Re: reminding the scope for this AI KR CG

The "blue bubbles" do NOT cover the entire general knowledge representation
framework.

I will refrain from going into mathematical and linguistic details why not.

But if we consider this schema

{<natural language >,<symbols>,<diagrams>,<pictograms/glyphs>, <other
visuals}->{<KR>}->{<KR
frameworks>}->{<formalisms>,<standards>,<schemas>,<computational
frameworks>}->{<machine learning frameworks>}->{<machine learning
>}->{<AI>} and for sake of brevity and simplicity just limiting ourselves
to visual input knowledge representation we will run into problems if we do
not discuss domains of discourse.
Because this entire house of cards needs to be constructible, which implies
using set theory to create the atomic units we are talking about in the
first place.

Large language models tried to circumvent the linguistic problems by using
tokenization.
If we want to create KR that is more general than just the conceptual
frameworks that use tokenization we have to understand that the core
elements must be well defined.
That element is missing in the "blue bubbles ".

Milton Ponson
Rainbow Warriors Core Foundation
CIAMSD Institute-ICT4D Program
+2977459312
PO Box 1154, Oranjestad
Aruba, Dutch Caribbean

On Wed, Nov 19, 2025, 02:35 Daniel Ramos <capitain_jack@yahoo.com> wrote:

> Paola, all,
>
> Thank you for sharing the updated diagram and the clarification on where
> you now see the focus of the AI‑KR CG.
>
> As I understand your updated diagram, the “in scope” areas are:
>
> KR languages / formalisms and upper / foundational ontologies,
> vocabularies and concept maps around “Web AI standards”,
> knowledge representation learning, and
> reliability engineering for AI systems,
> with domain‑specific KR / “domain ontologies, ODD” treated as out of scope
> for this group.
>
> I’m happy to respect that boundary and keep detailed discussion of
> concrete domain Houses and application‑specific ontologies outside this
> list.
>
> At the same time, it may be useful for the group to know that the work
> I’ve been sharing under the name Knowledge3D (K3D) is not just about domain
> Houses. It also lives squarely in several of the blue “in‑scope” bubbles in
> your diagram:
>
> KR languages / formalisms and upper‑level structures
> K3D defines a structural vocabulary for multi‑modal KR (Houses, Rooms,
> Nodes, Doors, Galaxy, Tablet) that is independent of any one domain and
> substrate. The intent is to provide a shared “spatial KR language” into
> which different domains of discourse can be embedded. This is documented
> here:
>
> Spatial KR visual encoding (domains, concepts, relations, time):
>
> https://github.com/danielcamposramos/Knowledge3D/blob/main/docs/SPATIAL_KR_VISUAL_ENCODING.md
> Vocabulary specifications:
> https://github.com/danielcamposramos/Knowledge3D/tree/main/docs/vocabulary
> Knowledge representation learning
> K3D implements an explicit KR‑learning pipeline (Galaxy ↔ House
> consolidation, ternary RPN reasoning, procedural compression) rather than
> treating knowledge as an undifferentiated vector space. That sits very
> close to what Dave called out in his Plausible Knowledge Notation (PKN)
> note: reasoning over imperfect, contextual knowledge with explicit
> structure. Some relevant technical notes are:
>
> RPN ternary / three‑valued logic and domains of discourse:
>
> https://github.com/danielcamposramos/Knowledge3D/blob/main/docs/RPN_TERNARY_SETUN_CHAIN.md
> Overall architecture whitepaper:
>
> https://github.com/danielcamposramos/Knowledge3D/blob/main/K3D_Technical_White_Paper.md
> Reliability and adequacy
> The architecture is designed around bounded domains, adequacy rather than
> “scaling will fix it”, and explainable spatial traces (for both humans and
> machines). There is also an explicit energy / carbon angle, which Milton
> and others have highlighted as important for AI‑for‑Good:
>
> Carbon blueprint for a 10‑year horizon if K3D‑style architectures are
> adopted:
>
> https://github.com/danielcamposramos/Knowledge3D/blob/main/docs/CARBON_BLUEPRINT_10_YEAR_PROJECTION.md
> On the natural language side: my intent is not to replace formal KR work
> with a numeric trick, but to provide a substrate where natural‑language
> statements (including PKN‑style plausible knowledge and reified JSON‑LD/RDF
> patterns like the “Action adheresTo Rule” example Adam shared) can be
> represented, learned over, and inspected spatially. Milton’s emphasis on
> domains of discourse and linguistic plurality maps directly to how K3D
> treats each House.
>
> Given your updated scope, I propose the following as a way to avoid noise
> on the list:
>
> I will keep domain‑specific Houses and application ontologies (e.g., BIM,
> disaster response, etc.) off this list unless explicitly requested.
> When I do share K3D material here, I will focus strictly on the parts that
> overlap with your blue bubbles:
> KR vocabularies and structural patterns,
> KR‑learning and plausibility layers,
> and reliability / adequacy considerations that may be relevant to Web AI
> standards work.
> If that still doesn’t fit what you now want AI‑KR to do, I’m happy to
> treat K3D as an external case study that people can look at (or ignore) as
> they wish, and I’ll focus on other venues for the rest.
>
> In any case, I appreciate the clarification and the thoughtful
> contributions from Dave, Milton and others on this list. My aim is simply
> to contribute one concrete implementation example to the broader discussion
> on AI‑KR, within whatever boundaries the CG agrees are appropriate.
>
> Best regards,
> Daniel
>