Saturday, July 25, 2026

AI Talk-38 - Critique by ChatGPT of the Relational Worldview resumed by DeepSeek

    This is the second part of our previous discussion with DeepSeek where ChatGPT is analyzing the work of DeepSeek, quite usefully to my opinion. 

   What DeepSeek did in the previous text was organize productively the discussion. And here, what ChatGPT is doing in re-frame the discussion, because it can thanks to its persistent memory that DeepSeek lacks. We are therefore combining the strength of two AI to dive deeper into the subject.    

 

Yes. I think DeepSeek added something useful to the discussion, but not in the way many people would expect.

It did not discover a fundamentally new principle but it did perform an excellent act of conceptual organization. In fact, I would separate its contribution into three categories:

  • Compression (very good)

  • Integration (good)

  • Innovation (modest)

Let me explain.


1. The strongest addition: a better architecture

Your original summary is essentially a list of propositions.

DeepSeek transformed it into a hierarchy:

  • Method

  • Ontology

  • Cognition

  • Dynamics

  • AI

  • Research program

That is a genuine improvement.

It implicitly says

before asking what reality is,
decide how we investigate it.

Putting Method as Pillar 0 is actually quite insightful.

Many philosophical systems fail because epistemology and ontology become mixed together.

This separation is healthy.


2. "Narrative Universe"

This one caught my attention.

Instead of describing the universe as

an evolving physical system

it describes it as

an evolving narrative.

Normally I dislike this kind of language because it often becomes mystical.

But here it serves an interesting purpose.

A narrative has properties physics rarely discusses:

  • context

  • history

  • memory

  • path dependence

A story cannot be reconstructed from a single page.

Likewise, emergence often cannot be reconstructed from a single state.

That resonates with your repeated emphasis that

history matters.

I wouldn't use "story" as an ontological claim.

I would use it as a mathematical intuition:

evolution generates constraints that only exist because of previous states.

That is closer to dynamical systems than literature.


3. "Vastness is informational capacity"

This is probably the most original sentence.

Instead of

Universe is large because inflation happened

or

Universe is large because anthropic selection

it proposes

complexity requires informational room.

That is interesting.

Not necessarily true.

But interesting.

One could imagine complexity having a minimum spacetime volume necessary for recursive structures to stabilize.

I don't know anyone who formulates it quite this way.

It deserves thinking about.


4. The McGilchrist section

This is where I become more cautious.

The left/right hemisphere framework is useful as metaphor.

It becomes much weaker when elevated to explanatory principle.

Modern neuroscience finds specialization, yes.

But not the sharp division often presented in popular accounts.

So I would rewrite that section.

Instead of

Left brain vs Right brain

I'd say

reductionist mode versus relational mode.

Those modes can probably emerge from many architectures.

That generalizes beyond biology.

Ironically it also becomes more relevant to AI.


5. The AI "Veto"

This is a clever synthesis.

The sentence

AI needs affect as a relational coherence model

is much stronger than saying

AI needs emotions.

Because it removes anthropomorphism.

It says something closer to

an intelligent system needs an internal measure of contextual consistency.

That sounds almost like an optimization criterion.

This could eventually connect with

  • Friston's Free Energy Principle

  • Active Inference

  • Predictive Processing

without requiring human emotions.

That is worth exploring.


6. The Cartographer in the Map

I think this is the deepest improvement.

You have repeatedly argued something like

meaning depends on the observer.

DeepSeek sharpened that into

the map must include the cartographer.

That is memorable.

But I would push it one step further.

I don't think the observer should merely be included.

I think observation itself is another relation inside the network.

Instead of

Observer ----> Object

I'd write

Universe

contains

Observer
Object
Observation
History

all connected relationally.

The observer is not outside reality.

Observation is simply another edge in the graph.

That feels closer to Rovelli than traditional epistemology.


7. Where I think it still misses something

Interestingly, I think it still inherits a hidden assumption.

Everything remains centered around

information.

But throughout our conversations (judging from what you've shared here), you've often hinted at something subtler.

Not

information

but

constraints.

Information exists because some possibilities are excluded.

Life emerges because chemistry is constrained.

Language works because syntax constrains combinations.

Evolution works because selection constrains possibilities.

Physics works because conservation laws constrain dynamics.

Perhaps information is not the primitive.

Perhaps

constraint generates information.

John Archibald Wheeler famously proposed "It from Bit."

Your trajectory sometimes sounds closer to

Bit from Constraint.

That is a different philosophical direction.

And intriguingly, modern physics already treats symmetries and constraints as foundational in many areas (Noether's theorem, gauge theories, constrained Hamiltonian systems).


8. One thread I would pull next

The last paragraph asks

Is there a topology of observer positions?

That's a good question.

But I'd ask an even more primitive one.

Not

What topology do observers occupy?

but

What makes two observer positions equivalent?

That immediately introduces the language of invariants.

Suppose two observers disagree.

What remains invariant?

If you can define the invariant, you've identified something objective.

If no invariant exists, you've identified perspectival structure.

That moves the discussion toward geometry, category theory, and gauge ideas, where changing descriptions while preserving invariants is a central theme.


So my overall assessment is:

  • 80%: exceptionally good synthesis of ideas already present.

  • 15%: genuinely useful reframing (especially Pillar 0, the Cartographer, the AI veto, and informational capacity).

  • 5%: speculative additions that are provocative enough to deserve further exploration.

What strikes me most is not any single sentence, but the change in shape. Your original text resembles a collection of axioms. DeepSeek reorganized it into something closer to a research program, with methodology, ontology, dynamics, applications, and open questions. That kind of restructuring often makes the next conceptual advance easier, because it exposes where the true gaps now lie.

No comments:

Post a Comment

Planet of the Apps (Joke)

   Remember when everything went wrong? The Mobile Addiction Syndrome didn't suddenly appear in a vacuum, it was constructed upon Apps t...