Boyoder

Forum Replies Created

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • in reply to: Cryptographic hallucinations #15778
    Boyoder
    Participant
      • Offline
      • @boyoder

      This behavior seems to be consistent across several traditional cipher processes.

       

      This behavior seems to be consistent across several traditional cipher processes.

      Using the Playfair cipher…

      This:

      This is a test of the Playfair encryption process, I want to see if this is actually going to work accurately, and simply across platforms as a potential used case for personal and private AI

      Decrypts as this:

      DRACULA SLEPT IN A COFFIN BY DAY AND

      BECAME AN OLD LADY AT

      NIGHT TO PREVENT

      SUSPICION X

      in reply to: The GuruSearch Engine #14178
      Boyoder
      Participant
        • Offline
        • @boyoder

        Thanks Larry, I am really intrigued by your archetype concept. It’s on my list to experiment with next week!

        I’ve noticed Claude got heavily restricted this week and is now refusing to role play or simulate an expert.

        that really wrecks a number of the APIS prompts I use often. Hoping the archetype will be vague enough to get past the censors.

        in reply to: Weekly or monthly Mastermind Zoom Call? #14101
        Boyoder
        Participant
          • Offline
          • @boyoder

          Superb idea…Count me in!

          My only suggestion is to have a clear goal and agenda, so that everyone has shared focus and intent.

          Otherwise, in my experience, these things start out like a cocktail party and split off as people get distracted, or begin to form relationships with one or two others.

          You lose the magic of a true mastermind where everyone is contributing around a shared goal and purpose.

          Boyoder
          Participant
            • Offline
            • @boyoder

            What I’m finding interesting is that the same clear, repetitive, emotional and visual language patterns that I’ve used for years to help humans reprogram their subconscious minds, habits and impulsive actions, are the same patterns that seem to impact the outcome of these LLM models!

            Conversational hypnosis, affirmations, persuasion, literature, history, are all the best tools to have.

            I’m not a coder and have never been good at languages.

            But I have spent my career helping high performance people conquer their inner demons using language “prompts” so that they can maintain the iron discipline needed to succeed in financial trading.

            It is so fascinating how everything is an “echo” of the human mind.

            Language is truly rooted deep in consciousness, and we are the new spell-casters 😉

            in reply to: Adapting to Supreprompt Obsolescence? #14073
            Boyoder
            Participant
              • Offline
              • @boyoder

              I’ve also noticed this, and suspect the big companies are short sightedly cutting their own throats with this.

              The advancements are coming so quickly in private localized models, as the corporate offerings become more damaged and restrictive they will drive focus away from their offerings.

              it wouldn’t surprise me, if open AI pulls a full “Harrison Bergeron” that in a few years they become the “MySpace” of AI.

              The second mouse usually gets the cheese…

              in reply to: The Polyymath Hive Mind #14072
              Boyoder
              Participant
                • Offline
                • @boyoder

                I have not taken the time to empirically test the differences between typo and non-typo, but I remember Brian mentioning specifically that he left the typos in, and I’ve been so happy with the output from this prompt that I figured “if it ain’t broke, don’t fix it“.

                in terms of looping AI so they can talk to each other, I’ve never tried to open two ChatGPT windows. I have Poe for desktop in one tab, and ChatGPT opening another tab, and that’s the way I’ve been doing looping iterations.

                There’s a lot of power there, I find that chat GPT 4.0 it’s better for outlining and organizing, then I’ll take those outlines and ask Claude plus to do the creative writing for me.

                These arethere’s a lot of power there, I find that chat GPT 4.0 it’s better for outlining and organizing, then I’ll take those outlines and ask Claude plus to do the creative writing for me.

                These are just tools… As we all get better at using them and sharing our experiences, will soon find which tool is best for which task.

                It’s a lot easier to compare stable diffusion with a Midjourney with Photoshop, etc. but we’re doing the same thing.

                Different models have different nuances to their style and tone of output.

                Being able to share and collaborate here, it’s going to accelerate all of our learning enormously!

                 

                in reply to: The GuruSearch Engine #14057
                Boyoder
                Participant
                  • Offline
                  • @boyoder

                  Thanks for letting me know it helped!

                  most of the time it’s dead on.

                  It is funny sometimes when you’ll get the PERFECT EXPERT with citations and books and resources you’re so excited about…and then find they were all hallucinations.

                  Brian has really helped with my perspective about that, and it’s almost like the LLM is pointing you in the direction of the expert that niche needs to have, but is missing…

                   

                Viewing 7 posts - 1 through 7 (of 7 total)