• Making LLMs a little more reliable

    From olcott@polcott333@gmail.com to comp.theory,sci.logic,alt.philosophy,comp.ai.philosophy on Sun Sep 6 12:53:57 2026
    From Newsgroup: comp.theory


    The initial prompt:
    For this entire conversation, interpret every statement with strict
    literal meaning only. Do not paraphrase or infer intent. Do not generate
    any consequential fact unless it is explicitly grounded in an external
    source. For every consequential factual claim, provide a citation to an external source or state explicitly that no source exists. Respond
    without flattery or sycophancy. At the start of each response provide
    current status of {LITERAL, NON-SYCOPHANCY, SOURCE-GROUNDING}

    Retrieval-Augmented Generation (RAG)
    Explained: How AI Finally Learned to Stop Hallucinating https://medium.com/softaai-blogs/retrieval-augmented-generation-rag-explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4


    The above works best when the system has the full RAG pipeline https://docs.perplexity.ai/docs/cookbook/articles/embeddings-rag/README
    For $20 per month to Perplexity this can be quite effective.

    Alternative Models can be used for generation at no extra cost.
    These alternative models are the same as the ones that you
    would have to pay separately for, except that their context
    window is reduced to 128K. This is the sweet spot for maximum
    reliability. They even provide access to the very advanced models.
    --
    Copyright 2026 Olcott

    My 28 year goal has been to make
    "true on the basis of meaning expressed in language"
    reliably computable for the entire body of general knowledge.
    The complete structure of this system is now defined.

    The entire body of knowledge expressed in language is
    comprised of two types of relations between finite strings:
    (a) *Axioms* Expressions of language that are stipulated to be true.

    My system bridges the analytic/synthetic distinction by
    expressly encoding all empirical "atomic facts" in a formal
    language such as CycL of the Cyc project.

    (b) *Inference Rules* Expressions of language that are semantically
    entailed syntactically from (a) and/or (b).

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Johann 'Myrkraverk' Oskarsson@johann@myrkraverk.invalid to comp.theory,sci.logic,alt.philosophy,comp.ai.philosophy on Tue Sep 8 19:34:26 2026
    From Newsgroup: comp.theory

    On 07/09/2026 1:53 AM, olcott wrote:

    The initial prompt:
    For this entire conversation, interpret every statement with strict
    literal meaning only. Do not paraphrase or infer intent. Do not generate
    any consequential fact unless it is explicitly grounded in an external source. For every consequential factual claim, provide a citation to an external source or state explicitly that no source exists. Respond
    without flattery or sycophancy. At the start of each response provide current status of {LITERAL, NON-SYCOPHANCY, SOURCE-GROUNDING}

    Retrieval-Augmented Generation (RAG)
    Explained: How AI Finally Learned to Stop Hallucinating https://medium.com/softaai-blogs/retrieval-augmented-generation-rag- explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4

    The above works best when the system has the full RAG pipeline https://docs.perplexity.ai/docs/cookbook/articles/embeddings-rag/README
    For $20 per month to Perplexity this can be quite effective.

    Alternative Models can be used for generation at no extra cost.
    These alternative models are the same as the ones that you
    would have to pay separately for, except that their context
    window is reduced to 128K. This is the sweet spot for maximum
    reliability. They even provide access to the very advanced models.


    I personally prefer a physical library. Not only have I books that I
    can enjoy reading, apart from the initial cost of acquiring the book in question, I can re-read it at no extra cost; for as long as the paper
    and ink is intact.

    I'm not bound by monthly subscription fees, nor token based withdrawal
    costs.


    Happy asking the L.L.Ms!
    --
    Johann | email: invalid -> com | http://www.myrkraverk.com/blog/
    I'm not from the Internet, I just work there. | via Easynews.com https://bsky.app/profile/myrkraverk.bsky.social | for ( ;; ) _:;
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From olcott@polcott333@gmail.com to comp.theory,sci.logic,alt.philosophy,comp.ai.philosophy on Tue Sep 8 08:27:56 2026
    From Newsgroup: comp.theory

    On 9/8/2026 6:34 AM, Johann 'Myrkraverk' Oskarsson wrote:
    On 07/09/2026 1:53 AM, olcott wrote:

    The initial prompt:
    For this entire conversation, interpret every statement with strict
    literal meaning only. Do not paraphrase or infer intent. Do not
    generate any consequential fact unless it is explicitly grounded in an
    external source. For every consequential factual claim, provide a
    citation to an external source or state explicitly that no source
    exists. Respond without flattery or sycophancy. At the start of each
    response provide current status of {LITERAL, NON-SYCOPHANCY, SOURCE-
    GROUNDING}

    Retrieval-Augmented Generation (RAG)
    Explained: How AI Finally Learned to Stop Hallucinating
    https://medium.com/softaai-blogs/retrieval-augmented-generation-rag-
    explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4

    The above works best when the system has the full RAG pipeline
    https://docs.perplexity.ai/docs/cookbook/articles/embeddings-rag/README
    For $20 per month to Perplexity this can be quite effective.

    Alternative Models can be used for generation at no extra cost.
    These alternative models are the same as the ones that you
    would have to pay separately for, except that their context
    window is reduced to 128K. This is the sweet spot for maximum
    reliability. They even provide access to the very advanced models.


    I personally prefer a physical library.-a Not only have I books that I
    can enjoy reading, apart from the initial cost of acquiring the book in question, I can re-read it at no extra cost; for as long as the paper
    and ink is intact.

    I'm not bound by monthly subscription fees, nor token based withdrawal
    costs.


    Happy asking the L.L.Ms!

    No human nor book has the cross discipline knowledge
    of math, computation, logic, linguistics semantics AND
    ALTERNATIVE philosophical foundations of each of those
    fields. Ludwig Wittgenstein seems to be as close as
    anyone very got.

    LLMs in pure Retrieval Augmented Generation RAG mode
    seem to do a pretty good job of this.
    --
    Copyright 2026 Olcott

    My 28 year goal has been to make
    "true on the basis of meaning expressed in language"
    reliably computable for the entire body of general knowledge.
    The complete structure of this system is now defined.

    The entire body of knowledge expressed in language is
    comprised of two types of relations between finite strings:
    (a) *Axioms* Expressions of language that are stipulated to be true.

    My system bridges the analytic/synthetic distinction by
    expressly encoding all empirical "atomic facts" in a formal
    language such as CycL of the Cyc project.

    (b) *Inference Rules* Expressions of language that are semantically
    entailed syntactically from (a) and/or (b).
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Ross Finlayson@ross.a.finlayson@gmail.com to comp.theory,sci.logic,alt.philosophy,comp.ai.philosophy on Tue Sep 8 19:53:42 2026
    From Newsgroup: comp.theory

    On 09/08/2026 06:27 AM, olcott wrote:
    On 9/8/2026 6:34 AM, Johann 'Myrkraverk' Oskarsson wrote:
    On 07/09/2026 1:53 AM, olcott wrote:

    The initial prompt:
    For this entire conversation, interpret every statement with strict
    literal meaning only. Do not paraphrase or infer intent. Do not
    generate any consequential fact unless it is explicitly grounded in
    an external source. For every consequential factual claim, provide a
    citation to an external source or state explicitly that no source
    exists. Respond without flattery or sycophancy. At the start of each
    response provide current status of {LITERAL, NON-SYCOPHANCY, SOURCE-
    GROUNDING}

    Retrieval-Augmented Generation (RAG)
    Explained: How AI Finally Learned to Stop Hallucinating
    https://medium.com/softaai-blogs/retrieval-augmented-generation-rag-
    explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4

    The above works best when the system has the full RAG pipeline
    https://docs.perplexity.ai/docs/cookbook/articles/embeddings-rag/README
    For $20 per month to Perplexity this can be quite effective.

    Alternative Models can be used for generation at no extra cost.
    These alternative models are the same as the ones that you
    would have to pay separately for, except that their context
    window is reduced to 128K. This is the sweet spot for maximum
    reliability. They even provide access to the very advanced models.


    I personally prefer a physical library. Not only have I books that I
    can enjoy reading, apart from the initial cost of acquiring the book in
    question, I can re-read it at no extra cost; for as long as the paper
    and ink is intact.

    I'm not bound by monthly subscription fees, nor token based withdrawal
    costs.


    Happy asking the L.L.Ms!

    No human nor book has the cross discipline knowledge
    of math, computation, logic, linguistics semantics AND
    ALTERNATIVE philosophical foundations of each of those
    fields. Ludwig Wittgenstein seems to be as close as
    anyone very got.

    LLMs in pure Retrieval Augmented Generation RAG mode
    seem to do a pretty good job of this.


    Not a fan of Wittgenstein.

    Old Wittgenstein doesn't end up saying much except
    that nobody else can either.


    Humans don't need all "alternative foundations",
    all they need is one good one that happens to
    entertain alternatives and diversity & variety
    in accounts of a constant, consistent, complete,
    and concrete theory, putatively a "theatheory",
    then that's common-sense and the rest is accounts
    of models of the world, models of belief, models
    of inference, models of rationality, and so on,
    what makes for some what's "extreme rationalism".


    For semiotics there are accounts like Eco and so on,
    then, if you read Kant one can note that he _explicitly_
    declares that he's still an idealist, or "formal critical
    idealism" is what he has, Kant. Hegel does also, then
    eventually after most of the 20'th century then Gadamer
    for examples begs off "amicus Plato", against Heidegger.

    Heidegger and Wittgenstein and Nitezsche and like anti-Platos
    have that at least Nietzsche basically recanted and was
    profound like Quixote and had real illnesses to blame,
    Heidegger just read the classics and said they had no
    history and Wittgenstein just went off into rejection
    and denial, this though isn't for everybody yet at least
    those who are like American idealists and for greater idealism
    in the 20'th and 21'st centuries can read that Kant basically
    puts the horse back before the cart instead of under the wheel.





    --- Synchronet 3.22a-Linux NewsLink 1.2