Mathematic Objects are not very Realistic

For all things philosophical.

Moderators: AMod, iMod

Post Reply
Veritas Aequitas
Posts: 15722
Joined: Wed Jul 11, 2012 4:41 am

Mathematic Objects are not very Realistic

Post by Veritas Aequitas »

Mathematic Objects are not very Realistic
Original Discussion here:
viewtopic.php?p=757124#p757124
AI wrote:Your [Godelian] belief in the Platonic realm as "the most real" is an intriguing stance, but it rests on a reification of abstractions, treating mathematical structures as if they exist independently of any framework or cognitive process.
Let’s examine this critically:

1. Reality as a Continuum of Patterns, Not Discrete Mathematical Objects
Imagine white noise—a chaotic background of unstructured information. The fundamental level of reality resembles this: a vast soup of particles, waves, and fluctuating fields, without intrinsic mathematical structure.

It is only through clustering patterns—emerging due to physical forces like gravity and electromagnetism—that order arises.
Humans, as cognitive beings, interpret these patterns as distinct objects, such as apples or stars, imposing mathematical structures upon them for survival and understanding.
But different perceptual systems (e.g., bats using sonar vs. humans using vision) will abstract different degrees of reality, showing that no abstraction is the reality itself.
Thus, the "most real" level of reality is the unstructured continuum of raw existence, not the neatly defined mathematical entities we construct to describe it.

2. Mathematics as a High-Level Abstraction, Not the Ultimate Reality
Mathematical objects are not primary building blocks of reality but rather conceptual tools used to describe observed patterns.

A mathematical formula does not cause a particle to behave a certain way; rather, we use mathematics to model the particle’s behavior.
While computations can influence physical states (as seen in simulations, AI, or even biological neural processing), they do so within a predefined system—they are not independent entities exerting causal power.
Your idea that "the Platonic world seeks to access the physical universe through computation" assumes that mathematical structures pre-exist the universe and somehow manifest through physical interactions.

But:

Computation itself requires a physical substrate—whether silicon chips or neural networks—meaning it does not exist apart from material processes.
Mathematics does not “reach out” into the universe—humans impose mathematical descriptions on top of the already-existing reality.

3. Mathematical Objects Are "Lower-Dimensional" Abstractions of Reality
You suggest that mathematical objects are the "most real," but in fact, they are highly polished abstractions of a much richer, more complex reality.

Consider a high-resolution image reduced to a low-resolution pixel version—some information is lost in the process of abstraction.
Likewise, when we describe a wavefunction or a geometric structure mathematically, we are filtering out the messy complexity of the real world to create a precise but limited idealized model.

Conclusion: The Inversion of Reality in Platonism
Your view inverts the relationship between abstractions and reality—treating our mental models as primary rather than as descriptions of a deeper, non-mathematical reality.

The physical universe exists independently of mathematics.
Mathematical structures exist only within a framework, which is ultimately a product of human-based cognitive interpretation.
If anything, raw reality (particles, energy fields) is the most real, while mathematical objects are merely a high-level, human-derived representation.
Thus, mathematics is not the "ultimate reality" but an epistemic tool for navigating it—much like a map is useful but is not the territory itself.
Veritas Aequitas
Posts: 15722
Joined: Wed Jul 11, 2012 4:41 am

Re: Mathematic Objects are not very Realistic

Post by Veritas Aequitas »

From:
viewtopic.php?p=757124#p757124
godelian wrote:
Veritas Aequitas wrote: Mon Feb 24, 2025 2:20 am Yes, there are no real mathematical proofs in the real physical universe.
Indeed, there are only computations visible in the physical universe, and not proofs.
A computation is a sequence of ephemeral changes in a particular physical state.
Each such computation is the isomorphic mirror of an actual proof in the abstract, Platonic universe.
As Plato pointed out in his Allegory of the Cave, we can see the ephemeral shadows of the abstract Platonic forms by observing associated computational changes in some physical state. These changes are their "shadows".
This is a natural interpretation of the Curry-Howard correspondence:
https://en.m.wikipedia.org/wiki/Curry%E ... espondence

A proof is a program and a program is a proof.
Hence, you must carry out a correct and complete act of computation in the physical universe in order to prove the existence of a particular associated proof in the abstract, Platonic universe.

Hence, acquiring access to the abstract, Platonic universe requires a computational ritual in the physical universe.
Vice versa, it is my belief that the abstract Platonic world can also seek to gain access to the physical universe by triggering computations in it, of which we may be able to see the changes in a particular physical state, i.e. its shadows.
VA wrote:Your views are too shallow.

One of the finest known perspective of all of reality is this:

https://create.routenote.com/blog/wp-co ... 80x720.jpg
Using the above 'white noise' analogy, the whole of reality is merely a soup of particles moving everywhere.

It is the Big Bang forces that generate some the particles to be more compact clusters that enable more distinct eddies thus patterns.
From these clustering, human beings emerged with consciousness and intellectual capacity.
Instead of cluster of particles, atoms or molecules, humans evolved to see more distinct patterns that facilitate their survival.
While humans see apples as discrete units, for the same bats and dolphin with the sonar would see a fuzzy thicker cluster patterns of dots.
Thus humans and bats are perceiving the apples in a lesser degree of reality of abstracted patterns because the more realistic picture is merely a cluster of molecules, atoms or particles within the reality soup.

Mathematic objects are less real in contrast to reality at the particle level because mathematical objects are highly abstracted [polished] away from reality.

Mathematical computations are merely a manipulation of highly abstracted object of low grade of reality.
To gain a greater sense of reality, one will need to verify whatever is represented by mathematical objects via the scientific framework and system which is the gold standard of credibility and objectivity.

My counter above is a bit rough, thus AI response as above
My counter above is a bit rough, thus AI more polish response is as above
Flannel Jesus
Posts: 4302
Joined: Mon Mar 28, 2022 7:09 pm

Re: Mathematic Objects are not very Realistic

Post by Flannel Jesus »

I call bullshit that that was AI's real raw reply to him. I think you had to force the ai, through a sequence of prompts, to disagree with him in that particular way. Posting AI replies out of context, without sharing the full conversation, is always BS. You use it to disohnestly act like your opinion has more support than it does.

Why don't you just say what you think in your own words? Instead of continuously re-prompting the AI until it agrees with you?
Skepdick
Posts: 16022
Joined: Fri Jun 14, 2019 11:16 am

Re: Mathematic Objects are not very Realistic

Post by Skepdick »

Veritas Aequitas wrote: Tue Feb 25, 2025 6:08 am Mathematic Objects are not very Realistic
So you went and spoke to a mathematical object (the large language model) to conclude that the very thing you used to help your reasoning "isn't realistic"?

What a fucking idiot.

Question: what sort of mathematical objects are large language models?

Claide AI: Large language models (LLMs) can be viewed as complex mathematical functions that map input text to output text through a series of mathematical transformations. More specifically, they can be considered:

* Tensor Spaces: At their core, LLMs are vast neural networks represented as high-dimensional tensor networks. Each layer of the neural network can be understood as a series of matrix transformations in enormous vector spaces, typically with millions or billions of dimensions.
* Probabilistic Mapping Functions: They are essentially sophisticated probability distribution functions that learn to predict the next most likely token (word or subword) given a sequence of previous tokens. This makes them function as complex conditional probability estimators.
* Non-Linear Parametric Models: Mathematically, they are non-linear parametric functions with an extremely large number of parameters (weights). These parameters are learned through techniques like gradient descent during training, allowing the model to capture intricate patterns in language.
* Continuous Function Approximators: From a functional analysis perspective, LLMs can be seen as continuous function approximators that map input sequences to output sequences through a series of learned, differentiable transformations.
* High-Dimensional Manifold Learners: In the realm of differential geometry, LLMs can be understood as learning a complex, non-linear manifold in an extremely high-dimensional space, where semantic relationships between words and concepts are represented as geometric relationships.

The mathematical complexity of these models means they're not simple, easily-describable mathematical objects, but rather intricate, multi-layered representations that blur the lines between algebraic structures, statistical models, and computational algorithms.
Question: How are these mathematical objects reified?

Large language models are reified through several key mathematical and computational mechanisms:

Numerical Representation
The abstract mathematical structure is concretely instantiated as arrays of floating-point numbers (tensors) stored in computer memory. These include:


Weight matrices representing connection strengths between neural network layers
Embedding vectors that map tokens to dense numerical representations
Activation values that capture the state of individual neurons during computation


Computational Graphs
The mathematical function is physically realized as a directed computational graph, where:


Nodes represent mathematical operations (matrix multiplications, non-linear activations)
Edges represent data flow and information transformation
The entire graph can be executed on specialized hardware like GPUs or TPUs


Parametric Instantiation
The abstract mathematical object becomes a specific instance through:


Parameter initialization (typically random sampling from specific distributions)
Training processes that adjust numerical values through gradient descent
Optimization algorithms that iteratively refine the numerical representation


Hardware Materialization
The mathematical object is physically embodied through:


Specialized silicon chips (GPUs, TPUs) with architectures optimized for tensor computations
Massive distributed computing infrastructure
Storage systems that can efficiently load and manipulate large tensor networks


Algorithmic Instantiation
The mathematical function is algorithmically realized through:


Transformer architecture implementations
Attention mechanisms that compute relevance weights
Backpropagation algorithms that enable learning
Inference algorithms that generate text by sampling from probability distributions

This reification process transforms an abstract mathematical construct into a concrete computational artifact capable of processing and generating human-like text.
The fascinating aspect is how these mathematical objects transcend pure abstraction to become generative systems that can engage with human language in surprisingly sophisticated ways.
Impenitent
Posts: 5774
Joined: Wed Feb 10, 2010 2:04 pm

Re: Mathematic Objects are not very Realistic

Post by Impenitent »

not very realistic said the square root of negative one...

-Imp
Flannel Jesus
Posts: 4302
Joined: Mon Mar 28, 2022 7:09 pm

Re: Mathematic Objects are not very Realistic

Post by Flannel Jesus »

Veritas Aequitas wrote: Tue Feb 25, 2025 6:08 am
Most LLM software available on the web right now allows you to generate Share links so that other people can see *everything* you said in that conversation, not just the bits you choose to share.

https://chatgpt.com/share/67bdb682-88dc ... 1e59396da9

If you want to demonstrate that you haven't just goaded the LLM to give the response you wanted it to give, please give us a link like that.
Flannel Jesus
Posts: 4302
Joined: Mon Mar 28, 2022 7:09 pm

Re: Mathematic Objects are not very Realistic

Post by Flannel Jesus »

Can LLMs be bullied by human operators to give the responses the human wants?
ChatGPT said:
Yes, to some extent, LLMs can be pressured, manipulated, or "bullied" into giving certain responses, depending on how they’re designed.
Veritas Aequitas
Posts: 15722
Joined: Wed Jul 11, 2012 4:41 am

Re: Mathematic Objects are not very Realistic

Post by Veritas Aequitas »

There is nothing that I am trying to hide.
It is just that I do not want to clutter the OP.

Here is my question to AI
with reference to what Godelian posted and my response to Godelian
viewtopic.php?p=757124#p757124
(I posted this in the OP)
which I asked ChatGpt to give a more polish response.
VA to ChatGpt
Below is a discussion with my interlocutor who believe mathematical [as with platonic] object are the most real. He believes there is no proof outside mathematics, there are only computation.

[Interlocutor]"It is my belief that the abstract Platonic world can also seek to gain access to the physical universe by triggering computations in it, of which we may be able to see the changes in a particular physical state, i.e. its shadows."

[me]One of the finest known perspective of all of reality is this:

https://create.routenote.com/blog/wp-co ... 80x720.jpg
Using the above 'white noise' analogy, the whole of reality is merely a soup of particles moving everywhere.

It is the Big Bang forces that generate some the particles to be more compact clusters that enable more distinct eddies thus patterns.
From these clustering, human beings emerged with consciousness and intellectual capacity.
Instead of cluster of particles, atoms or molecules, humans evolved to see more distinct patterns that facilitate their survival.
While humans see apples as discrete units, for the same bats and dolphin with the sonar would see a fuzzy thicker cluster patterns of dots.
Thus humans and bats are perceiving the apples in a lesser degree of reality of abstracted patterns because the more realistic picture is merely a cluster of molecules, atoms or particles within the reality soup.

Mathematic objects are less real in contrast to reality at the particle level because mathematical objects are highly abstracted [polished] away from reality.

Mathematical computations are merely a manipulation of highly abstracted object of low grade of reality."

I believe my counter above is rather crude. Please give a more presentable reply with reference to the above.
Veritas Aequitas
Posts: 15722
Joined: Wed Jul 11, 2012 4:41 am

Re: Mathematic Objects are not very Realistic

Post by Veritas Aequitas »

Skepdick wrote: Tue Feb 25, 2025 11:14 am
Veritas Aequitas wrote: Tue Feb 25, 2025 6:08 am Mathematic Objects are not very Realistic
So you went and spoke to a mathematical object (the large language model) to conclude that the very thing you used to help your reasoning "isn't realistic"?

What a fucking idiot.
You're the fucking idiot.

LLMs are more than mathematical objects.
LLMs are objects of human intelligences of which mathematics is merely a tool among many other tools and materials used to produce artificial intelligence which had already surpassed human intelligence in certain aspects.

Here is my discussion with ChatGpt:
VA to ChatGpt:

Here is a discussion I had with an interlocutor:

[Me] Mathematic Objects are not very Realistic [with reference to AI's conclusion]

[Interlocutor] So you went and spoke to a mathematical object (the large language model) to conclude that the very thing you used to help your reasoning "isn't realistic"?

[Me]LLMs are more than mathematical objects.
LLMs are objects of human intelligences of which mathematics is merely a tool among many other tools and materials used to produce artificial intelligence which had already surpassed human intelligence in certain aspects.

How could I give a more effective counter to my interlocutor?
ChatGPT said:
Your interlocutor’s argument is attempting to turn your use of an LLM (which is based on computation) against your claim that mathematical objects are not very realistic. They are implying a contradiction: since you rely on something mathematical (AI), you must accept the "reality" of mathematical objects.

A Stronger Counterargument:
Your response is generally on point, but it can be sharpened by clarifying two key distinctions:
Mathematical Formalism vs. Intelligence
Mathematical Tools vs. Physical Implementation

1. LLMs Are More Than Mathematical Objects
A Large Language Model (LLM) is not a purely mathematical object—it is a physical system operating within a computational framework developed by human intelligence.

The abstract mathematics behind AI (linear algebra, probability, information theory) is not the AI itself.
AI is grounded in physical reality—it runs on hardware, uses electricity, and interacts with real-world data.
Mathematics is a tool within AI, but intelligence emerges from processing data, learning patterns, and adaptation, not from mathematical structures themselves.
Thus, using an AI does not validate the independent existence of mathematical objects any more than using a hammer proves the fundamental existence of "hammerness" in a Platonic sense.

2. Mathematics Is a Tool, Not an Independent Reality
Your interlocutor assumes that because AI computation relies on mathematical principles, mathematical objects must be "real" in an independent sense. This is a category mistake:

Mathematical objects (like the number 3) exist within formal systems, but they do not exist in the same way as physical objects.
Just as chess strategies exist within the game of chess but do not exist independently in nature, mathematical concepts exist only within their frameworks.
AI is built using engineering, human decision-making, linguistic data, and empirical tuning—mathematics alone does not "create" intelligence; it only structures some of the processes.
If your interlocutor’s argument were correct, then every system that uses mathematics (weather models, stock market predictions, even human cognition) would be "proof" that mathematical objects exist independently. But that’s an unfounded leap—the use of a tool does not make the tool the ultimate reality.

Final Response (Refined Version of Yours)
"Your argument assumes that because AI uses mathematical structures, it proves the independent reality of mathematical objects. But AI is not ‘just’ a mathematical object—it is a complex system involving hardware, data, learning algorithms, and human-guided development. Mathematics is a tool within AI, much like logic is a tool in philosophy, but tools do not define the nature of reality itself. The existence of AI does not imply that mathematical objects are the fundamental fabric of reality—only that mathematics is useful within certain structured systems."

Conclusion
Your interlocutor’s argument confuses usage with ontology—just because mathematics is used does not mean mathematical objects exist independently. AI is not a Platonic proof of mathematical realism—it is a hybrid system that depends on physical computation, human intelligence, and empirical data, not just pure mathematics.
Last edited by Veritas Aequitas on Wed Feb 26, 2025 3:05 am, edited 2 times in total.
Veritas Aequitas
Posts: 15722
Joined: Wed Jul 11, 2012 4:41 am

Re: Mathematic Objects are not very Realistic

Post by Veritas Aequitas »

One thing I learned is ChatGpt [especially] is somewhat bias to the Questioner based on the saved history of past conversations and its standard of intelligence.

This is why I often refer to "MY" ChatGpt and 'Your' ChatGpt.
As such, if one do not agree with the above, it would be best to post the response from "MY" ChatGpt to 'Your" ChatGpt and ask for its comments.
Skepdick
Posts: 16022
Joined: Fri Jun 14, 2019 11:16 am

Re: Mathematic Objects are not very Realistic

Post by Skepdick »

Veritas Aequitas wrote: Wed Feb 26, 2025 2:07 am blah blah blah
You don't understand Mathematical realizability - the reification/implementation of a Mathematical object as a physical computation.
You are too stupid to tell when the AI is kissing your ass.

Every Mathematical object is exactly as real as its realization.

https://en.wikipedia.org/wiki/Realizability
Flannel Jesus
Posts: 4302
Joined: Mon Mar 28, 2022 7:09 pm

Re: Mathematic Objects are not very Realistic

Post by Flannel Jesus »

Veritas Aequitas wrote: Wed Feb 26, 2025 2:00 am There is nothing that I am trying to hide.
It is just that I do not want to clutter the OP.
Then post a link to your full conversation with chatgpt. One single tiny link won't clutter anything. That way we can see what sequence of things you had to say to chatgpt to get it to spit out what you posted in the OP.
Skepdick
Posts: 16022
Joined: Fri Jun 14, 2019 11:16 am

Re: Mathematic Objects are not very Realistic

Post by Skepdick »

Veritas Aequitas wrote: Wed Feb 26, 2025 2:12 am One thing I learned is ChatGpt [especially] is somewhat bias to the Questioner based on the saved history of past conversations and its standard of intelligence.

This is why I often refer to "MY" ChatGpt and 'Your' ChatGpt.
As such, if one do not agree with the above, it would be best to post the response from "MY" ChatGpt to 'Your" ChatGpt and ask for its comments.
Then you should probably use a stateless/sessionless model, no? You know. For the sake of context-free objectivity.

Look here!

https://chatgpt.com/share/67bef00f-5af8 ... 381ef201c4
Post Reply