Current AI is trapped in the Stochastic Paradigm: high-dimensional probabilities produce hallucinations, inconsistency, and fragile reasoning. To address these issues, this paper introduces the rule-based mechanism of human logical thinking, which follows a set of universal rules to perform the corresponding types of thinking. Behind this lies a mechanism through which neural activity follows the objective interrelationships to establish the corresponding conceptual relations within the neural network. Thus, the relationships of serial, parallel, convergence, divergence and symmetry are correspondingly translated into causal thinking, parallel thinking (analogy), convergent thinking (inductive reasoning and generalization), divergent thinking (deductive reasoning) and symmetrical thinking (opposite thinking). The investigation of logical thinking is part of the much broader ontological research project on A Theory of Everything. This research reveals that the rules of logical thinking are the fundamental rules underlying everything in the universe. These rules include serial, transition, transition point, continuation, discontinuation, parallel, similarity, commonality, difference, common mechanism, convergence, divergence, contraction, expansion, symmetry, asymmetry, limitation, limitlessness, order, disorder, hierarchy and interconnectedness. These concepts can be represented by an ontological framework: a geometric model – the Fundamental Interrelationships Model (IRM) and its associated ontological-mathematical formulation. This framework can be applied to logical thinking to the various issues of AI, such as Epistemic Instability (Hallucinations), Stochastic Variance (Inconsistency), Pattern Overgeneralization, Input Fragility (Prompt Sensitivity), Recursive Decay and so on…

You can download the full, new version of the article by clicking the link: https://philpapers.org/rec/HUATMP-2

I looked at on of the graphics cards and it was $5,000.

But I think there is a demand here. If someone can utilize the traditional processors for better parallel processing than Nvida they would save the pc industry.

Of course I am not on board with Ai. It makes changes to the computer while you program with it. But I can’t ignore it. I just don’t understand why traditional processors can’t be as efficient.

https://boardgamegeek.com/boardgame/374055/nukumiwen-quan-kai-tuo-ji-nukumi-onsen-kaitakuki

I took the screenshot (below) from a Japanese YT video about the game, and fed it into a Google Images search box, then used the inbuilt OCR facility to capture the text and pipe it into Google Translate, and here is what the AI told me

AI Overview

The image is a screenshot from the game No, I'm not a Human. 

• The card "意図せぬ来訪者" (Unintended Visitor) is a special event card in the game.

• The text "全プレイヤーは" (All players) indicates that the effect of this card applies to every player.

• The image of a person waving and the missing text suggest an action that all players must perform.

• The game is a horror adventure where players must distinguish between humans and monsters.

It also suggested that the game was available on Steam ! Most of which is complete and utter nonsense. The game is actually themed around building onsen (hot-spring spas) in a seaside town called Nukumi in southern Japan. The card text simply allows all players to draw an extra ‘helper’ card at the start of a round.

To be fair, the AI did significantly better a second time round when asked to do a ‘deeper dive’ but the fundamental problem seems to be that AI systems have a self-denying ordinance which forbids them from replying  “I don’t know” or “Insufficient data”.

Am after re-purposing a retired/orphan 'smart' phone with no service, for surveillance/wildlife/trail snapshot photo camera triggered by motion; help!

If you know of any application that can work, please advise.
No video, no service provider, no wifi. Just hang a retired 'smart' phone on a tree; later grab the phone hanging from the tree and look at pictures taken since last hung. Ignore supplying power to it.

This is not about a smart phone to communicate with a wifi wildlife camera at a trail to download images nor remote viewing. It is a walk-to-the-tree, look at the images taken, erase the unwanted. Hang it up again. Simple basics. Cannot find such ☹️

Is it too much to ask ? Decommissioned old phones filling drawers doing nothing when they have on-board the camera, the memory for saving, the brains to do it, the connectors/links to do it... Is there a web site dedicated for re-purposing these very featured useable 'junk' ?

Found many complex applications all aimed to the lazy press a wifi button to get video remotely. Plenty of surveillance camera applications too, all for video. No motion triggered photos ☹️

Can this image be altered (or create a new one) rotating the tracks 90 degrees so the tracks go on a slope from land to river entering underwater and mount the turbine on wheels, also rotated 90 degrees so the river flow can spin it ?

English is not my language and perhaps cannot express properly to GPT to create the image as desired. As far as I got :

Changes desired :

• 'rollercoaster' type tubular rails/tracks instead :

• turbine mounted on rollercoaster 'wheels' :

• sloping from ground to river.

• Thank you for your patience...

I’m currently researching a new model I call OpenSymbolic, based on symbolic communication units I’ve named conceptrons. The idea is to encode meaning using color, shape, and tone — forming structured “symbolic chains” similar to words or data packets.

I’d like to discuss its possible applications in communication, assistive technologies, and information systems.

Would it be appropriate to share a demo or a short paper for peer feedback here?

So what do you think, is it a new era for AI or will we forever be in control of AI?

I read an article some time ago about a Chinese AI based on spiking neural networks that requires ~90% less computational resources but it's an early stage model and I don't remember where the article was published. I also cannot comment on the merits of spiking neural networks themselves as my knowledge of comp sci is modest in general.

Here is the log: link removed by moderator per Rule 2.7

https://tomboy-pink.co.uk/projectgemini/Browser%20Tool%20Functionality%20Check/index.html

#AI #LLM #MachineLearning #EthicsInAI #Tech

Is there a specific language that would be good for this type of interest. Where can a newb start learning about building the tools to develop skills to develop stochastic-type models?

Thanks!

I’ve been following recent developments in AI and how they're being integrated into software development workflows. Tools like GitHub Copilot, ChatGPT, and AI-powered testing suites are changing the way developers write, review, and optimize code.

What fascinates me is how AI isn’t just assisting with code completion—it’s also influencing architecture decisions, bug detection, and even team collaboration. In your opinion, how far can we go with AI in the dev lifecycle?

Do you think we’ll ever reach a point where AI handles most of the actual coding, leaving humans to focus mainly on high-level logic and design?

Would love to hear your thoughts or any research you've come across.

Best

Suppose my bsp compiler chooses f₁ as the first splitting plane.  Then f₄ will be split into a triangle (f_4a) and a square (f_4b).   F₂, f₃, f_4b, f₇ will be "in".  F₆, f_5. f_4a will be "out"

This is a bad bsp tree for collision detection using offset surface.  The decomposition of the object into two convexes (a ramp and a block) will result in discontinuities that an avatar may get stuck on.  

If, however, I chose f_6a as the first splitting plane, I will not have that problem because the ramp-block connection will be part of the same convex.  How do I prevent this (and any generalization of this topology) from being compiled into a bad bsp tree?  

Beveling is not the problem, because you are only supposed to bevel if the angles are acute.  The ramp-block interface is obtuse.  

Thank you.

🔍 What Is It?

MathAI takes symbolic math inputs, evaluates them using sympy, and augments that logic with a few very powerful features:

🧱 Grounding Truths

MathAI has a core idea: it learns from grounding truths — equations that are known to be dimensionally and symbolically valid. These include classical physics equations (like kinetic energy, work with angle, Coulomb’s law, etc.) and mathematical identities from trigonometry.

Once an equation is solved and recognized as consistent (or verified manually), it’s promoted to a “grounding truth,” meaning it can be used to verify future equations by structure or result similarity.

🧪 Partial Truths & Learning

When MathAI evaluates an equation that doesn't match a known truth, but still produces a valid result, it stores it as a partial truth. These are reviewed and can be promoted to grounding truths — allowing the system to grow over time. This mimics a human learner testing what "seems right" until enough confidence is built.

⚖️ Dimensional Reasoning (Partial Unit Balance)

One of the script’s standout features is that it checks symbolic dimensional consistency. Using a lookup table of physical units, it compares the units involved in each term of an equation. If it detects that values of incompatible dimensions being added or equated, it raises a warning — something even many modern CAS tools don’t attempt symbolically.

🔁 Adaptive Ratio Logic

MathAI uses an adaptive ratio system. It doesn’t just take two variables and divide them blindly — instead, it checks their relative magnitude. If the first value is greater, it divides; if not, it multiplies. This feature is useful when modeling real-world systems where relationships change based on scale.

🤔 Why Does This Matter?

Traditional symbolic engines process math strictly. MathAI tries to reason about math — not just compute it. It asks:

• Is this equation built out of truths I already know?

• Do the units make sense?

• Is there a deeper pattern here worth remembering?

I believe symbolic tools should reflect and adapt — not just return numbers.

🧪 Call for Collaboration

If you’re a:

• Scientist working with symbolic systems

• Educator building teaching tools for physics or math

• Researcher modeling physical laws or quantum interactions

• Developer looking to integrate symbolic AI

…then I’d love to connect.

Whether you're working on difficult math modeling, scientific calculators, physics simulators, or something completely new — I think this tool could be incredibly useful.

Feel free to reach out, ask technical questions, or challenge the engine with hard problems. I'm always improving it and I plan to share it as widely as I can but I have no real means right now. 

Here's something to peak interest. Take into consideration that my code will only utilize a single core of a processor and works on my Ryzen 7 7800 with sub one second results. (Soon maybe GPUs will have support, but I need data sets with thousands of mathematical problems to run before its even relatively needed.)

The picture below has a scrollable 1503 decimal digits. Ran on my Ryzen.

If anyone has a math problem I can attempt to solve it! Just input it as comment.

I can also add almost every field of math! I can even invent a MathAI that works on specific but large areas of math like a version for medical sciences, or one for astronomy perfectly optimized by only working the required math.

It's even made to be able to function areas of math which are "bleeding edge", I plan to add more than a derivative, but a whole system that will allow creation of your own symbols as this is a closed application it will evolve to you!

One more post to get 5. So lets say you want to find a specific value for something seemingly infinite, consider what infinite even means, it means "indefinite" or "forever" but that is a rare function in reality, take a wide perspective and consider what positive and negative means. Positives indicate things like, kinetics, volume, mass, or anything physical at that point. SO how can something be infinite or indefinite? My perspective is that it is a simple factor of functions, since we know decay happens then something is removing values, so look into division and diffusion, it produces infinite wells of value depreciation, so then what can be inferred about finding infinite values? IF its positive, its not infinite, if its function is to decay or diffuse then yes it can be indefinite. This might be controversial because of how many things we think are infinite, but I think some values are a simply LONG.

Okay I gotta add this. Stars are BIG, imagine how much energy is conducted inside of a star! With all of that energy then light could be a reflection of the complexity in the conduction of plasma. Now calculate that.... Just kidding, as you can imagine a number which has as much complexity that is reflected in a star could be a value so loooong its unrealistically long. Point is, light can be narrowed to specific values based on the complexity of the waveform being produced by the particle shining because that particle has mass and is in combustion which is all align with positive function. At least that's what I believe.

• BSP offset as described by Melax in his paper for environment collisions.
• GJKEPA, or SAT for object on object collisions. 
• Use collision response formula for bouncing objects. 
• Use PGSLCP solver for sliding rolling objects. 

PGSLCP solver allows small inter-penetrations, which are corrected using Baumgarte stabilization.  This means I don't have to subdivide time intervals every time a collision occurs.  This works well with GJKEPA and separating axis, but BSP offset returns the exact time and location of the first collision.  BSP offset requires subdivision of the time interval, because if the object falls through the floor an infinitesimal amount the contact will not register on the next frame. My question; is BSP offset incompatible with PGSLCP solver? 

This has performed fine for several years now.

However just recently I can't access it.

Most of the references have disappeared in Windows File Explorer and I can't access the one that is left.

Any idea please ?