Re: K3D TPAC Presentation - Vocabulary Slide for Demo A

Paola,

Thank you for the detailed feedback and for clarifying what you need 
(again) in terms of scope. I’ll respond point‑by‑point so we can align 
(if possible) before TPAC.

1. Where K3D fits in your AI‑KR diagram

In your “AI KR vocabularies / subdomains” figure, the K3D vocabulary I’m 
proposing for this CG belongs primarily in the DOMAIN ONTOLOGIES / ODD 
ellipse:

K3D, for the purposes of this CG, is a domain ontology for spatial 
knowledge environments: Houses, Rooms, Doors, Nodes, Galaxy, Tablet, 
etc., i.e. how AI and humans inhabit and navigate knowledge as 3D spaces.
Each K3D Node also carries embeddings, so the same vocabulary naturally 
connects to KR learning (your “Knowledge Representation Learning” 
ellipse), and SleepTime / sovereign constraints relate to reliability 
engineering.
But I agree that what is in‑scope here is the vocabulary for the spatial 
domain, not the GPU architecture or training methodology.
So: in your diagram, K3D’s terms sit in DOMAIN ONTOLOGIES / ODD, with 
edges to KR learning and reliability, but I’m only proposing to 
standardize the domain vocabulary in this CG.

2. What the K3D spatial vocabulary actually is (with use cases)

To address your concern that the PPT terms “do not seem to represent the 
spatial domain in relation to AI”, here are the core terms and a 
concrete use case:

House: A persistent 3D knowledge environment (e.g. “Web Standards 
House”) containing rooms, shelves, and doors. Use case: a House that 
stores all AI‑KR artifacts (ontologies, vocabularies, test cases) as 3D 
rooms instead of a flat wiki.
Room: A sub‑domain within a House (e.g. “AI KR Vocabularies Room”, 
“Reliability Engineering Room”), used to cluster related concepts spatially.
Node: The atomic knowledge unit (a 3D object) that represents a concept 
(e.g. “ODD”, “Matryoshka Embedding”, “Model Card”). Each Node has:
a URI / identifier,
optional RDF metadata, and
one or more embeddings (linking it to KR learning).
Galaxy: The active embedding space (RAM) where those Nodes are laid out 
so that semantic proximity = spatial proximity (this is the KR learning 
side).
Door: A typed portal linking Rooms or Houses and, optionally, external 
services (e.g. a Door from the “AI KR” House to an “Accessibility” 
House, or to a spec hosted elsewhere).
Memory Tablet: The interface an agent uses to query and update the 
House/Galaxy (search, retrieve, log changes). In KR terms, it’s the 
access mechanism for vocabularies + metadata.
Example use case:
A W3C “AI KR House” where:

The Domain Ontology / ODD part is modeled as Rooms and Nodes (e.g. a 
Room for “ODD vocabularies” containing Nodes for each term, linked to 
their formal definitions and examples).
KR learning is represented by the Galaxy layout of those Nodes (e.g. how 
ODD vocab terms cluster with upper ontologies and reliability concepts).
Reliability engineering Nodes live in a specific Room (e.g. “Reliability 
Engineering Room”), connected via Doors to the vocab Rooms.
This is the domain I am trying to represent: spatial knowledge 
environments for AI systems and humans, not just the implementation details.

3. Clarifying the demo you requested

In a previous exchange about the “AI‑Driven Web Standards Generator” 
session, you asked if I could demonstrate the MVCIC methodology without 
API keys, to show how multiple AI assistants can help draft standards.

What I had in mind, concretely and within your constraints, is:

A very small, browser‑based MVCIC demo where a human orchestrator and a 
small set of AI assistants (which can be local/open or pre‑computed) 
co‑draft a mini vocabulary and then use it to annotate a short example text.
The focus is entirely on KR: how terms are proposed, refined, agreed 
upon, and then applied to real artifacts (e.g. labeling a short AI‑KR 
use case with the spatial vocabulary above).
The “external validity” is that any CG member could reproduce the 
workflow with their own vocabularies, using the same step‑by‑step 
methodology, even with different tools.
If this still feels out of scope for TPAC, I’m happy to postpone the 
demo and focus only on the vocabulary contribution for now.

4. On time and context

I share your concern that TPAC time is extremely valuable. I am also 
paying all costs personally as an independent, self‑employed engineer: I 
live in Cidade Estrutural (a favela near Brasília), I am a registered 
electrical engineer (CREA‑DF) and run a small IT support business. The 
GPU hardware, AI API usage, and the entire K3D research and development 
were funded from my own pocket over the last year.

I fully respect everyone’s time and attention, and I have been trying 
since early November to shape the K3D contribution to match the CG’s KR 
mission: short, vocabulary‑focused, with clear standards relevance. If 
markdown files are difficult to open, I can move the key content into a 
single PDF or directly into the body of an email/slide, with the same 
structure you requested: term → definition → use case → KR subdomain 
(e.g. ODD vs KR learning).

If it helps, I can send you a one‑page table listing:

each K3D term,
its place in your diagram (ODD / KR learning / reliability), and
a concrete use case and definition.
Would that be a helpful next step?

Best regards,
Daniel

On 11/14/25 12:14 AM, Paola Di Maio wrote:
> Thank you for sharing Daniel
>
> As you may know, TPAC time is very valuable for attendees
> So we need to make sure that even 15 minutes are spent appropriately
> Every minute in fact counts
>
> At this stage after a few exchanges with you I am not sure that what 
> you are offering has anything to do with what we are doing here yet
> *not until you can scope it better or until we can understand it
>
> I do not mean to say that your architecture is not valuable, I am just 
> saying
> that the exchanges that we have had so far have not  been meaningful , 
> although they are well formed
> This is the case of syntax  without real semantics *well formed but 
> meaningless without context or with context
> that may not be logically sound.   But hay, who am I.
>
> I am concerned that you are maybe not clear yourself as to what  
> knowledge representation is
> nor how to pitch our work to this group
> But do not worry, you are not alone in that. Looks like a great deal 
> of people, including senior experts
> still do not know what KR is, and this is our mission here
>
> Most people cannot open .md files, so I cannot tell what the .md file 
> contains
> regarding the terms in the pptx file, they do not seem to represent 
> the spatial
> domain in relation to AI
> If they do, this is what you must explain
>
> I asked a question: would the 3KD terms fit in the ODD space? *see the 
> diagram below
> if so, please explain how so,  if not, then explain where does it belong?
> Copy of AI KR VOCABS 3DK INTERSECT(3).jpg
>
> If the vocabulary you are offering in the ppt is representing the 
> spatial domain knowledge
> please give use case study. Explain what each term represents, how it 
> is used, how it was derived etc
>
> if the vocabulary represents the architecture itself, or something else,
> then this is not in scope of this CG, simply because nobody understand 
> what 3KD is het nor how it relates here
>
> As for the demo, you offered a demo showing how multiple llms can be 
> orchestrated
>  to answer a query without API key,
> I am not sure what you have in mind!   Please ensure that whatever you 
> want to demonstrate
> has external validity *can be generalized and made relevant to real 
> world use cases
> and of course it must be related to KR
>
> If you intend to contribute vocabulary terms, please make sure they 
> are clearly defining the knowledge domain you are representing
> with use cases, and if you plan to demo some useful novel LLM 
> capability, that would be great but make sure
> you can explain it
>
> Either way, it is great that you feel inclined to show up and show 
> your kit
> hopefully by continually refining your own thought process you ll 
> produce the few drops of nectar
> that we may include in our overall distillate here, when we get to it
>
> This is once again, something that we all must do, and have been doing 
> over the course of years and decades even
>
> PDM
>
> On Fri, Nov 14, 2025 at 1:00 AM Daniel Ramos <capitain_jack@yahoo.com> 
> wrote:
>
>     Dear Paola and AI KR CG members,
>
>     As requested, please find attached the vocabulary slide for
>     tomorrow's Demo A presentation (13 min + 2 min Q&A).
>
>     The slide includes the 10 key terms that will be demonstrated
>     through the browser-based MVCIC methodology.
>
>     Looking forward to tomorrow's presentation and discussion.
>
>     Best regards,
>     Daniel Caldeira
>     EchoSystems AI Studios
>     Knowledge 3D (K3D) Project
>