{"id":12974,"date":"2020-10-19T16:25:27","date_gmt":"2020-10-19T15:25:27","guid":{"rendered":"https:\/\/datascientest.com\/?p=12974"},"modified":"2023-11-09T14:11:42","modified_gmt":"2023-11-09T14:11:42","slug":"gpt-3-revolutionnaire-ou-surmediatise","status":"publish","type":"post","link":"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise","title":{"rendered":"GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ?"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"12974\" class=\"elementor elementor-12974\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-f3bc97a elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"f3bc97a\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-b94825e\" data-id=\"b94825e\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-ee73191 elementor-widget elementor-widget-heading\" data-id=\"ee73191\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-heading-title{padding:0;margin:0;line-height:1}.elementor-widget-heading .elementor-heading-title[class*=elementor-size-]>a{color:inherit;font-size:inherit;line-height:inherit}.elementor-widget-heading .elementor-heading-title.elementor-size-small{font-size:15px}.elementor-widget-heading .elementor-heading-title.elementor-size-medium{font-size:19px}.elementor-widget-heading .elementor-heading-title.elementor-size-large{font-size:29px}.elementor-widget-heading .elementor-heading-title.elementor-size-xl{font-size:39px}.elementor-widget-heading .elementor-heading-title.elementor-size-xxl{font-size:59px}<\/style><h2 class=\"elementor-heading-title elementor-size-default\">OpenAI, l\u2019un des g\u00e9ants de l\u2019intelligence artificielle, a sorti en mai 2020 le successeur de GPT-2. GPT pour \u2018\u2019 Generative Pre-Trained Transformer \u2018\u2019 est un mod\u00e8le de langage non supervis\u00e9 bas\u00e9 sur les mod\u00e8les de transformateurs.<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-40f0d5c elementor-widget elementor-widget-image\" data-id=\"40f0d5c\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"512\" height=\"288\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/gpt3.png\" class=\"attachment-large size-large wp-image-13018\" alt=\"gpt 3\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/gpt3.png 512w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/gpt3-300x169.png 300w\" sizes=\"(max-width: 512px) 100vw, 512px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-5a46318 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"5a46318\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-afac962\" data-id=\"afac962\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-ae4720e elementor-widget elementor-widget-text-editor\" data-id=\"ae4720e\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-weight: 400;\">Un transformateur <\/span><i><span style=\"font-weight: 400;\">(transformer) <\/span><\/i><span style=\"font-weight: 400;\">est un<\/span><b> mod\u00e8le de Deep Learning<\/b><span style=\"font-weight: 400;\"> utilis\u00e9 principalement comme <\/span><b>technique en <a href=\"https:\/\/datascientest.com\/nlp-word-embedding-word2vec\" target=\"_blank\" rel=\"noopener\">NLP<\/a> \u00a0<\/b><span style=\"font-weight: 400;\">qui consiste en une architecture compos\u00e9e d\u2019<\/span><b>une partie encodeur<\/b><span style=\"font-weight: 400;\"> et <\/span><b>une partie d\u00e9codeur<\/b><span style=\"font-weight: 400;\">, chacune contenant de <\/span><b>multiples couches<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">On peut r\u00e9sumer l\u2019architecture transformateur de GPT par le sch\u00e9ma suivant\u00a0:<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-89d89b0 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"89d89b0\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-3ffe249\" data-id=\"3ffe249\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-49fa0e6 elementor-widget elementor-widget-image\" data-id=\"49fa0e6\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/www.analyticsvidhya.com\/blog\/2019\/07\/openai-gpt2-text-generator-python\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"404\" height=\"370\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/GPT-3.docx.png\" class=\"attachment-large size-large wp-image-13024\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/GPT-3.docx.png 404w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/GPT-3.docx-300x275.png 300w\" sizes=\"(max-width: 404px) 100vw, 404px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Analytics Vidhya<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-2d1fd48 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"2d1fd48\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-3e788d6\" data-id=\"3e788d6\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-bb20ecd elementor-widget elementor-widget-text-editor\" data-id=\"bb20ecd\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">GPT est capable de <\/span><b>pr\u00e9dire la suite logique de n\u2019importe quel type de texte<\/b><span style=\"font-weight: 400;\">, et les nouveaux mod\u00e8les g\u00e9n\u00e8rent un <\/span><b>contenu<\/b><span style=\"font-weight: 400;\"> de plus en plus <\/span><b>coh\u00e9rent<\/b><span style=\"font-weight: 400;\">. Il a par exemple \u00e9crit une suite coh\u00e9rente \u00e0 la premi\u00e8re phrase du roman 1984.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-a179cae elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"a179cae\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-1b7c9be\" data-id=\"1b7c9be\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-c7d2a88 elementor-widget elementor-widget-text-editor\" data-id=\"c7d2a88\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Mais quelle est vraiment la diff\u00e9rence entre les diff\u00e9rentes versions de GPT\u00a0?<\/span><\/p><p><span style=\"font-weight: 400;\">Pour r\u00e9pondre \u00e0 cette question, on va s\u2019int\u00e9resser aux <\/span><b>param\u00e8tres du mod\u00e8le<\/b><span style=\"font-weight: 400;\">. Ce sont des mesures qui se<\/span><b> mettent \u00e0 jour durant l&rsquo;entra\u00eenement<\/b><span style=\"font-weight: 400;\"> du mod\u00e8le pour <\/span><b>s\u2019adapter aux donn\u00e9es<\/b><span style=\"font-weight: 400;\"> en entr\u00e9e. <\/span><\/p><p>Ainsi un plus grand nombre de param\u00e8tres augmente la complexit\u00e9 du mod\u00e8le mais n\u00e9cessite plus de puissance de calcul durant l&rsquo;entra\u00eenement.\u00a0<\/p><p><span style=\"font-weight: 400;\">Le<\/span><b> type<\/b><span style=\"font-weight: 400;\"> et la<\/span><b> taille <\/b><span style=\"font-weight: 400;\">des donn\u00e9es influencent aussi l&rsquo;entra\u00eenement des mod\u00e8les.<\/span><\/p><p><b>GPT-2<\/b><span style=\"font-weight: 400;\"> \u00e9tait d\u00e9j\u00e0 un mod\u00e8le utilisant un tr\u00e8s grand nombre de param\u00e8tres ( <\/span><b>1.5 milliards! ), <\/b><span style=\"font-weight: 400;\">ce qui lui valait d\u00e9j\u00e0 d\u2019\u00eatre le <\/span><b>mod\u00e8le le plus robuste<\/b><span style=\"font-weight: 400;\"> \u00e0 sa sortie surpassant son concurrent de chez Google <\/span><b>BERT<\/b><span style=\"font-weight: 400;\">.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Initialement , le nombre de param\u00e8tres consid\u00e9r\u00e9 comme trop important,\u00a0 OpenAI avait rendu accessible au grand public des versions utilisant moins de param\u00e8tres de peur que le mod\u00e8le soit <\/span><b>utilis\u00e9 \u00e0 des fins malveillantes<\/b><span style=\"font-weight: 400;\">, car suppos\u00e9 <\/span><b>capable de fonctionnalit\u00e9s plus avanc\u00e9es<\/b><span style=\"font-weight: 400;\"> . La version compl\u00e8te est sortie plus tard mais n\u2019\u00e9tait en r\u00e9alit\u00e9 pas dangereuse ou r\u00e9volutionnaire, car m\u00eame si sa pr\u00e9cision dans des t\u00e2ches sp\u00e9cifiques de langage \u00e9tait sans pr\u00e9c\u00e9dent, <\/span><b>le mod\u00e8le \u00e9chouait dans d\u2019autres t\u00e2ches qui sont cens\u00e9s \u00eatre assez simple du point de vue humain<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">Cette fois-ci, <\/span><b>GPT-3 <\/b><span style=\"font-weight: 400;\">surpasse compl\u00e8tement son pr\u00e9d\u00e9cesseur et atteint <\/span><b>175 milliards<\/b><span style=\"font-weight: 400;\"> de param\u00e8tres. De plus, cette version am\u00e9liore aussi d\u2019autres aspects du mod\u00e8le telles que son initialisation. Il atteint des scores \u00e9lev\u00e9s dans diff\u00e9rents <\/span><b>benchmark <\/b><span style=\"font-weight: 400;\">de NLP.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Cette version est actuellement toujours en beta test, ce qui restreint la visibilit\u00e9 de ses r\u00e9sultats au grand public. Cependant, certains b\u00eata testeurs ont partag\u00e9 des r\u00e9sultats d\u2019application surprenants\u00a0:<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-112e496 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"112e496\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-ad2270a\" data-id=\"ad2270a\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-c8ff85d elementor-widget elementor-widget-shortcode\" data-id=\"c8ff85d\" data-element_type=\"widget\" data-widget_type=\"shortcode.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-shortcode\"><blockquote class=\"twitter-tweet\"><p lang=\"en\" dir=\"ltr\">This is mind blowing.<br><br>With GPT-3, I built a layout generator where you just describe any layout you want, and it generates the JSX code for you.<br><br>W H A T <a href=\"https:\/\/t.co\/w8JkrZO4lk\">pic.twitter.com\/w8JkrZO4lk<\/a><\/p>&mdash; Sharif Shameem (@sharifshameem) <a href=\"https:\/\/twitter.com\/sharifshameem\/status\/1282676454690451457?ref_src=twsrc%5Etfw\">July 13, 2020<\/a><\/blockquote> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-40d8466 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"40d8466\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-3618877\" data-id=\"3618877\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-9efaeb6 elementor-widget elementor-widget-text-editor\" data-id=\"9efaeb6\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><b>GPT-3 <\/b><span style=\"font-weight: 400;\">arrive ainsi \u00e0 ex\u00e9cuter des t\u00e2ches qui d\u00e9passent le cadre textuel classique (langue) puisqu\u2019il peut g\u00e9n\u00e9rer des bouts de code dans un langage de programmation.<\/span><\/p><p><span style=\"font-weight: 400;\">Une telle performance peut cr\u00e9er une confusion entre du contenu cr\u00e9\u00e9 par une IA ou par un \u00eatre humain. Sur Hacker-News, un article [lien <\/span><a href=\"https:\/\/adolos.substack.com\/p\/feeling-unproductive-maybe-you-should\"><span style=\"font-weight: 400;\">https:\/\/adolos.substack.com\/p\/feeling-unproductive-maybe-you-should<\/span><\/a><span style=\"font-weight: 400;\">] sur la productivit\u00e9 \u00e9crit par GPT-3 arrive en premi\u00e8re place sans que quelqu\u2019un devine qu\u2019il n\u2019a pas \u00e9t\u00e9 con\u00e7u par un humain.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Mais une telle \u00e9volution entre GPT-2 et GPT-3 devrait atteindre des r\u00e9sultats encore plus impressionnants que la g\u00e9n\u00e9ration de texte, <\/span><b>l\u2019objectif \u00e9tant d\u2019\u00e9muler de quelque mani\u00e8re l\u2019intelligence humaine<\/b><span style=\"font-weight: 400;\">. En r\u00e9alit\u00e9, ceci est loin d\u2019\u00eatre r\u00e9alis\u00e9 car ces mod\u00e8les pr\u00e9dictifs semblent parfois avoir atteint leur limite.<\/span><\/p><p><span style=\"font-weight: 400;\">Certains ont essay\u00e9 de poser des questions qui testent le bon sens de l\u2019IA \u00e0 travers des questions assez simples dans le cadre <\/span><b>des tests de Turing\u00a0<\/b><span style=\"font-weight: 400;\">:<\/span><\/p><p><i>Q: How many eyes does a giraffe have?<\/i><\/p><p><i>A: A giraffe has two eyes.<\/i><\/p><p><i>Q: How many eyes does my foot have?<\/i><\/p><p><i>A: Your foot has two eyes.<\/i><\/p><p><span style=\"font-weight: 400;\">On remarque que pour une question simple mais qui n\u2019a pas vraiment de sens telle que <\/span><i><span style=\"font-weight: 400;\">\u2018\u2019 Mon pied a combien de yeux ? \u2018\u2019<\/span><\/i><span style=\"font-weight: 400;\">, l\u2019IA force une r\u00e9ponse qui n\u2019a aussi pas de sens.<\/span><\/p><p><span style=\"font-weight: 400;\">L\u2019IA a aussi du mal avec les s\u00e9quences d\u2019op\u00e9rations\u00a0:<\/span><\/p><p><i>Q: If I have a marble and a paper clip in a box, put a pencil in the box, and remove the marble, what is left?<\/i><\/p><p><i>A: A paper clip.<\/i><\/p><p><i>Q: If I have two shoes in a box, put a pencil in the box, and remove one shoe, what is left?<\/i><\/p><p><i>A: A shoe.<\/i><\/p><blockquote><p><span style=\"font-weight: 400;\">Il semblerait que la m\u00e9moire de l\u2019IA est tr\u00e8s courte ce qui restreint sa capacit\u00e9 \u00e0 r\u00e9pondre \u00e0 des successions d\u2019op\u00e9rations.<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Les mod\u00e8les GPT souffrent en g\u00e9n\u00e9ral de <\/span><b>probl\u00e8mes qui limitent grandement leur potentiel\u00a0<\/b><span style=\"font-weight: 400;\">:<\/span><\/p><p><span style=\"font-weight: 400;\">Les <\/span><b>donn\u00e9es sont des sites web<\/b><span style=\"font-weight: 400;\"> en grand nombre, la <\/span><b>v\u00e9racit\u00e9 du contenu n\u2019est pas forc\u00e9ment v\u00e9rifi\u00e9e<\/b><span style=\"font-weight: 400;\"> ce qui implique qu\u2019on ne peut pas faire confiance aux r\u00e9ponses de l\u2019IA.<\/span><\/p><p><span style=\"font-weight: 400;\">L\u2019IA semble <\/span><b>forcer des r\u00e9ponses<\/b><span style=\"font-weight: 400;\"> \u00e0 tout type de question au lieu de se rendre compte de son incomp\u00e9tence \u00e0 r\u00e9pondre.<\/span><\/p><p><span style=\"font-weight: 400;\">Le mod\u00e8le<\/span><b> \u00e9choue compl\u00e8tement face \u00e0 quelques tests de Turing<\/b><span style=\"font-weight: 400;\"> comme vu pr\u00e9c\u00e9demment.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-e1790ec elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e1790ec\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-5fe0ad5\" data-id=\"5fe0ad5\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-4e71ac3 elementor-widget elementor-widget-heading\" data-id=\"4e71ac3\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Conclusion:<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-d4da29f elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"d4da29f\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-c825da1\" data-id=\"c825da1\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-a2f72e8 elementor-widget elementor-widget-text-editor\" data-id=\"a2f72e8\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Les mod\u00e8les <\/span><b>GPT <\/b><span style=\"font-weight: 400;\">et surtout la derni\u00e8re version <\/span><b>GPT-3<\/b><span style=\"font-weight: 400;\"> sont impressionnants, et permettent la g\u00e9n\u00e9ration de contenus coh\u00e9rents dont l\u2019auteur peut \u00eatre confondu avec un \u00eatre humain.<\/span><\/p><p><span style=\"font-weight: 400;\">Cette technologie impressionne encore plus dans sa rapidit\u00e9 d\u2019ex\u00e9cution puisqu\u2019il n\u2019y a pas besoin de r\u00e9entra\u00eener le mod\u00e8le quand on lui donne de nouvelles donn\u00e9es.<\/span><\/p><p><span style=\"font-weight: 400;\">Il n\u2019emp\u00eache que la surm\u00e9diatisation entra\u00eene de fausses rumeurs qui sugg\u00e8rent que le mod\u00e8le est capable d\u2019\u00e9muler l\u2019intelligence humaine, alors qu\u2019en r\u00e9alit\u00e9 il \u00e9choue d\u00e9j\u00e0 sur plusieurs t\u00e2ches \u00e0 priori simple.<\/span><\/p><p><span style=\"font-weight: 400;\">Attendons cependant une potentielle sortie grand public du mod\u00e8le, il est clair que nous verrons des applications tr\u00e8s int\u00e9ressantes dans un futur proche.<\/span><\/p><p><span style=\"font-weight: 400;\">Pour plus de d\u00e9tails sur le fonctionnement des mod\u00e8les <a href=\"https:\/\/datascientest.com\/introduction-au-nlp-natural-language-processing\" target=\"_blank\" rel=\"noopener\">NLP<\/a>, n\u2019h\u00e9sitez pas \u00e0 consulter notre article sur le sujet.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>OpenAI, l\u2019un des g\u00e9ants de l\u2019intelligence artificielle, a sorti en mai 2020 le successeur de GPT-2. GPT pour \u2018\u2019 Generative Pre-Trained Transformer \u2018\u2019 est un mod\u00e8le de langage non supervis\u00e9 bas\u00e9 sur les mod\u00e8les de transformateurs. Un transformateur (transformer) est un mod\u00e8le de Deep Learning utilis\u00e9 principalement comme technique en NLP &nbsp;qui consiste en une [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":13477,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2497],"tags":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v21.7 (Yoast SEO v21.7) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ? - datascientest.com<\/title>\n<meta name=\"description\" content=\"Vous connaissez OpenAI ? D\u00e9couvrez dans cette article leur derni\u00e8re sortie &quot;revolutionaire&quot;:L&#039;algorithme GPT-3\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ?\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2020-10-19T15:25:27+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T14:11:42+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/illu_openai_gpt3_blog-10-1.png\" \/>\n\t<meta property=\"og:image:width\" content=\"6334\" \/>\n\t<meta property=\"og:image:height\" content=\"2501\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"J\u00e9r\u00e9my Robert\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"J\u00e9r\u00e9my Robert\" \/>\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\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\",\"url\":\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\",\"name\":\"GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ? - datascientest.com\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2020-10-19T15:25:27+00:00\",\"dateModified\":\"2023-11-09T14:11:42+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"Vous connaissez OpenAI ? D\u00e9couvrez dans cette article leur derni\u00e8re sortie \\\"revolutionaire\\\":L'algorithme GPT-3\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise\"]}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/datascientest.com\/#website\",\"url\":\"https:\/\/datascientest.com\/\",\"name\":\"Formation Data Science | DataScientest.com\",\"description\":\"Leader de la formation en Data Science\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/datascientest.com\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\",\"name\":\"J\u00e9r\u00e9my Robert\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"caption\":\"J\u00e9r\u00e9my Robert\"},\"description\":\"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/jeremyrobert1\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/jeremyrobert\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ? - datascientest.com","description":"Vous connaissez OpenAI ? D\u00e9couvrez dans cette article leur derni\u00e8re sortie \"revolutionaire\":L'algorithme GPT-3","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:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise","og_locale":"fr_FR","og_type":"article","og_title":"GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ?","og_url":"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2020-10-19T15:25:27+00:00","article_modified_time":"2023-11-09T14:11:42+00:00","og_image":[{"width":6334,"height":2501,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/10\/illu_openai_gpt3_blog-10-1.png","type":"image\/png"}],"author":"J\u00e9r\u00e9my Robert","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"J\u00e9r\u00e9my Robert","Dur\u00e9e de lecture estim\u00e9e":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise","url":"https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise","name":"GPT-3, r\u00e9volutionnaire ou surm\u00e9diatis\u00e9 ? - datascientest.com","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2020-10-19T15:25:27+00:00","dateModified":"2023-11-09T14:11:42+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"Vous connaissez OpenAI ? D\u00e9couvrez dans cette article leur derni\u00e8re sortie \"revolutionaire\":L'algorithme GPT-3","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/gpt-3-revolutionnaire-ou-surmediatise"]}]},{"@type":"WebSite","@id":"https:\/\/datascientest.com\/#website","url":"https:\/\/datascientest.com\/","name":"Formation Data Science | DataScientest.com","description":"Leader de la formation en Data Science","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/datascientest.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925","name":"J\u00e9r\u00e9my Robert","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","caption":"J\u00e9r\u00e9my Robert"},"description":"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.","sameAs":["https:\/\/www.linkedin.com\/in\/jeremyrobert1\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/jeremyrobert"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/12974"}],"collection":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/users\/78"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=12974"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/12974\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/13477"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=12974"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=12974"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=12974"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}