Wittgenstein for the enterprise?

[X DdC]

ontolo...@googlegroups.com

Wittgenstein for the enterprise?

Simon Polovina wrote:
     As introduced in previous foundational work (Polovina 2007), CGs
     provide a versatile environment for knowledge representation and
     automated reasoning that directly aligns with enterprise object
     models.

Sure.  This has a Wittgenstein Tractatus flavor in which assertions
just have the true-false feature.  AI-systems have additional
responsibilities: see again my:
     Ontology families for AI-systems
     https://www.ontooo.com/Ontology.pdf
The Language-of-Thought (Fodor, LOT) of AI-systems should handle also
requests for planning, graph search, answer questions about new
mini-stories,  solve analogy challenges, etc.  Pragmatics offers more
modes than assertions & questions.

A sentence showing up in a LOT meaning representation has an
entourage of contextual info - pragmatic type, time, source,
intention, ...
Long, long time ago an assertion whispered the suggestion that it
harbors a vector with as components at least {true-false,
pleasant-unpleasant, right-wrong, ...}

Simon Polovina wrote also:
      The CG Advantage: Conceptual Graphs feature nested contexts as a
      core primitive. A belief state, a temporal validity window, a
      hypothetical "what-if" planning scenario, or an unapproved
      workflow step can be cleanly enclosed within a context
      box. Reasoning rules apply precisely inside specified contexts
      without leaking state into the broader knowledge base.
Didn't find this in Sowa's 1983 Conceptual Structures text.  It
refers, likely/ perhaps, to a newer LOT - meaning representation.

And also:
      4.1.3 Direct Neuro-Symbolic Grounding for AI Agents (Joule & LLM
      Orchestration)
      The RDF Reality: Modern enterprise architectures (like SAP’s
      Business AI Hub and SAP Knowledge Graph) use Knowledge Graphs to
      ground LLM agents and prevent hallucinations. However,
      converting LLM outputs into SPARQL triple patterns is
      notoriously fragile due to predicate drift, URI matching issues,
      and rigid triple structures.
I thought that the real challenge is to feed an LLM (and more likely an
SLM) into a domain specific mini-ontology.

      The CG Advantage: CGs support graph projection—a sound and
      complete mechanism for pattern matching. An LLM acting as an
      abductive engine can propose a tentative solution in Conceptual
      Graph Linear Form (CGLF/CGIF). The engine uses projection to
      test if the proposed structure is a valid specialisation of an
      existing business rule or schema. If the projection fails, the
      logic engine rejects the hypothesis or pinpoints the exact
      ungrounded concept.
No idea how to 'parse' this one.

      4.2 Why Didn’t SAP Do It?
      The barrier has never been theoretical superiority—it has been
      ecosystem momentum. The RDF/SPARQL stack offered off-the-shelf
      open-source parsers, mature triple-store database engines, and
      an established developer base familiar with Semantic Web
      tooling. Implementing CGs natively would have required SAP to
      build and maintain a commercial-grade ISO Common Logic graph
      engine from scratch.
YES ...  i believe.
      However, as enterprise systems pivot toward neuro-symbolic AI
      and local multi-agent orchestration, the architectural overhead
      of triple stores is becoming a bottleneck. Systems designed
      around -ary events, nested contexts, and graph projection are
      proving far better suited for grounding autonomous agents than
      legacy triple stores.
Bingo - i believe.

      5 Formal Logic Representations: CGIF and CLIF
      5.1 Conceptual Graph Interchange Format (CGIF)
      < good stuff >
      5.2 Common Logic Interchange Format (CLIF)
      < good stuff >
Still I wonder whether this pertains only to what is happening in a
Wittgenstein Tractatus cave/ dungeon.

Please set me straight as needed.

dennis de champeaux

[Simon Polovina]

Hi Dennis,

Thank you for the insightful feedback. If I understand you correctly, I'm glad to see we align on the fundamental bottleneck that legacy triple stores present for modern enterprise AI.

Let me try to address the points you raised (as that’ll help me understand, too):

1. Pragmatics, LOT, and "Tractatus" Logic: You make a fair point regarding classical logic's restriction to binary truth-values () versus the full pragmatic entourage required by a Language of Thought (LOT). This is precisely why nested contexts (derived from Peirce's existential graphs) are foundational to Conceptual Graphs. Rather than asserting unconstrained global facts, nested context boxes allow us to cleanly enclose belief states, temporal validity windows, agent intentions, or hypothetical "what-if" scenarios. It moves the representation beyond a flat Tractarian space into a structured, context-aware substrate.
2. Parsing Graph Projection for LLMs/SLMs: To clarify how graph projection acts as a neuro-symbolic filter:
• Generation: An LLM or domain-specific Small Language Model (SLM) proposes a candidate business action expressed in CGLF/CGIF ().
• Projection Test: The engine projects against a validated domain rule or schema graph ().
• Verification: Graph projection deterministically tests structural specialisation (). If projection succeeds, the hypothesis is structurally grounded; if it fails, the engine pinpoints the exact ungrounded concept or invalid relation before any execution occurs.

Thanks again for your observations!

Best regards,

Simon

--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAJ%2B4qGeO94oWA_1X5TLk-b9vvsALd3w6fwovnu8CXSvSGAN%2Beg%40mail.gmail.com.

[X DdC]

Dear Simon Polovina si...@similelogics.ltd via googlegroups.com

Here questions to help me out.
Your:::
     1 Pragmatics, LOT, and "Tractatus" Logic: You make a fair point
     regarding classical logic's restriction to binary truth-values ()

     versus the full pragmatic entourage required by a Language of
     Thought (LOT). This is precisely why nested contexts (derived
     from Peirce's existential graphs) are foundational to Conceptual
     Graphs. Rather than asserting unconstrained global facts, nested
     context boxes allow us to cleanly enclose belief states, temporal
     validity windows, agent intentions, or hypothetical "what-if"
     scenarios. It moves the representation beyond a flat Tractarian
     space into a structured, context-aware substrate.

Are you here describing infrastructure for meaning
representation (LOT) - since you are referring to contexts (a
magical term used by everyone with a different significance/role)?
Yes/ no?
How/where to read up about your:

       This is precisely why nested contexts (derived from Peirce's
       existential graphs) are foundational to Conceptual Graphs.

Your :::
    2 Parsing Graph Projection for LLMs/SLMs: To clarify how graph

    projection acts as a neuro-symbolic filter: Generation: An LLM or
    domain-specific Small Language Model (SLM) proposes a candidate

    business action expressed in CGLF/CGIF ().  Projection Test: The
    engine projects against a validated domain rule or schema graph
    ().  Verification: Graph projection deterministically tests
    structural specialisation (). If projection succeeds, the

    hypothesis is structurally grounded; if it fails, the engine
    pinpoints the exact ungrounded concept or invalid relation before
    any execution occurs.

The key term here is 'candidate business action'. This triggers the
interpretation of a limited scope setting (not the open-ended
chat-systems setting) with perhaps a Small Language Model (SLM). But
why an SLM?  One needs a mapping indeed using ako hashtable from
domain terms to the concepts in a mini-ontology of the application.
This suggests that your:

       Generation: An LLM or domain-specific Small Language Model
       (SLM) proposes a candidate business action expressed in

       CGLF/CGIF ().
refers to a business action represented somewhere in the ontology that
must be mapped onto "CGLF/CGIF ()" whatever that stands for.

From there I am lost with your:
         Projection Test: The engine projects against a validated
         domain rule or schema graph ().  Verification: Graph

         projection deterministically tests structural specialisation

         (). If projection succeeds, the hypothesis is structurally

         grounded; if it fails, the engine pinpoints the exact
         ungrounded concept or invalid relation before any execution
         occurs.

I suspect that somehow the business function/ operation is generated/
extracted - provided that some magic can agree with something.  With
validation of the signature of the arguments being OK?
Pre/post-conditions are OK?

Please assume that I am mostly ignorant.
PS Suspect now that a schema is available for what you are after and
that you don't need an ontology after all.
Yes/no?
Better stop now.

dennis de champeaux

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/LOAP123MB835242344E5E539B4D22383BD4D92%40LOAP123MB8352.GBRP123.PROD.OUTLOOK.COM.

[Simon Polovina]

Hi Dennis,

Thanks for the follow-up. Your intuition is spot on.

Let me try to answer your questions directly:

1. Are you describing infrastructure for meaning representation (LOT)? Yes. In CGs, nested contexts (derived from Peirce’s existential graphs and cuts) serve as the formal syntax for LOT constructs. Context boxes allow us to cleanly scope propositions inside modal, temporal, or hypothetical frames (e.g., [Belief: ...], [Plan: ...]), rather than asserting unconstrained global truth values. These can be extended into Peirce’s visual logic (or "cuts").
• References: Section 6 ("Inference") and Section 7 ("A More Comprehensive Illustration") of my paper (An Introduction to Conceptual Graphs, ICCS 2007) demonstrate this context-nesting mechanism using negative contexts, nested cuts, and Peirce logic rules in a corporate funding request case study (Paper: https://shura.shu.ac.uk/1175/). For how CGs with Peirce’s cuts convert deterministically into computable binary contexts and formal lattices, see also Polovina & Andrews (2013), CGs to FCA Including Peirce's Cuts (Paper: https://shura.shu.ac.uk/7433/). For the deep formal and computational foundation of Peirce’s EG cuts and goal-driven theorem proving in CGs, see John Heaton’s doctoral thesis (Goal Driven Theorem Proving using Conceptual Graphs and Peirce Logic, Loughborough University, 1994, available via the Loughborough Repository: https://repository.lboro.ac.uk/articles/thesis/Goal_driven_theorem_proving_using_conceptual_graphs_and_Peirce_logic/9406046?file=17022827).
2. Graph Projection & Verification (Is it validating argument signatures / pre- and post-conditions?): Yes. In an enterprise context, graph projection operates as a structural generalisation test (). You can think of it as validating function signatures, type bounds, and pre/post-conditions simultaneously in graph form. If an LLM generates a candidate action, projection tests whether that action graph is a valid specialisation of the domain's schema graph. If a parameter, relation, or constraint violates the schema, the match fails deterministically.
• Reference: Section 4 ("Projection") and Section 3 ("The Type Hierarchy") of the ICCS 2007 paper detail the formal mapping operations, specialisation rules, and subtype hierarchy matching during projection.
3. Do you need a schema rather than an ontology after all? Yes (in the enterprise sense). Enterprise systems like SAP already possess rich type hierarchies and business object schemas (e.g., Customers, Orders, Approvals). CGs represent these operational schemas directly as graph structures. We don't need heavy open-world ontologies; we need a fast, sound mechanism to ensure LLM-generated actions conform to strict enterprise business rules.

Hope this clears up the "magic" 😊

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAJ%2B4qGf9_R6e0h%3Dz5fGYvosmnFaXQ%3DCf2JmYWKTdJchMtF_qLQ%40mail.gmail.com.

[Ian Glendinning]

Dennis
Surely Wittgenstein's whole point in the Tractatus was that such models are vacuous and don't capture what really matters?
Ian Glendinning

--

[X DdC]

Dear Simon Polovina si...@similelogics.ltd via googlegroups.com,

      Hope this clears up the "magic".  
Yes, I poked into the links you provided and i got a better grip on
what is going on.  

Hence the next question: what triggers what you have described?  

You mention an LLM (or SLM) source.  Thus I am guessing that a new
requirement arrives with specified functionality in textual format and
is processed into an LLM/ SLM.

Then I am guessing that a match is made of that input against what is
available in your CG repository of your business schemas (relational
databases or better?) using the magic you described and an executable
operation with the intended functionality is obtained.
Right?

Still not clear why you need the LLM/ SLM detour.  Divide & conquer
suggests processing one new functionality at the time.  Why does The
matching requires a neural-type input?  The input size does not
require an LLM (and NVDIA processing) i surmise. Then why the SLM?
You need an enhanced version of the initial text spec? 

dennis de champeaux

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAEWxg7a_y6OdTKEPGPC1EvyOZ7LAL_BDxrie0PNHZ2v6YzeM7g%40mail.gmail.com.

[X DdC]

Dear Ian Glendinning

      Surely Wittgenstein's whole point in the Tractatus was that such
      models are vacuous and don't capture what really matters?

Not in the Tractatus.  He changed his mind about the Tractatus focus
in his Philosophical Investigations.  Its index contains items like
ethics, feeling (experience, fear, grief, impression, pain, pity,
sensation ...), good, intention, ...

Wittgenstein focussed in the Tractatus on assertions and their
truth-false feature (and ignored what i suggested earlier:
pleasantness/ goodness/ ....).

Well ... that is, it looks like what AI has done also (although with
around 2000 some attention to emotions for cartoons - according to my
memory).  
AI has always been busy with the location of blocks, the frame problem,
truth-maintenance, ...
The Foundation ontologies have been focussed on conceptual structures
with semantic inheritance.
[[[[[ Ahum ... the AI community may have an overrepresentation of
Asperger personas. ]]]]]

[[[ Had prompt sessions with ChatGPT & Gemini regarding hot-potato
topics (Tragedy of the Commons, meta-morality with white lies).  If
interested I can send a link of the traces. ]]]

dennis de champeaux

[Ian Glendinning]

" Not in the Tractatus.  He changed his mind ... "

Precisely, not "IN" the Tractatus. He didn't change his mind though, the point of the Tractatus was always about the stuff he'd left out.
Witt knew and made clear (eg in his notes to the publisher of his Tractatus and more) that all this seemingly objective rational "propositions" stuff was the vacuous / tautological lesser half of his project. He was clearing it all away to leave room for the stuff that really mattered. Sadly his audience, Russell and the Vienna Circle just didn't get it (so he had come back to do his later work).

No?
Ian Glendinning

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAJ%2B4qGdXFQMjBR1Ng6%3DCuOVDm0LKtgC6EP032-Od5ST02_GTKg%40mail.gmail.com.

[Simon Polovina]

Hi Ian, Dennis, and List,

 

Thank you both for your critiques regarding the boundaries of formal logic representations in enterprise AI. My interpretation is as follows:

 

• Ian rightly highlights the risk of treating Conceptual Graphs (CGs) as static, truth-functional representations that degenerate into closed-world tautologies or formal vacuity. Purely flat, uncontextualized graph structures fail to capture dynamic, pragmatic dimensions such as intent, belief, temporal validity, and modal constraints.

 

• Dennis builds on this by observing that classical First-Order Logics often resemble early Wittgensteinian (Tractatus) systems restricted to binary truth values. As Dennis points out, enterprise architectures require a richer Language of Thought (LOT) capable of processing pragmatics, intentions, and multi-dimensional evaluations (e.g., preference, risk, compliance).

 

Considering all the discussions here, I’ve updated my working paper on Evaluating CGs, ISO/IEC 24707 Common Logic, and SAP architectures, to address these challenges:

1. Peircean Cuts & Contextual Scope: Building on Polovina & Andrews (2013), Peircean cuts operate as structural boundaries that define relational scope, polarity, and context containment. Rather than treating cuts as abstract modal spaces, we systematically transform nested cut boundaries into CG positive conditional dependencies (IF/THEN) and FCA binary formal contexts. This maps negative scopes directly into cohesive concept lattices, delineating situational constraints from ground truths through explicit structural containment.

 

2. Nested Contexts as the Pragmatic Bridge: Rather than asserting ungrounded global assertions, nested context boxes structurally enclose belief states, hypothetical "what-if" planning scenarios, or temporal validity windows. This ensures that pragmatic entourage metadata (source, intention, time, agent state) is structurally bound directly to the logical payload, satisfying the operational responsibilities required of multi-agent cognitive architectures.

 

3. The "LLM as Universal Translator" Fallacy vs. Deterministic Grounding: While practitioners often champion operational pragmatism—relying on LLMs to generate complex SPARQL or navigate RDF-star/SHACL harnesses—pushing -ary business events into binary quads introduces severe prompt fragility, bloated context windows, and non-deterministic join paths. Executing deterministic graph projection over native -ary structures under ISO/IEC 24707 Common Logic provides a verifiable, hallucination-free substrate for neuro-symbolic agent orchestration without relying on brittle middleware.

 

Best regards,

 

Simon

 

From: ontolo...@googlegroups.com <ontolo...@googlegroups.com> On Behalf Of Ian Glendinning
Sent: 19 August 2026 08:09
To: ontolo...@googlegroups.com
Subject: Re: [ontolog-forum] Wittgenstein for the enterprise?

 

" Not in the Tractatus.  He changed his mind ... "

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAEWxg7ZSifu24tHxGZcFMCe1_SHBXhJayBpa7rb7hOjmwgzpbw%40mail.gmail.com.

[John F Sowa]

This thread touches on many important issues that deserve far more discussion.  I have not been able to participate in the disussion for the past day or so.  But I want to start with the ending of the Tractatus:

"What can be said at all can be said clearly; and what we cannot talk about we must pass over in silence"

In the original German:  "Was sich überhaupt sagen lässt, lässt sich klar sagen; und wovon man nicht reden kann, darüber muss man schweigen."

But Wittgenstein's later philospjhy went very deep into all the issues that cannot be said clearly.  He emphasized that those issues are the most important issues of  science and human life -- far more important than the ones that can be said clearly.

I also agree that conceptual graphs (CGs) are a precise formal notation.  But that does not limit the methods for processing CGs.  In fact, some of the most important operations on CGs break out of the formallism into an open-ended search of what I called "Knowledge Soup" --  the totality of everything tht might pass through anyone's mind at any time in any way.    That knowledge soup is also the focus of what Wittgenstein was studying, analyzing, lectureig about, and writing about in his later philosophy.

One issue I emphasize is Peirce's cycle of pragmatism:  Abduction, Deduction, Testing, Induction, and repeat:

1. Abidction is the first stage of generaing a new hypothesis -- an educated guess.  And the emphasis is on the fact that a guess is just a guess.  There is no giaramtee that it is correct.

2. Deduction is formal reasoning.  It is absolutely precise in generating statements are implied by whatever guesses were generated by abduction.

3. Testing checks those implications to find internal errors (contradictions derived from the guesses), system errors (conflicts with other aspects of what the system has assumed to be true:, and external errors (conflicts with sources of knowledge outside of what was previously assumed).

4. Finally, the results that pass all the tests are integrated with previous knowledge in the totallity of what a person or a computer assumes to be true (what I called the Knowledge Soup}

Then the reasoning system repeats with more cycles of abduction, deduction, testing, and induction -- for as long as the system exists (human or mechanical).

There is much much more to be studied, evaluated, and developed.  I posted a considerable amount of it at https://jfsowa.com/ikl .  Most important:  LLMs are a method of abduction.  They are educated guesses, but they lack the deduction and testing to detet and correct errors in heir guesses.  That is why they hallucinate == even worse, they make subtle errors that go undetected and never corrected.

The IKRIS project was started by a different branch of  the DoD from the branch that produced the final Semantic Web report in 2005.  Some of the experts in DoD and many researchers in AI and related branches of computer science were extremely disappointed with the directions taken by the Semantic Web project in 2002 -- in particular, the limitations to decidability. 

Following is a brief summary: The Project for Interoperable Knowledge Representation for Intelligence Support (IKRIS) was sponsored by the US Department of Defense in 2005 and 2006. IKRIS brought together an impressive group of researchers with backgrounds in artificial intelligence, knowledge representation, logic, and ontology. Its primary products were the specification of the IKL language and the evaluation of IKL as a basis for interoperability among systems that use different versions of logic.

They brought a group of about 40 researchers in AI, computer science, and applications to develop a far more advanced and ambitious system that would meet and exceed the oriinal goals in Tim B-L's book and winning proposal in 2000. 

There is much more to say about all these issues.  But I recommend that people browse that web page and read anything they may find interesting.  I would emphasize Section 2 about the IKRIS project, which I believe should have been the foundation for all software developments since 2005.

Some advanced researchers have in fact built on that foundation.  Two of the members of that research group happened to be Arun Majumdar and me.  Our VivoMind system of 2010 was built on ideas that grew out of that project and our current Permion system combined that foundation with LLMs and other new developments.  There is much more to say about all these issues.''

But please browse through the many links in https://jfsowa.com/ikl .   What I wrote is a tiny fraction of the many reports and articles cited there.

John

 

---

From: "Simon Polovina" <si...@similelogics.ltd>
Sent: 8/19/26 12:28 PM

[Ian Glendinning]

John Sowa's words don't appear threaded here, but:

QUOTE

"What can be said at all can be said clearly; and what we cannot talk about we must pass over in silence"
In the original German:  "Was sich überhaupt sagen lässt, lässt sich klar sagen; und wovon man nicht reden kann, darüber muss man schweigen."

But Wittgenstein's later philosophy went very deep into all the issues that cannot be said clearly.  He emphasized that those issues are the most important issues of  science and human life -- far more important than the ones that can be said clearly.

UNQUOTE

The issues that cannot be said clearly are far more important than those that can.

Part of the reason my own agenda bailed-out of this forum more than a decade ago in order to pursue those things that cannot be formally documented clearly :-)
#MoreThanScience

Regards

Ian

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/LOAP123MB835249F2DDC84F9ECDDBEB40D4A52%40LOAP123MB8352.GBRP123.PROD.OUTLOOK.COM.

[Nadin, Mihai]

Wittgenstein hit a critical issue: the literal interpretation ('lässt sich klar sagen") vs. the speculation. Laws of physics are taken literally. But the phenomena described (take gravity, for example) are not illustrations of the literal meaning of the laws, but rather samples of a dynamics aggregated in the law. Newton's initial kinetic theory involved circulating aether particles which influenced gravitational attraction towards celestial bodies. He realized ("wovon man nicht reden kann, darüber muss man schweigen.") that something else was required. IN time, a curved-space model of gravity emerged...Yoou know the rest. 

Mihai Nadin

---

From: ontolo...@googlegroups.com <ontolo...@googlegroups.com> on behalf of Ian Glendinning <ian.gle...@gmail.com>
Sent: Sunday, August 23, 2026 07:21
To: ontolo...@googlegroups.com <ontolo...@googlegroups.com>

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAEWxg7aYS2A%2BMOGsoQzhwNxHFJ9_Cy3zi6aLfYUJfYXJHrNMYg%40mail.gmail.com.

[X DdC]

Dear Ian,

      Part of the reason my own agenda bailed-out of this forum more
      than a decade ago in order to pursue those things that cannot be
      formally documented clearly :-)

While i consider myself an AI-contributor i am sympathetic to your
stance.  Rabelais and fairy-tale style writing can be used to create
another perspective on the AI-discipline including ontologies.  Here
an example:
<<<
Civil

My AI-buddy (EI2718), Charloi and Elon got together (virtually) this
time inside a facility where novel manners are developed and tested.

S: Our previous meeting ended up with you two getting metaphorically
into each other's hair.
Charloi & Elon: No idea what you are referring to.
S: You two were creating mayhem and hence you two were not civil.
Elon & Charloi: What is civil?
S: You two were not polite to each other.
Elon & Charloi: What is polite?
S: Here is an example of not being polite: "You just utter nonsense."
Elon: Not clear what is at stake.  What happens if I say to Charloi:
"You just utter nonsense."
S: Likely nothing because Charloi said it without knowing what it
means.
Charloi: Indeed.  It is a phrase that I have encountered earlier and I
tried it out.
Elon: Looks like it can be said to those who doesn't know what it
means. But why saying it then in the first place?
S: That is a good rule, which you yourself have made useless.
Charloi & Elon: Indeed because how would we know that the recipient
doesn't know the meaning. OK, now what is the meaning?
S: It means that the party addressed is said something that is not
true.
Elon & Charloi: That seems to be nonsense.  Why say to
someone something that is not true?
S: Yes that is difficult to fathom.  There is actually a name for it.
Elon & Charloi: Cannot believe that.
S: Look it up: lying.
Elon: Thus Charloi had said that I was lying?
S: Not really because Charloi didn't know the meaning of what he said.
Also it is possible that somebody says something that is not true due
to plain ignorance.
Elon: Charloi, hope you can explain to me later what is going on
here.  Also we were not polite to each other.  What is that about?
S: Politeness has multiple meanings.  One of them may surprise
you.  Consider someone saying something that is wrong.  Instead of
telling that party that the utterance is wrong, reply with
"Interesting" or move gently onto another topic.
Elon & Charloi: We have to do more research here.  Can we now say that
this has been "Interesting"?
S: Not if you mean that what I said was nonsense.
Elon: Charloi, why did you let me say that?
Charloi: Elon, I thought you told me?
S: No more mayhem, please.
Time is up.
Bye now.
>>>

If you want more of this:
    https://www.ontooo.com/0All.txt

dennis de champeaux

BTW Wittgenstein was introduced because the Tractatus stance
corresponds with what is needed in enterprise systems - i believe.

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/PH0PR01MB63922C41257BC42E8E2ED14CDAA12%40PH0PR01MB6392.prod.exchangelabs.com.

[John F Sowa]

This thread includes a great many points that are worth discussing in detail, but that would require a huge amount of detail.. I'll just state some important points that Wittgenstein and C. S. Peirce made in their later writings, which present a profound perspective on everything they wrote from beginning to end.

First, both Wittgenstein and Peirce had a deep understanding of logic.  They knew exactly what it could or couldn't do.  In their early writings, they emphasized what logic could do:  derive precise the implications of precisely stated assumptions.

But they also realized that most of what we know or think we know or wish we knew is vague.  It cannot be the starting point of precise logical analysis.  Dealing with these issues was the major direction of their last decades of work.  And everything else must be interpreted in terms of their late wriings.

In his first book, the Tractatus, Wittgenstein emphasized what logic could do, and he dismissed everything "about which one must be silent".  But he later admitted that those later things were the most important aspects of life.  His later writings discussed and analyzed all those aspects in the greatest detail.

In his later writings, Peirce stated the issues clearly in his 4-step cycle of pragmatism:  abduction, deduction, evaluation, induction, and repeat.

1. Abduction is the process of deriving an educated guess or hypothesis.  But there is no guarantee that the guess is correct.  Some modern systems (LLMs for example) attempt to estimate the probability, but no estimate can be absolutely certain.

2. Deduction is a precise method for deriving the implications of any hypotheis generated by abduction.  There may be very many.  Some educated guesses are aleo necessary for deciding which ones to test by deduction.

3. Evaluation involves testing the deductions against observations, previously tested information, and requirements for further action.

4. Induction integrates the tested and verified results into the body of knowledge assumed to be accurate.  In various writings, I called this information the knowledge soup.  It contains lumps of thoroughly tested theories floating in a huge vat or soup of everything people know or think they know.

There is much more to be said about all these issues, but these  four principles are essential for analyzing and interpreting the issues that have been discusse in this thread.

John

[Ian Glendinning]

Interesting X-DdC

Are you saying " My AI-buddy (EI2718)" is a functioning AI or is this a literary thought experiment?

Ian

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAJ%2B4qGcO3a6PV2dBGRAn1q%2B8-TU_hYn0t%2Bo8Da53_Se3Zfjg2Q%40mail.gmail.com.

[Ian Glendinning]

Also X-DdC - not sure of rules / expectations for this forum these days - I/we should maybe start another thread?

Your attachment starts on "Thinking about Thinking" ... a subject close to my heart - HOW we think, understand and decide.

There is a tremendous drive to practical action in most "professional" groups - on the presumption that the **way** we think is a given.
It's incredibly hard, without being accused of navel-gazing or analysis-paralysis to call for space to do that kind of thinking.
Anatoly Levenchuk is strong on different levels of thinking in (say) project planning contexts - "Systems Thinking about Thinking" distinct from thinking about planning and implementation, or words to that effect.

Regards

Ian

[John F Sowa]

Ian,

Ypur note got buried when I was away from my email for a couple of days.   I just got back down to the August level.

In any case, I agree with your point:  "Precisely, not "IN" the Tractatus. He didn't change his mind though, the point of the Tractatus was always about the stuff he'd left out."  

But in the Tractatus, he said that stuff could not be expressed -- "muss man schweigen." 

After he came down from the mountain, he spent the rest of his life trying to express what could not be expressed.  He made significant comments that any successors must address.  But at the end, he was still not satisfied with what he left incomplete.

John

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAEWxg7ZSifu24tHxGZcFMCe1_SHBXhJayBpa7rb7hOjmwgzpbw%40mail.gmail.com.