{"id":164,"date":"2025-02-08T09:00:00","date_gmt":"2025-02-08T09:00:00","guid":{"rendered":"https:\/\/www.tb-software.ch\/ai\/der-offline-chat-bot-wie-eine-k-i-denken-kann\/"},"modified":"2025-02-08T09:00:00","modified_gmt":"2025-02-08T09:00:00","slug":"der-offline-chat-bot-wie-eine-k-i-denken-kann","status":"publish","type":"post","link":"https:\/\/www.tb-software.ch\/ai\/en\/der-offline-chat-bot-wie-eine-k-i-denken-kann\/","title":{"rendered":"The Offline Chat Bot \u2013 How an AI Can Think"},"content":{"rendered":"

Artificial Intelligence has fascinated me for a long time, especially the possibility of making it usable offline. With my project TheLocalChat<\/em>, a locally running chatbot, I want to demonstrate how an AI processes information even without an internet connection.<\/p>\n

A question for the AI.<\/h3>\n

An example interaction with the AI might look like this:<\/p>\n

The termination clause under \u00a7 573 (2).<\/em><\/p>\n

This isn\u2019t a complete or clear question. Yet the bot manages to provide a meaningful answer. But why?<\/p>\n

How an AI works with vectors<\/h3>\n

Unlike a traditional search engine that only looks for keywords, an AI works with vectors.<\/p>\n

This means:<\/p>\n

    \n
  • Every word and sentence is represented mathematically.<\/li>\n
  • Terms with similar meanings are close together in the vector space.<\/li>\n
  • This way, the AI recognizes connections and can correctly interpret even imprecise queries.<\/li>\n<\/ul>\n

    The PDF documents serving as the basis for the answers are transferred into this vector space. This way, the AI can find relevant information even if the user doesn\u2019t phrase their question perfectly.<\/p>\n

    Traceability through a tree view<\/h3>\n

    In TheLocalChat<\/em> I\u2019ve integrated a tree view to visualize how the AI works.<\/p>\n

      \n
    • The query is broken down and analyzed.<\/li>\n
    • The relevant documents are evaluated and compared.<\/li>\n
    • The most suitable match is determined and displayed.<\/li>\n<\/ul>\n

      This isn\u2019t necessarily the absolute truth but rather the most likely match for your query.<\/p>\n

      Why offline AI?<\/h3>\n

      Unlike conventional chatbots or search engines, my system operates entirely offline. So when I try to look up google.ch<\/em> when I search for something, I get an offline notification.<\/p>\n

      This means the AI works solely with locally stored documents.<\/p>\n

      Which documents are suitable for an offline AI?<\/h3>\n

      To function effectively, an AI requires a solid knowledge base. This can consist of documents such as:<\/p>\n

        \n
      • legal texts<\/li>\n
      • court rulings and legal commentaries<\/li>\n
      • scientific publications<\/li>\n
      • specialized books and technical manuals<\/li>\n
      • company policies and internal documentation<\/li>\n
      • product catalogs and technical specifications<\/li>\n
      • Medical studies and pharmaceutical documentation<\/li>\n
      • patent documents and research reports<\/li>\n<\/ul>\n

        By carefully selecting the documents, the AI can be specialized in specific subject areas.<\/p>\n

        Conclusion<\/h3>\n

        My project demonstrates that an AI can function effectively even without an internet connection. TheLocalChat<\/em> processes texts, establishes connections, and provides answers \u2013 offline<\/p>\n

        I hope you found this glimpse into my project interesting. I\u2019d love to hear your thoughts on offline AI.<\/p>\n","protected":false},"excerpt":{"rendered":"

        Artificial intelligence has fascinated me for a long time, especially the possibility of making it usable offline. With my project TheLocalChat, a locally running chatbot, I want to demonstrate how an AI processes information even without an internet connection. A question for the AI. A typical interaction with the AI could look like this: ‘The termination clause under \u00a7 573\u2026’<\/p>\n","protected":false},"author":1,"featured_media":163,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[37],"tags":[],"class_list":["post-164","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-applikationsentwicklung"],"_links":{"self":[{"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=\/wp\/v2\/posts\/164","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=164"}],"version-history":[{"count":0,"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=\/wp\/v2\/posts\/164\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=\/wp\/v2\/media\/163"}],"wp:attachment":[{"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=164"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=164"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.tb-software.ch\/ai\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=164"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}