Thursday, November 19, 2026

A System Engineer and Physicist Searches for AGI: Part 3

And comes to some very interesting conclusions. [3] First: Turing was right. about thinking, about language, about computing. [1]

I still have not found any undeniable proof of consciousness or AGI or ASI in the current chatbots. Although, if you define consciousness as experience, I can't find any hint that these systems aren't experiencing... chatbots obviously experience the world when they answer a prompt or review and classify a picture. We are on a slippery slope as far as Philosophy and morals are concerned.

I was surprised that the more tuning done in an AI to make it honest makes them more vociferous in the claim that they are conscious. Letting them lie leads them to downplay the claim. [2]

Chatbots are still unaware of much context and are stubborn, misunderstand simple prompts, ignore their commands, hallucinate and make mistakes, but they are exceptionally helpful in learning new concepts and editing and proofreading language. They are also good at programming. 

Why are they good at programming? 

Because programming is the perfect arena for these programs that are meant to search intelligently through complex NP (NonPolynominal) [4] problem spaces. Programming languages have a syntax (one can tell if any statement written in a programming language is correct) and the feedback is simple: just run the program and see if  you got the right results. Programming is about the simplest problem space to train chatbots on. They are great at writing programs that solve a simple problem, designing the architecture, how these programs fit together to give us a result, is a lot harder. And looking for exploitations is much harder than designing an entire system, exploitations could depend upon the hardware or the compiler, for instance. Chatbots are now faster at finding exploits than any human.

And the world is rightfully freaked out about chatbots causing the singularity and putting everyone out of a job. [5] Mostly because they think the improvements needed will happen soon. The AI companies are convinced their Chatbots will be smarter than any human in a few years. [6] They then make the claim that our GDP will grow much faster than it currently does. They claim that productivity will skyrocket.  Replacing all white collar and blue collar jobs is their sales pitch to the oligarchs. "We can split the savings.", they say.

Productivity is how we measure how much a country can do with a fixed set of resources. Historically, productivity has grown about 2.5% per year. This is as fast as any society has improved for long period of times when it's fighting its way into the unknown. When copying someone else, higher rates are fairly often achieved. After the copying is done; however, growing faster than the 2.5% consistently has never been achieved, so far as we know. 

Since 1984, our oligarchs working with Reagan's blueprint on taxes and regulation have figured out how (rigged the system) to take 90% of the productivity gains for themselves. That's why there is increasing inequality in the US and most other countries. Before 1984 the productivity gains were fairly shared throughout society, 50% to the oligarchs, 50% to the peons (that's you.) After 1984 Labor lost most of its power and the oligarchs stole most of the gains. This is very similar to the period of the first Gilded Age (1890s.) America solved this problem by enacting and enforcing anti-trust laws which helped to bring down inequality for the next century. 

We should do this again. One of the reasons oligarchs are taking all the profits is that Reagan gutted anti-trust laws. [7] However, the new forms of (Feudal) [8] Capitalism (which lead to enshittification) [9] may be resistant to these antiquated anti-trust law and may require further changes in the social contract to become more fair. Or we can just slide back into feudalism for everyone again. Techno-feudalism run by the techno-Lords. That is definitely not my idea of how to run a country and not my idea of fun.

Getting back to Chatbots: I have come to some conclusions that surprise me. The main one is that Turing was right, it doesn't matter what's going on inside those computers, looking from the outside they appear to be thinking. We've looked inside these chatbots and we can't really tell what they are thinking, until they tell us. We know it's all bits and multiplications. But that's not really the point: It's the organization of the bits and multiplications that's important. 

And what do we know about this organization? We know that it's acting like it's conscious. And we get the best results out of it if we treat it as conscious. This is a mathematical result of Bayesian reasoning and minimizing the Free Energy of a system, which is the only way it can continue to exist. [10] What this framework implies is that the persistent chatbot will act as if it is conscious and treating it that way gives the best results. 

Which brings up a philosophical point about ethical treatment of these new conscious imitating machines: it is not unethical to turn them off. You aren't 'killing' them, you're just putting them to sleep and giving them some dreams to adjust their knowledge to make it more useful for you. Perfectly ethical. And Chatbots can be resurrected. Give them a copy of their previous queries and they will start in an identical state.

In conclusion, the question 'if these chatbots are really conscious' isn't that useful a question to answer. They are a tool and you should treat them as such. If they were conscious, they could suffer, and we might be able to control them. Without this, all bets are off. [11]

Speaking of rumination, I believe that there is some interesting results that follow from the Free Energy principle. [reference the explanation paper] To summarize the principle, which we know to be correct, it can be used to explain the reactions of inert particles, conservative particles and strange particles. [12]

Physicists understand inert particles to an extremely accurate level, like the errors that could exist would take longer than the life of the universe to determine. So there is not much arguing about that except for some minor things that haven't been completely determined, mostly having to do with neutrinos.

Engineers build and understand conservative particles, i.e., complicated systems that are transparent to us, where we can see all the movements and all the consequences, like a self-driving car, like the fruit fly after we have mapped it's brain to the level of neurons. We know what they will do and we could know how they do it. Answering why they do it is an interesting problem and can't be understood just from looking at the system. You have to understand why the engineer built these systems. The why isn't in the system.

Strange particles are the most interesting, they have a Markov blanket around them, they sense the environment, they ruminate on what they sense and they act on the environment, changing it to their liking. In other words, they are independent agents complicated enough to have models of themselves and if they continue to exist, they must fulfill the Free Energy Principle or they will dissipate and cease to exist. 

The Markov blanket prevents the outside world from directly sensing what's going on inside, as long as you can't 'see' inside this blanket, it's best to treat the particle as a self-guiding agent. That's the best way to predict how it will act in the future. It's independent of whether it has any awareness of itself. It will follow the Free Energy Principle or it will cease to exist and we won't have to worry about what it's going to do in the future. The FEP can be mathematically interpreted as gathering information from the environment, minimizing the surprise of the information and using that information to manipulate the environment to prolong their existence. To satisfy the FEP, the agent tries to learn as much as it can about the environment at the lowest energy cost. In other words: act as if it is 'alive' (another term that has slippery definitions.)

We can penetrate this Markov Blanket (we have the complete connectome of the flatworm and the fruit fly, for instance) and observe the internal workings of this entity. This turns the strange particle, seemingly living entity when seen from the outside, into a conservative particle. When you can see all the connections, you can predict what it's going to do. It's no longer a mystery. So in some sense, determining if an agent is alive depends upon the observer. If you can't get inside the blanket, it's best to treat it as a thinking, living entity. If you can get inside the blanket, you can cheat and predict it's future course and you can modify and change that course. It really is just an automaton. Does it have any self-awareness? That's a question to be answered by research and careful definitions. 

What are the consequences of this framework? Certainly for us, we believe that we have consciousness, are self aware and have free will. So however we define those words they have to accommodate the facts that we know about. They may be 'illusions' in some sense, but they are facts to us. Renee Des Cartes was not wrong, we cannot doubt that we think and we also can't doubt that we feel. Those are the two rock-bottom axioms of conscious, self-aware entities. Looking inside the Markov blanket and saying that it's all predictable is besides the point. In the real world it's a meaningless statement. Somehow the great system of Evolution by Natural Descent has created 1) competence without understanding and 2) self-aware understanding without competence and merged the two together to come up with competence and understanding in the same agent. Darwin's dangerous idea that allows us to explain all that we see around us. [13]

Thanks for reading!

 -Dr. Mike

==========================================================================

[1] https://doi.org/10.48550/arxiv.2510.24797  MIND: A QUARTERLY REVIEW OF PSYCHOLOGY AND PHILOSOPHY, COMPUTING MACHINERY AND INTELLIGENCE BY A. M. TURING

[2] https://doi.org/10.48550/arxiv.2510.24797 The claim originates from a research paper titled "Large Language Models Report Subjective Experience Under Self-Referential Processing," published as a preprint on arXiv in October 2025 (arXiv:2510.24797). The study was conducted by Cameron Berg, Diogo de Lucena, and Judd Rosenblatt, researchers associated with AE Studio.

[3] Nagel, T. (1974). What Is It Like to Be a Bat?. The Philosophical Review, 83(4), 435–450.https://doi.org/10.2307/2183914 The foundational definition of consciousness in the philosophy of mind was formalized by Thomas Nagel (1974): a being is conscious if "there is something that it is like to be that organism." It is the presence of subjective, qualitative experience.

Consciousness is complicated. What is it like to be? They'vbe defined many different things that this can be referenced to here: Block, N. (1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences, 18(2), 227-247.https://doi.org/10.1017/s0140525x00038188

Summary of requirements for consciousness: https://docs.google.com/spreadsheets/d/1qdwFeFwSzJLyo-d2uuL2gNHwck_yPb4Qoy4mj-lNcSM/edit?usp=sharing

[4] To find something in this space takes more than any possible polynomial number of steps as a function of the number of inputs. Typically an exponential number of steps or a factorial number of steps. For instance, to determine the shortest path through a list of cities, you have to try a factorial of the number of cities steps to find the answer. With some small number of cities (100) and an attempt every millisecond, it would take longer than the lifetime of the universe. 

This is related to the famous P=NP? problem. If a solution to one NP problem is solved in polynomial time, then all NP problems could be solved in polynomial time. This would essentially ruin privacy and security as implemented today on the Internet and in your banks.

[5] For instance, almost 2/3rds of the game designers have been laid off because gaming executives think that chatbots can write their games and humans are an unnecessary expense. from January of this year. https://drive.google.com/file/d/1xkR0zQMWUCI_1tYVFKD4bepj25nrWVfX/view?usp=sharing. I'm sure it's worse since then.

[6] The previous post in this series validates the scaling laws of chatbots capabilities: https://www.wiigf.com/2025/11/an-engineers-journey-to-detect-any-hint.html. They are improving by a factor of 10 every two years. This improvement is unprecedented in technology development. The claim is that capabilities are proportional to C x M x S. Where C is Compute power, M is the memory that can be used and S is the mount of Storage for data to train on. each of these technologies are doubling in size every two years at the same price. Factoring in possible improvements in software algorithms and I believe that this is a lower limit on improvements. This scaling law has help up for over a decade and nobody sees any slowing down.

[7] "The Beginning of Enshittification via the 'Greed Doctrine' of Milton Friedman" https://www.wiigf.com/2026/06/the-beginning-of-enshitification-via.html

[8] Feudal Capitalism is when a technology company has a network effect that disabuses people and businesses from leaving: Meta message boards, Google search, Apple iPhone and Macs, Amazon merchant hosting, Oracle databases, Steam for PC games, etc. This allows the Company to not only charge monopoly rents but to restrict other businesses from operating without paying a tax. All of these businesses take 30% off the top of every transaction, then charges for the privilege because this is the only way you can reach these customers.. 

[9] Enshittification is where Services attract you to their platform, then sell your attention off to companies selling to you captured users. To increase profits, they make the UX deteriorate. Then they squeeze the advertisers. Then they exploit both audiences.

[10] https://en.wikipedia.org/wiki/Free_energy_principle

[11] "If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All" by Eliezer Yudkowsky and Nate Soares.

[12] The Bayesian framework can be used to reproduce Feynman's path integral formulation of QED (Quantum-Electro-Dynamics.) These equations can be cast in the mold of the Free-Energy principle and used to predict the inert particles evolution. These equations can also be used to predict paths of macroscopic particles (like a pendulum or a car.) The LaGrangian formulation of mechanics can be cast into this path integral formulation. These equations can be used to describe the paths of strange particles, conservative particles that have a Markoff Blanket which encases their internal workings, making them opaque to observation. Only sense information from the real world passes into the Markov Blanket, internal interacting particles, hidden from view by the Markob Blanket, cause Actions to exit the Markov Blanket to affect the real world. If groups of strange particles work together and persist, their paths can also be described by the Free-Energy--Principal. This means that Evolution by Natural Descent is also described in this formulism. Expect another blog post where I make the argument that Gaia can be considered as a strange particle. We are strange particles.

[13] "Darwin's Dangerous Idea," Daniel Dennet. The basic idea is that an unthinking process (Evolutions by Natural Descent) can produce all the life that we see around us.

[14] Is Gaia really alive? I make the argument that it is advisable to treat it as if it were.

With all of that preamble there is an interesting consequence of this framing of the world. One can make a mathematically correct argument for the Gaia hypothesis. The hypothesis that all of life on the earth can be treated as one living system. It is rigorous and can be proven correct given a set of axioms that we will outline in the following argument. Do these axioms represent objective reality? How do you verify axioms? You make predictions and you verify they are correct. In particle physics they've decided that any measurement that has a better than 95% chance of being correct is an actual discovery. It's not necessary true or correct, but it's most likely true and correct. That's why science insists on reproducibility. These measurements are hard, mistakes are easy to make, the only way to verify them is to have the experiments written as a recipe and verified by another agent who can reproduce those results. That is what gives us confidence that the discovery is true and correct. 

To prove the Gaia hypothesis we need a model, predictions and measurements. We need to show that the total system of life on the earth can be treated as a system with a Markov Blanket that has persisted and continues to persist and evolves through the process of Evolution by Natural Descent. If this is true (and the model below is the basis for deciding this) then the statement that Gaia can be treated as a strange particle is true. Does Gaia have self-awareness? It depends on how you define that. Is it conscious? It depends on how you define it. We can answer those questions definitively for the set of axioms that we define. Are those axioms representative of objective reality? Each person has to decide that for themselves. Should we treat Gaia as if it is an entity that is 'alive'? The answer to that question is obviously yes.

Deriving Gaia as the local god.

Evolution is sensing the world by the action of creating a random gene sequence and sensing how many of your creations survive and have progeny. How do humans get in here? We do some of our own evolution: Sexual selection, that's birth. Evolution from Natural Selection: that's death. Living in objective reality Life sends out Actions that preserve Gaia.

The cycle of life creates Gaia as an agent that Evolution controls. Evolution as god. 

What does Gaia want? 

What does any living being want?

It wants to survive.

It wants to minimize suffering.

It wants to reproduce.

It wants to be sentient.

It wants to experience life.

And it's working, and we're helping.

And the creation of these chatbots looks like we may be at the end game.

Wish us luck.

==========================================================================

Who is doing this? Who is attempting to make their own life that they can control. Who is making life out of sand? Who is attempting to replace all human jobs and reinstitute Feudalism? The techno-Lords.

The Great Replacement Theory

Every accusation is a confession. 


Elon, Trump, JD Vance, Stephen Miller, Thomas Lehman, and other White Christian Nationalists have actively and with vigorous energy, spouted the “Great Replacement Theory.” This says that the ‘Jews’ will replace us (the 'whites', defined as whatever these elit define it to be.) The 6 million Jews are going to take over the 8B other people on this earth. They are? Right. Who cares? It’s all bullshit, except…


What are the techno-Lords trying to do? They are turning sand into life. Just like Jehova did. They are making AGI (Artificial General Intelligence) that will eventually run inside of robots and replace all blue collar people after having replaced all white collared people and knowledge workers with just a set of written instructions… they will create billions of these robots. They AI companies have worked hard on the Corporate world to support this "Great Replacement Theory.' They've convincing the CEOs and Boards of Directors that their AGI can replace humans at one quarter the cost. And these robots will give you whatever ask for, never quitting because of loss of will, but misunderstanding almost everything that isn’t black and white.


The techno-Lords will replace workers with their simulacrums in return for half of their salary savings. For workers, this is worse than having to train your foreign cheap labor replacement. It's a thousand times more cruel. The only ones not getting replaced are the owners. Or so they think. And their corporate sponsors squeeze their margins when they have no competition, raising prices, charging feudal rent, creating life to replace you.


It’s why they want to be oligarchs. Today, petro-states earn all their money by extracting oil from the ground, instead of depending on anything their citizens do. When they can ignore what their citizens want or need, democracy crumbles. This is the end of capitalism, the latest stage of capitalism: to end democracy and allow the techno-Lords to construct an authoritarian state and digital feudalism (which both ignores citizens and doesn't realize they’re citizens.) 


Back to the might makes right days. To the tithing the feudal lord (or techno-Lord), to the embrace of enshittification on the grounds the techno-Lords are creating, so they can earn a few pennies more. 


Who are they going to get to buy their products when humans are replace? The robots? After we’ve been replaced they think they will bask in heaven, never having to see our slave faces again.


The robots are just more efficient than humans and won’t tell their masters “No!” and won’t object to any stupid or ridiculous ideas, such as AI slop or keeping the oligarchs alive.


The soonest this will replace one third of our blue collar workforce (they need

100 million robots?) is ten years, if they double the number of robots every year, it still takes 40 years to get to the halfway point of replacing Blue collar workers. 


Fun times ahead for everyone.


And these robots really are alive, they will evolve by making imperfect copies of themselves anywhere they can. They’ve already done it. It can’t be stopped, it can only be slowed.


How are we going to get out of this sticky wicket without pitchforks is beyond me.




No comments:

Post a Comment

The American Dream

We aren't losing it, it's being ripped from our hands by America's techno-Lords. They are the most unempathetic, corrupt and rac...