[et_pb_section fb_built=\u00a0\u00bb1″ _builder_version=\u00a0\u00bb4.16″ da_disable_devices=\u00a0\u00bboff|off|off\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb da_is_popup=\u00a0\u00bboff\u00a0\u00bb da_exit_intent=\u00a0\u00bboff\u00a0\u00bb da_has_close=\u00a0\u00bbon\u00a0\u00bb da_alt_close=\u00a0\u00bboff\u00a0\u00bb da_dark_close=\u00a0\u00bboff\u00a0\u00bb da_not_modal=\u00a0\u00bbon\u00a0\u00bb da_is_singular=\u00a0\u00bboff\u00a0\u00bb da_with_loader=\u00a0\u00bboff\u00a0\u00bb da_has_shadow=\u00a0\u00bbon\u00a0\u00bb][et_pb_row _builder_version=\u00a0\u00bb4.16″ background_size=\u00a0\u00bbinitial\u00a0\u00bb background_position=\u00a0\u00bbtop_left\u00a0\u00bb background_repeat=\u00a0\u00bbrepeat\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb][et_pb_column type=\u00a0\u00bb4_4″ _builder_version=\u00a0\u00bb4.16″ custom_padding=\u00a0\u00bb|||\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb custom_padding__hover=\u00a0\u00bb|||\u00a0\u00bb][et_pb_text _builder_version=\u00a0\u00bb4.27.4″ background_size=\u00a0\u00bbinitial\u00a0\u00bb background_position=\u00a0\u00bbtop_left\u00a0\u00bb background_repeat=\u00a0\u00bbrepeat\u00a0\u00bb hover_enabled=\u00a0\u00bb0″ global_colors_info=\u00a0\u00bb{}\u00a0\u00bb sticky_enabled=\u00a0\u00bb0″]<\/p>\n
Apr\u00e8s l\u2019enthousiasme initial, le premier sentiment qui saisit les exp\u00e9rimentateurs de ChatGPT est\u2026 la d\u00e9sillusion, devant la masse d\u2019informations erron\u00e9es qu\u2019on le voit \u00e9noncer avec un aplomb complet pour satisfaire son interlocuteur (https:\/\/www.numerama.com\/tech\/1251982-ne-faites-pas-confiance-a-chatgpt-il-ment-tout-le-temps.html<\/a>). Rappelons aussi le fiasco tr\u00e8s co\u00fbteux de la d\u00e9mo de Bard de Google sur la pr\u00e9tendue premi\u00e8re photo d\u2019exoplan\u00e8te par le t\u00e9lescope James Webb\u2026 qui n\u2019en est pas l\u2019auteur (https:\/\/www.bbc.com\/news\/business-64576225<\/a>).<\/p>\n Aux fausses informations g\u00e9n\u00e9r\u00e9es par des algorithmes d\u2019IA, non intentionnelles, on pourrait opposer, bien plus inqui\u00e9tantes, les hordes de deepfakes (https:\/\/fr.wikipedia.org\/wiki\/Deepfake<\/a>), qui risquent de faire passer les \u00ab\u202ffake news\u202f\u00bb de nuisances mineures \u00e0 armes de destructions massives\u202f; citons l\u2019exemple connu de Barak Obama d\u00e9clamant des insultes contre Donald Trump.<\/p>\n Quant \u00e0 Twitter, l\u2019\u00e9chec de son \u00ab\u202fbadge bleu\u202f\u00bb qui permet \u00e0 n\u2019importe qui d\u2019usurper une identit\u00e9 faute de v\u00e9rification, participe \u00e0 la longue d\u00e9ch\u00e9ance de la plate-forme (https:\/\/www.slate.fr\/story\/241904\/service-premium-twitter-blue-elon-musk-echec-abonnement-payant-badge-bleu-certification-reseau-social<\/a>).<\/p>\n Quel point commun peut-on mettre en avant sur ces probl\u00e9matiques\u202f? Celui de la\u00a0confiance en la source<\/strong>. Dans le cas des propos erron\u00e9s de ChatGPT, l\u2019\u00e9metteur de l\u2019information n\u2019est pas identifi\u00e9. Dans le cas des deepfakes, on tente d\u2019usurper l\u2019identit\u00e9 de cet \u00e9metteur. Quant \u00e0 Twitter, on y donne l\u2019illusion de garantir l\u2019identit\u00e9\u2026sans aucune v\u00e9rification r\u00e9elle.<\/p>\n Or, avoir un moyen d\u2019authentifier de fa\u00e7on fiable et s\u00e9curis\u00e9e la source d\u2019une information ou d\u2019un contenu num\u00e9rique, est un sujet \u00e9pineux connu depuis longtemps par les archivistes et notamment trait\u00e9 par le r\u00e8glement eIDAS (https:\/\/www.ssi.gouv.fr\/administration\/reglementation\/confiance-numerique\/le-reglement-eidas\/#:~:text=Le%20r%C3%A8glement%20eIDAS%20s\u2019applique,march%C3%A9%20de%20la%20confiance%20num%C3%A9rique<\/a>). Cela part d\u2019un constat simple\u202f: archiver des documents \u00e9lectroniques, c\u2019est bien, les signer \u00e9lectroniquement pendant l\u2019archivage, c\u2019est mieux,\u00a0mais v\u00e9rifier leur authenticit\u00e9 avant de les archiver<\/strong>, c\u2019est encore mieux\u202f! C\u2019est l\u00e0 qu\u2019intervient le complexe processus de\u00a0v\u00e9rification de signature \u00e9lectronique<\/strong>, incluant par exemple le contr\u00f4le de non-r\u00e9vocation des certificats, qu\u2019il faut pratiquer en amont de l\u2019archivage d\u2019un document.<\/p>\n Ce sujet regorge n\u00e9anmoins de paradoxes. Citons ce paragraphe, qui m\u2019a particuli\u00e8rement marqu\u00e9, de l\u2019\u00e9tude \u00ab\u202fInt\u00e9grit\u00e9, signature et processus d\u2019archivage\u202f\u00bb de Fran\u00e7oise Banat-Berger et Anne Canteaut (https:\/\/www.rocq.inria.fr\/secret\/Anne.Canteaut\/Publications\/BaCa07.pdf<\/a>):<\/p>\n \u00abDemande-t-on aujourd\u2019hui \u00e0 un archiviste de v\u00e9rifier les signatures manuscrites qui figurent sur les documents qu\u2019il re\u00e7oit ? ce qu\u2019on demande \u00e0 l\u2019archiviste, c\u2019est de garantir le processus d\u2019archivage y compris paradoxalement pour les documents qui seraient des faux : l\u2019archiviste doit pouvoir prouver qu\u2019il a parfaitement conserv\u00e9 des faux !\u202f\u00bb<\/em><\/p>\n Pour revenir sur notre parall\u00e8le avec ChatGPT ou des deep fakes, il s\u2019agirait de prouver de fa\u00e7on infalsifiable que l\u2019information donn\u00e9e proviendrait bien de la NASA, ou que la vid\u00e9o de Barak Obama montre bien le r\u00e9el pr\u00e9sident des USA (on peut citer l\u2019article \u00ab\u202fProtecting World Leaders Against Deep Fakes\u202f\u00bb qui \u00e9voque une solution de d\u00e9tection bas\u00e9e sur une signature biom\u00e9trique des personnes\u202f:https:\/\/openaccess.thecvf.com\/content_CVPRW_2019\/papers\/Media%20Forensics\/Agarwal_Protecting_World_Leaders_Against_Deep_Fakes_CVPRW_2019_paper.pdf<\/a>). Le m\u00e9canisme de cha\u00eene de certificats de la signature \u00e9lectronique pourrait tr\u00e8s bien \u00eatre une source d\u2019inspiration\u202f!<\/p>\n Supposons qu\u2019on a r\u00e9gl\u00e9 le probl\u00e8me de l\u2019authenticit\u00e9. Tout comme les archives, authentifier la source ne signifie pas que cette source a \u00e9mis des propos v\u00e9ridiques. On connait les limites de \u00ab\u202fl\u2019argument d\u2019autorit\u00e9\u202f\u00bb. Ainsi, les prix Nobel ayant \u00ab\u202fd\u00e9raill\u00e9\u202f\u00bb vers des th\u00e9ories farfelues ou des pseudo-sciences sont si nombreux qu\u2019ils ont \u00e9t\u00e9 regroup\u00e9s sous le vocable de \u00ab\u202fla maladie du Nobel\u202f\u00bb (https:\/\/www.lalibre.be\/planete\/sciences-espace\/2022\/02\/10\/la-maladie-du-nobel-quand-les-scientifiques-deraillent-UZ32J66SORBJ5MSWMXYKKISPJI\/<\/a>).<\/p>\n Il s\u2019agit donc de compl\u00e9ter cette\u00a0preuve d\u2019authenticit\u00e9<\/strong>, dont le m\u00e9canisme resterait \u00e0 construire, par une\u00a0preuve de v\u00e9racit\u00e9<\/strong>. Le \u00ab\u202ffact checking\u202f\u00bb en serait une brique. Toutefois, il faut prendre conscience que la science du \u00ab\u202fveracity assessment\u202f\u00bb n\u2019en est qu\u2019\u00e0 ses balbutiements (comme le souligne la publication\u00a0https:\/\/www.sciencedirect.com\/science\/article\/pii\/S0167923619301617<\/a>). Contrairement \u00e0 la signature \u00e9lectronique, ce n\u2019est pas un aspect qui semble encore pr\u00e9occuper les producteurs d\u2019archives num\u00e9riques. Mais jusqu\u2019\u00e0 quand\u202f?<\/p>\n Une autre piste possible est le contr\u00f4le humain \u00e0 la soumission des archives. C\u2019est l\u00e0 que la notion de workflow en entr\u00e9e, qui ne permet l\u2019entr\u00e9e dans le syst\u00e8me d\u2019archivage que si un validateur humain a explicitement accept\u00e9 l\u2019archive, peut jouer un r\u00f4le.<\/p>\n On peut r\u00eaver qu\u2019une synergie f\u00e9conde se mette progressivement en place dans le domaine de l\u2019information num\u00e9rique sur le th\u00e8me de la confiance. C\u2019est peut-\u00eatre m\u00eame la condition de la survie du secteur!<\/p>\n [\/et_pb_text][\/et_pb_column][\/et_pb_row][et_pb_row column_structure=\u00a0\u00bb1_2,1_2″ _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb module_alignment=\u00a0\u00bbcenter\u00a0\u00bb custom_margin=\u00a0\u00bb2%||||false|false\u00a0\u00bb custom_padding=\u00a0\u00bb|2%||2%|false|false\u00a0\u00bb border_width_all=\u00a0\u00bb2px\u00a0\u00bb border_color_all=\u00a0\u00bb#2EA3F2″ global_colors_info=\u00a0\u00bb{}\u00a0\u00bb locked=\u00a0\u00bboff\u00a0\u00bb][et_pb_column type=\u00a0\u00bb1_2″ _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb][et_pb_text _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb hover_enabled=\u00a0\u00bb0″ global_colors_info=\u00a0\u00bb{}\u00a0\u00bb sticky_enabled=\u00a0\u00bb0″]<\/p>\n Pour obtenir des informations d\u00e9taill\u00e9es, t\u00e9l\u00e9chargez gratuitement nos documentations techniques.<\/strong><\/p>\n<\/div>\n<\/div>\n [\/et_pb_text][et_pb_button button_url=\u00a0\u00bbhttps:\/\/infotel-software.com\/fr\/documentations\/\u00a0\u00bb button_text=\u00a0\u00bbT\u00e9l\u00e9charger la documentation\u00a0\u00bb button_alignment=\u00a0\u00bbcenter\u00a0\u00bb _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb][\/et_pb_button][\/et_pb_column][et_pb_column type=\u00a0\u00bb1_2″ _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb][et_pb_text _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb]<\/p>\n Vous avez un projet ? Nos experts sont l\u00e0 pour vous accompagner.\u00a0<\/strong>Cliquez ci-dessous pour nous contacter.<\/strong><\/p>\n<\/div>\n<\/div>\n [\/et_pb_text][et_pb_button button_url=\u00a0\u00bbhttps:\/\/infotel-software.com\/fr\/nous-contacter\/\u00a0\u00bb button_text=\u00a0\u00bbContactez-nous\u00a0\u00bb button_alignment=\u00a0\u00bbcenter\u00a0\u00bb _builder_version=\u00a0\u00bb4.27.4″ _module_preset=\u00a0\u00bbdefault\u00a0\u00bb global_colors_info=\u00a0\u00bb{}\u00a0\u00bb][\/et_pb_button][\/et_pb_column][\/et_pb_row][\/et_pb_section]<\/p>\n","protected":false},"excerpt":{"rendered":" Apr\u00e8s l\u2019enthousiasme initial, le premier sentiment qui saisit les exp\u00e9rimentateurs de ChatGPT est\u2026 la d\u00e9sillusion, devant la masse d\u2019informations erron\u00e9es qu\u2019on le voit \u00e9noncer avec un aplomb complet pour satisfaire son interlocuteur (https:\/\/www.numerama.com\/tech\/1251982-ne-faites-pas-confiance-a-chatgpt-il-ment-tout-le-temps.html). Rappelons aussi le fiasco tr\u00e8s co\u00fbteux de la d\u00e9mo de Bard de Google sur la pr\u00e9tendue premi\u00e8re photo d\u2019exoplan\u00e8te par le t\u00e9lescope […]<\/p>\n","protected":false},"author":2,"featured_media":20297,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_et_pb_use_builder":"on","_et_pb_old_content":" After the initial enthusiasm, the first feeling that seizes those trying their hand at ChatGPT is\u2026 disillusionment, in front of the mass of erroneous information that we see it stating with complete aplomb to satisfy its interlocutor (https:\/\/flyingbisons.com\/blog\/hallucinations-of-chatgpt-4-even-the-most-powerful-tool-has-a-weakness<\/a>). Let us also recall the very expensive fiasco of Google Bard\u2019s demo on the supposed first photo of an exoplanet by the James Webb telescope\u2026 the source of the photo being from elsewhere (https:\/\/www.bbc.com\/news\/business-64576225<\/a>).<\/p> As for the non-intentionally false information generated by AI algorithms, we can stack up hordes of much more worrying deepfakes (https:\/\/en.wikipedia.org\/wiki\/Deepfake<\/a>). These risk the spread of \u201cfake news\u201d from minor nuisances to major destruction. One such known example shows Barak Obama proclaiming insults against Donald Trump.<\/p> As for Twitter (or now X), the failure of the \u201cblue badge\u201d system now allows anyone to usurp an identity due to lack of verification. This is adding to the slow decline of the platform. (https:\/\/slate.com\/technology\/2023\/04\/elon-musk-twitter-blue-check-marks-verification-lebron-james.html<\/a>).<\/p> What common point can we highlight on these issues? That of\u00a0trust in the source<\/strong>. In the case of ChatGPT\u2019s erroneous comments, the sender of the information is not identified. In the case of deepfakes, the identity of the subject is usurped. As for Twitter, it gives the illusion of guaranteeing identity\u2026without any real verification.<\/p> However, having a means of reliably and securely authenticating the source of information or digital content is a thorny subject known for a long time by archivists and notably addressed by the eIDAS regulation: (https:\/\/www.ssi.gouv.fr\/en\/regulation\/digital-confidence\/the-eidas-regulation\/<\/a>). It starts from a simple observation: archiving electronic documents is good, signing them electronically during archiving is better, but checking their authenticity before archiving them is even better! This is where the complex electronic signature verification process comes in, including, for example, checking the non-revocation of certificates, which must be carried out before archiving a document.<\/p> However, this subject is full of paradoxes. Let us quote this paragraph, which particularly struck me, from the study \u201cIntegrity, signature and archiving process\u201d by Fran\u00e7oise Banat-Berger and Anne Canteaut: (https:\/\/www.rocq.inria.fr\/secret\/Anne.Canteaut\/Publications\/BaCa07.pdf<\/a>):<\/p> \u201cDo we today ask an archivist to check the handwritten signatures that appear on the documents he receives? What we ask of the archivist is to guarantee the archiving process, including paradoxically for documents that are forgeries: the archivist must be able to prove that he has perfectly preserved forgeries!\u201d<\/em><\/p> To return to our parallel with ChatGPT or deep fakes, it would be a question of proving in an unfalsifiable way that the information given would indeed come from NASA, or that the video of Barak Obama shows the real president of the USA (we can cite the article \u201cProtecting World Leaders Against Deep Fakes\u201d which discusses a detection solution based on a biometric signature of people:<\/p> https:\/\/openaccess.thecvf.com\/content_CVPRW_2019\/papers\/Media%20Forensics\/Agarwal_Protecting_World_Leaders_Against_Deep_Fakes_CVPRW_2019_paper.pdf<\/a>). The certificate chain mechanism of the electronic signature could very well be a source of inspiration!<\/p> Suppose we have solved the problem of authenticity. Just like the archives, authenticating the source does not mean that this source has made truthful statements. We know the limits of the \u201cargument from authority\u201d. Thus, Nobel Prize winners who have \u201cderailed\u201d towards crazy theories or pseudo-sciences are so numerous that they have been grouped under the term \u201cNobel disease\u201d:\u00a0https:\/\/en.wikipedia.org\/wiki\/Nobel_disease<\/a>).<\/p>