers\u00e9s ou sensibles, \u00e9vitant parfois une r\u00e9ponse directe. Cela contraste avec l'IA constitutionnelle, o\u00f9 le mod\u00e8le devrait \u00e9lucider ses r\u00e9serves lorsqu'il est confront\u00e9 \u00e0 des requ\u00eates potentiellement dangereuses, en d\u00e9montrant activement un raisonnement bas\u00e9 sur ses r\u00e8gles fondamentales.<\/p>\n

En substance, alors que ChatGPT s'appuie fortement sur le retour d'information humain pour son orientation \u00e9thique, l'IA constitutionnelle utilise un cadre bas\u00e9 sur des r\u00e8gles \u00e9tablies avec des m\u00e9canismes d'auto-\u00e9valuation et met l'accent sur la transparence du raisonnement.<\/p>\n

En fin de compte, il n'y a probablement pas d'approche unique pour d\u00e9velopper des IA \"s\u00fbres\" - et certains, comme Elon Musk, critiquent la notion d'IA aseptis\u00e9e et \"\u00e9veill\u00e9e\". Des \u00e9tudes ont prouv\u00e9 que<\/a> que m\u00eame les IA constitutionnelles peuvent \u00eatre pirat\u00e9es et manipul\u00e9es pour adopter un comportement impr\u00e9visible.\u00a0<\/span><\/p>\n

Rebecca Johnson, chercheuse en \u00e9thique de l'IA \u00e0 l'universit\u00e9 de Sydney, a soulign\u00e9 que les ing\u00e9nieurs en IA et les informaticiens abordent souvent les probl\u00e8mes dans le but de trouver des solutions d\u00e9finitives, ce qui ne tient pas toujours compte des complexit\u00e9s de la nature humaine.\u00a0<\/span><\/p>\n

\"Nous devons commencer \u00e0 consid\u00e9rer l'IA g\u00e9n\u00e9rative comme une extension de l'homme, elle n'est qu'un autre aspect de l'humanit\u00e9\", a-t-elle d\u00e9clar\u00e9.\u00a0<\/span><\/p>\n

Le contr\u00f4le global de l'IA comme une sorte de simple syst\u00e8me technique ne fera que s'accentuer. <\/span>plus difficile \u00e0 mesure qu'il \u00e9volue<\/span><\/a>Il en va de m\u00eame pour les organismes biologiques que nous sommes. <\/span><\/p>\n

La divergence, provoqu\u00e9e ou non, est peut-\u00eatre in\u00e9vitable. <\/span><\/p>","protected":false},"excerpt":{"rendered":"

Alors que les mod\u00e8les d'IA continuent de s'int\u00e9grer dans notre vie quotidienne, les inqui\u00e9tudes concernant les limites et la fiabilit\u00e9 de leurs \"garde-fous\" se font de plus en plus vives. Les mod\u00e8les d'IA omnipr\u00e9sents tels que GPT-3.5\/4\/4V et autres sont dot\u00e9s de garde-fous et de mesures de s\u00e9curit\u00e9 int\u00e9gr\u00e9s qui les emp\u00eachent de produire des r\u00e9sultats illicites, contraires \u00e0 l'\u00e9thique ou ind\u00e9sirables. Ces dispositifs de s\u00e9curit\u00e9 sont toutefois loin d'\u00eatre imperm\u00e9ables, et les mod\u00e8les prouvent qu'ils peuvent se d\u00e9tacher de leurs garde-fous - ou d\u00e9railler, pour ainsi dire. Le probl\u00e8me r\u00e9side en partie dans le fait que les garde-fous ne suivent pas le rythme de la complexit\u00e9 et de la diversit\u00e9 des mod\u00e8les. Au cours des derni\u00e8res semaines, OpenAI, soutenu par Microsoft, a r\u00e9v\u00e9l\u00e9 des am\u00e9liorations majeures<\/p>","protected":false},"author":2,"featured_media":6204,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[84],"tags":[148,422,93],"class_list":["post-6203","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-anthropic","tag-constitutional-ai","tag-openai"],"yoast_head":"\nCan "constitutional AI" solve the issue of problematic AI behavior? | DailyAI<\/title>\n<meta name=\"description\" content=\"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called "guardrails" are mounting.\" \/>\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\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Can "constitutional AI" solve the issue of problematic AI behavior? | DailyAI\" \/>\n<meta property=\"og:description\" content=\"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called "guardrails" are mounting.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/fr\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2023-10-07T17:15:19+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-10-07T22:49:19+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1000\" \/>\n\t<meta property=\"og:image:height\" content=\"667\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Sam Jeans\" \/>\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=\"Sam Jeans\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\"},\"author\":{\"name\":\"Sam Jeans\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/711e81f945549438e8bbc579efdeb3c9\"},\"headline\":\"Can “constitutional AI” solve the issue of problematic AI behavior?\",\"datePublished\":\"2023-10-07T17:15:19+00:00\",\"dateModified\":\"2023-10-07T22:49:19+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\"},\"wordCount\":915,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"keywords\":[\"Anthropic\",\"Constitutional AI\",\"OpenAI\"],\"articleSection\":[\"Industry\"],\"inLanguage\":\"fr-FR\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\",\"name\":\"Can \\\"constitutional AI\\\" solve the issue of problematic AI behavior? | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"datePublished\":\"2023-10-07T17:15:19+00:00\",\"dateModified\":\"2023-10-07T22:49:19+00:00\",\"description\":\"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called \\\"guardrails\\\" are mounting.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#breadcrumb\"},\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"width\":1000,\"height\":667,\"caption\":\"Anthropic AI\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Can “constitutional AI” solve the issue of problematic AI behavior?\"}]},{\"@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\\\/711e81f945549438e8bbc579efdeb3c9\",\"name\":\"Sam Jeans\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"caption\":\"Sam Jeans\"},\"description\":\"Sam is a science and technology writer who has worked in various AI startups. When he\u2019s not writing, he can be found reading medical journals or digging through boxes of vinyl records.\",\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/in\\\/sam-jeans-6746b9142\\\/\"],\"url\":\"https:\\\/\\\/dailyai.com\\\/fr\\\/author\\\/samjeans\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"L'\"IA constitutionnelle\" peut-elle r\u00e9soudre le probl\u00e8me du comportement probl\u00e9matique de l'IA ? | DailyAI","description":"Alors que les mod\u00e8les d'IA continuent de s'int\u00e9grer dans notre vie quotidienne, les inqui\u00e9tudes concernant les limites et la fiabilit\u00e9 de ce que l'on appelle les \"garde-fous\" se font de plus en plus vives.","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\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","og_locale":"fr_FR","og_type":"article","og_title":"Can \"constitutional AI\" solve the issue of problematic AI behavior? | DailyAI","og_description":"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called \"guardrails\" are mounting.","og_url":"https:\/\/dailyai.com\/fr\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","og_site_name":"DailyAI","article_published_time":"2023-10-07T17:15:19+00:00","article_modified_time":"2023-10-07T22:49:19+00:00","og_image":[{"width":1000,"height":667,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","type":"image\/jpeg"}],"author":"Sam Jeans","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"\u00c9crit par":"Sam Jeans","Dur\u00e9e de lecture estim\u00e9e":"4 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/"},"author":{"name":"Sam Jeans","@id":"https:\/\/dailyai.com\/#\/schema\/person\/711e81f945549438e8bbc579efdeb3c9"},"headline":"Can “constitutional AI” solve the issue of problematic AI behavior?","datePublished":"2023-10-07T17:15:19+00:00","dateModified":"2023-10-07T22:49:19+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/"},"wordCount":915,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","keywords":["Anthropic","Constitutional AI","OpenAI"],"articleSection":["Industry"],"inLanguage":"fr-FR"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","url":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","name":"L'\"IA constitutionnelle\" peut-elle r\u00e9soudre le probl\u00e8me du comportement probl\u00e9matique de l'IA ? | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","datePublished":"2023-10-07T17:15:19+00:00","dateModified":"2023-10-07T22:49:19+00:00","description":"Alors que les mod\u00e8les d'IA continuent de s'int\u00e9grer dans notre vie quotidienne, les inqui\u00e9tudes concernant les limites et la fiabilit\u00e9 de ce que l'on appelle les \"garde-fous\" se font de plus en plus vives.","breadcrumb":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/"]}]},{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","width":1000,"height":667,"caption":"Anthropic AI"},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"Can “constitutional AI” solve the issue of problematic AI behavior?"}]},{"@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\/711e81f945549438e8bbc579efdeb3c9","name":"Sam Jeans","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","caption":"Sam Jeans"},"description":"Sam est un r\u00e9dacteur scientifique et technologique qui a travaill\u00e9 dans diverses start-ups sp\u00e9cialis\u00e9es dans l'IA. Lorsqu'il n'\u00e9crit pas, on peut le trouver en train de lire des revues m\u00e9dicales ou de fouiller dans des bo\u00eetes de disques vinyles.","sameAs":["https:\/\/www.linkedin.com\/in\/sam-jeans-6746b9142\/"],"url":"https:\/\/dailyai.com\/fr\/author\/samjeans\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts\/6203","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\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/comments?post=6203"}],"version-history":[{"count":12,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts\/6203\/revisions"}],"predecessor-version":[{"id":6216,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/posts\/6203\/revisions\/6216"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/media\/6204"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/media?parent=6203"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/categories?post=6203"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/fr\/wp-json\/wp\/v2\/tags?post=6203"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}