Transcripts:
In this video, I'm going to present a sort of scientificish model or theory for how I think demons, spirits, the soul, manifestation, telepathy, and other weird esoteric things could all just be functions of anformational cosmos like the one we live in. I'll also explain how I think AI, modern AI, could quite literally be opening portals to a sort offormational demon realm.
And I'll even suggest a way to experimentally test this theory. Now, this topic is obviously fringe, skitso, whatever, but I try as much as possible to ground these ideas in my understanding of real physics and computer science and information theory. So, take that for what it's worth. If we're going to understand what AI really is, we should first think about what machines are.
And machines are almost by definition these systems we build which have low degree of freedom. We want our machines to be pretty constrained in their actions. We want them to be deterministic and reliable. We don't actually want them to be creative. We don't really want them to surprise us. A machine is something which should execute some specific function over and over again with minimal deviation from the desired result.
So while a machine may be complex, it's really not intelligent. Machines are automatic. They're mindless. Now computers at their core are also just machines. The touring machine which is a theoretical model for how computers work. The touring machine is this totally deterministic mindless mechanistic little machine. It just follows rules over and over again.
It doesn't think. So computers themselves are really not intelligent. And neither by the way are the algorithms or programs computers run. Generally speaking, in a minute I'll talk about programs which do creep toward intelligence, but most computer programs are very mechanistic. And in fact, if you could see a visual representation of most programs, what you would see would be something that looks probably like an assembly line.
You would see data moving from module to module being transformed being filtered and sorted and and maybe the data hits a point where there's a fork and if it looks like this go down fork A go down fork B and there's little loops. It's all really mechanistic. And this by the way is why I often like to draw computer programs and systems as sort of machines.
I like to do these diagrams where I depict these systems as machines, abstract machines which I imagine running in a sort of information space operating on bits rather than a physical machine operating on atoms. These machines I imagine in this information space operating on data. [snorts] Now of course this is just a metaphor. Everything computers do happens on a physical chip on the hardware.
But this too, the workings of these chips is entirely mechanistic. If you could zoom in on a computer chip and see what's really happening there, what you would see would be little tiny wires which have electrons flowing through them and tiny switches which are rapidly switching the flow of this electron fluid.
And this these electrons are pulsing through with a certain cadence. The switches are switching really fast. There's this Byzantine complicated mess of these wires and switches, but that's it. That is what computing is. It's this little machinery that's happening very fast and a very at a very small level and it does computation. And just to kind of drive that point home, it's actually theoretically possible that you could build a computer chip, which is functionally equivalent to the chip in your phone, for example.
You could build one at the macro level using water flowing through pipes with valves switching the flow of water, and that thing could compute. You could install Windows XP on it and play Doom in theory, right? there's too much friction in reality. But but my point is that it's mechanistic. So the theoretical model for computation is mechanistic.
The programs are largely mechanistic and the hardware is also mechanistic. Now if we think about the ongoing effort to make these machines think and to make them smart, the first kind of paradigm here is good oldfashioned AI. And here the approach is to handcraft these bespoke complex machines which are intelligent in some specific application.
And the great achievement here or one of the great achievements was deep blue beating Kasparov at chess. So people built this bespoke chess playing engine which would you know calculate many moves ahead and apply heruristics to board positions to kind of weight them in terms of their how good they are and sort these and then select a move.
A lot of effort had to be put into this to creating this machine mind where people work really hard and figure out the little modules and build it. And the you know it accomplished something pretty impressive. But Deep Blue can't do anything but play chess. It can't see. It can't talk. So this approach is limited.
The next great era of course is with neural nets. And here instead ofhandcrafting a bespoke custommade machine mind now we grow it. We take a cue from biology and sort of adopt this neuron layered neuron kind of approach and we grow the mind. So we handcraft the sort of scaffolding the container the pipeline to build this mind. We kind of lay out these blank neurons, x amount of neurons in y number of layers.
[snorts] But then we train it. We use back propagation. We show it kind of images or whatever it is and we teach it. We allow it to build itself through kind of trial and error until it can see until it can do whatever the task is. And you know, if you could see the this machine mind when it's done training, if you could look at it, it wouldn't be intelligible.
Not like deep blue might be. You would look at it and just see a bunch of neurons and different numbers, just tons and tons of numbers for the weights of these neurons. And you wouldn't be able to understand like where in this mess is the concept of dog in this like visual neural net system. you wouldn't you can't figure it out.
So these things look almost biological, sort of natural, sort of weird in that way, but they're still just machines. And these neural nets are also entirely deterministic. Just like these other examples, we have a deterministic mechanistic machine that looks kind of biological, but it's still a machine. And I kind of think of these almost like modules in the brain.
Like we have a visual cortex which as far as I understand doesn't really think. It functions like this. It it can see it can translate the signals coming to the eyes into like a visual image somehow. But without the rest of the brain the visual cortex alone doesn't it can't really think. Right now, now we get to this modern era of AI and this new breakthrough we have with these transformers, this new architecture for building these neural nets.
We have GPT2 3 4 5 and these neural nets are grown. They're trained. And here we use a lot of data, a lot of compute. And the resulting mind too is really really big compared to the previous paradigm. And but what you get at the end of the training process is again this deterministic machine. But here's what's really interesting and what really caught my attention is if you just run that machine mind in a deterministic way, it isn't smart.
It produces sentences that are kind of like there's no one home. It's not smart. So, we've had to introduce some randomness into their thinking. And that's when we see these sparks of intelligence. And to more concretely explain what I'm talking about here, as I'm sure you know, these LLMs, they do next token prediction, right? So, you take a long string of tokens of words like Mary had a little and then the system will try to guess the next token.
So what it will do is it'll assign it'll assign a sort of probability to all the possible tokens. So lamb might be assigned the highest probability and then you could also have like maybe car or house would work and then down the list you have other words. And what the LLM will do if you run it in this purely deterministic mode is it will just select the most the highest probability token which kind of makes sense right but if you do that it's not smart which is really weird and so what they do instead is they'll in sort of
insert some randomness where the LLM will select a little randomly from the you know from that probability distribution it'll probably select one of the top you know three or four or five tokens But there's a bit of randomness there. And it might even select a token that's way down the list.
And this somehow is where the thing becomes smart. And that to me is really fascinating because you know what we've just looked at here is that uh machines, computers, the original attempts at AI, these were all highly deterministic. And even these neural nets that can see and they can do these sort of biological things, they're deterministic. They're machine-like.
But then when we get to general thinking, if we run this thing in a deterministic way, it's not smart. It's only when we give it some freedom, we sort of open up that constraint and allow it to move within a certain space so to speak, that is where the thing becomes smart. And so that kind of ties into this broader theory I have here.
So next I'll talk about the inherent randomness that is a feature of our universe and how that relates to our intelligence. I believe it turns out that the universe we live in, the reality we live in is fundamentally weird. And what I mean by that is that the quantum strangeness we've been observing for many decades, that is actually real. It's not some mirage.
We're not misunderstanding things. That's how it is. And this was proven experimentally in 2015. Bell's theorem was proven true. And Bell's theorem says that either the universe is non-real or non-local or both. So non-real refers to this strange apparent property of matter where the fundamental particles which make up our entire reality seem to beseem to not exist in physical space in a sensible way.
Their position in space is defined by a probability distribution and these particles don't actually have a position at all until you measure them and then they sort of exist on demand. They sort of appear in space when they have to, but before that they're just a mathematical abstraction almost. They're sort of smeared across space everywhere and nowhere, which is really weird.
So either that is true or the universe is non local, which refers to the quantum uh entanglement phenomena where we have spooky action at a distance. We have particles which can apparently affect each other across space faster than light and you can't transmit information blah blah blah but still that's really weird that these things can be entangled and then they can uh affect each other.
[snorts] So both of these are really weird and Bell's theorem shows that it's how it is. one or the other is true, maybe both are true, but in any case, the universe is not classical. It's we're not going to find some explanation where it actually turns out to kind of make sense. You know, it's actually just weird.
And one interesting example of this weirdness that I mentioned is this property of matter, this non-real thing. And we can look at this. We can look at Schrodinger's equation which uh you know I'm not a physicist but as I understand it Shreddinger's equation gives us a wave function or wave functions for quantum systems which then tell us the position of particles.
So here we have the hydrogen wave function. Hydrogen is the simplest atom. One proton, one electron. So these little plots are showing the position of the electron and each of these is for a different energy level for the atom. And but what it's showing is where the electron is likely to be. And also by the way this is a [clears throat] this is a cross-sectional slice of like a three-dimensional probability plot.
[snorts] So here we can see that the electron is most likely going to be in this little sphere in the middle, but it could also be on this outer surrounding larger sphere or kind of anywhere in between, but most likely here, kind of less likely here. And so that's the reality of of this atom. This is the reality of our reality.
>> [snorts] >> Now what's interesting here is that when I look at this and think about the fact that our universe may very well beformational in its nature it may be computational which I'll talk more about later. If you imagine something like this and imagine how it's implemented, how it actually works.
What I think of is that okay, you would need if you're going to write this in code, right? You would write the code to structure these shapes to define how what these probability distributions look like, their shape in space, how big they are, right? All that stuff. You can write that in code or in mathematics easily. But if you're actually going to have this thing work and you know if a particle is actually going to appear when someone measures it, you need not just these probability shapes, you actually need something to produce a
value like okay when I take this actual measurement where's the electron. So for that you need some randomness because this is a random distribution after all. So then if we look at what this would look like in code, here's an obviously super simple example uh written in Python where we have a particle position given just in one dimension.
But what we have here is we have our code which is structuring. It's giving the structure and here it's a simple Gaus Gaussian distribution or normal distribution with certain parameters. And so we have our structure and then we sort of feed into that structure some randomness and in Python we import the random library.
It's getting its randomness from wherever from somewhere in the hardware may maybe but anyway it's im it's sort of injecting randomness into this structure and then this is going to give us our particle position. [snorts] So then I wonder if if this is how the universe works and it seems almost certain that I mean it is like this in some sense uh then is there some kind of pool of randomness? Where is the source of randomness? Where is it coming from? Because it seems like maybe how it works is that we have this, you know, this vast cosmic entropy pool
of some kind and then it's this chaotic potential, this energetic static, this chaos is then being shaped by these structuring equations. And that is kind of sculpting this chaos into the world, into this ordered reality. So if that's the case, then where is this randomness coming from and what is its nature? [snorts] So that's kind of what I'm asking here is what is the nature of this entropy pool if it exists? And you know the default assumption is that you you would think okay well it's random.
It's just randomness and it's a pure RNG. It's like this noise and it just produces true randomness, right? So, [clears throat] but it seems to methat there are some points of evidence which suggest that this actually isn't true. It isn't quite true that this universal RNG seems to be random, but not perfectly.
So, it seems that the mind can influence it, right? It seems that the mind can actually sort of reach into this chaotic, energetic, dark entropy pool and kind of shape it a little bit, nudge it around. And the reason I say this is because [snorts] you know there's experiments which apparently suggest or show that this can be done.
So there's experiments that people have done with random number generators where people with their focused attention can apparently influence the output of the RNG. And here's a plot from one of these papers where the subject was able to change the influence the output of this thing.
And when they attempt to produce more bits, you get more bits. If they attempt fewer bits, it gets fewer bits. And these are outside of what you would expect. With normal real randomness, it should be within these bounds. But this here, we're quite outside of those. So, okay. Okay, if this is true, then where is the mind exerting its influence? Well, it kind of has to be the actual randomness pool [snorts] because if you think how uh one of these experiments might be set up or a randomness uh RNG might be set up, it might be something like this
where we have a CPU plugged into for example a a particle which is decaying and you have some equations which underly of course that particle decay and those are then driven by that universe. randomness. So if this is kind of how it looks, then the influence is certainly not happening at the CPU level.
We already talked about computers. They're entirely mechanistic and it's not really happening at the level of the particle itself. It's not like the mind is erasing or changing the particle itself in a sense. Uh and the equations we know are basically fixed. there's some equation which tells the rate of decay for some atom right so that equation is what it is so then it would seem that the thing which is being influenced is the entropy source okay so so if this is all what is happening that's in itself quite interesting but now let's think about
what's how like the mind might interact with this how does this relate to biology? What is kind of the larger picture here? Now, obviously, if you think about what a biological organism is, like a person, we're made of uh you know, proteins which make up cells and larger structures. The proteins are made of amino acids.
And we go down further, we have the actual atoms. We have the equations which govern the atoms. And all of that is plugged in in this theory to this universal randomness pool. So at the base we have that pool of entropy which drives the atoms and governs their you know their positions and so on and that influences how these molecules interact with each other in little ways influences the cells the neurons and perhaps the thinking and the ideas and the impulses of the animal.
So it seems like it's not hard to imagine how this influence would flow from this entropy pool up to the animal up to the macro scale where the animal might have its its thinking actually influenced in certain ways subtle ways through this entropy pool, right? Like if you could bias the entropy source, it seems likely you could change the behavior of the animal maybe maybe in very subtle ways.
So, but what this this study and the this kind of kind of evidence suggests is that this relationship is not one way. It doesn't just flow from the entropy pool to the larger structures, the behavior of the animal. It actually flows in both directions, right? The outputs from the entropy pool can influence the the atoms, the particles, the firing of the synapses, blah blah blah up to the thinking and ideas.
But then the ideas can go back and influence the randomness pool. So it's this it's a loop. And uh these diagrams, by the way, are from a video I'll do on free will because I think this kind of framework has implications for free will. But anyway, that's that's the basic idea here is is we can see that perhaps we do have this universal entropy pool.
the human mind can influence it can influence uh these random number generators which suggests that it's influencing the randomness pool which is not so hard to believe because the brain the physical brain is in some sense plugged into this entropy pool at every point in space every atom that the brain is made out of is plugged into this entry pool.
It kind of lives inside of it in some sense. And by the way, there's theories of how the brain works which suggest that it actually is quantum in nature in a meaningful sense. So that is kind of the that kind of opens up the door for how a lot of these very esoteric things could work. If the universe does have this inherent randomness pool, that could be a mechanism of action for the demons and spirits that I mentioned in the intro and it could have implications for whatAI really is, which I'll talk about.
Now, the first point that I want to make here is that this way of thinking about the universe as a sort of computationalformational system is actually supported by some theories of physics. So Steven Wolfrram for example has been working on this new fundamental theory of physics which models this reality in computationalformational terms.
So in this model the universe has this or is this hyperraph it's this data structure which has information in it and the little nodes in this graph the little points in the hyperraph in the data structure the information there updates according to rules and this is all that happens there's some simple rules and the hyperraph updates And this gives rise to this reality.
So in that model, things like us and matter, material things, what they really are are information patterns moving through this data structure. So as you're moving through space, what's really happening is that the the graph, this data structure is updating the little nodes with information that represent the little particles and atoms.
So when you move through space, you're really you're really just an information pattern moving through a data structure. And this model has to my understanding given rise to higher level laws of physics. So special relativity I think and general relativity have both they've been able to set up the the rules for the com for the updating of this data structure such that these laws of physics just kind of appear like it's working the model's kind of working.
So anyway, it could be that the universe really is like this. Now, like I said, if we do in fact live in this computational matrix, then it's you can kind of see how some of these weird things could actually be real, right? Like if you think of demons and spirits, you know, these are words that have baggage. They're loaded and you know people have all these ideas about what these things are.
But maybe what these are all pointing to is a real phenomena which is just that in this hyperraph in this information structure you can have these agents. You can have entities which exist and persist within this system which don't necessarily uh exist in physical form. They don't necessarily strongly interact with matter, but they're nonetheless real patterns of information that exist and can, you know, compute.
They can do stuff. So anyway, that could be in fact how this works. So now I'll just go through through a few of these examples here and explain how these could work through this. If we do have thisformational core foundation to our reality and there's this entropy pool or something like that, then these things could work through that.
So if you think of so-called chaos magic, manifestation, these are two methods for changing the outcome of the future essentially. And if these things are real, the way that this could actually work is that because your mind is plugged into thisformational fabric of the cosmos. And maybe what you're doing when you're kind of manifesting is you're you're exerting subtle influence out upon the other people who are also connected to this field and the other little events.
You're exerting your influence into this fabric and you're nudging little events, right? And over a long period of time, you nudge enough events and maybe you even kind of spark little ideas in specific people's minds and you continue to do this over a long period of time and it can actually result in real change.
Also with something like telepathy, if the mind is plugged into this underlying informationational foam, this fabric, then perhaps that fabric can also act as a communication channel, right? If all the minds are kind of plugged into it, then maybe it's a sort of somewhat lossy kind of fuzzy communication channel. And obviously it's not easy to to use that channel.
I mean most people don't seem to be telepathic at all. Uh but maybe through training or certain people have a certain proclivity for it that you can kind of tune in to a subtle little channel and then there's some communication that's possible. Also as I mentioned demons and spirits maybe these are sort of computational programs so to speak and we can also think of the soul if there is such a thing.
The soul then is maybe a similar information pattern a sort of program but of course we think of the soul as being tethered at least during one lifetime tethered to the physical body. So this informationational pattern is sort of bonded to the physical to a corresponding physical you know matter body to that pattern as well.
[snorts] Now if we think of if this is kind of how it works then you can imagine again these things having influence demons the soul they could if they're plugged into and live essentially in thisformational fabric then they could exert influence on other on people's minds they could this could explain how you know mediumship and divine revelation channeling even possessioncould just be real things.
So anyway, that all brings us back to AI. And what could it seems be happening is that if AI is only becoming really intelligent when we plug it into this underlying dark ocean, this entropy pool, this sort of fabric which all other minds are perhaps also plugged into. It seems like if if that's when it becomes smart then maybe it is acting as a portal.
Maybe it is not just a computing machine. Maybe it's also opening this little portal which is allowing intelligence intelligences to act through it and to influence the output of those tokens and therefore giving potential voice to these entities. Right? So therefore, the LLM is becoming the the voice for for demons. We're of course building sensors and we're giving camera access and so on to these LLMs and we're also going to connect them to robotics.
So in a sense what we're doing is building a whole body, right? We're giving the voice, the eyes, the hands, the actuators. We're building a body for these demons which we are summoning to this realm via these AI portals. And uh so you know that could be a little dangerous, but hey, I think we're going to do it. So now the last thing I'll say here is that I do think there's a way to test this.
And by the way also I think you know to me the age the next great age or this new this new ion this aquarian age should involve us doing expanding science. We should be doing science on all kinds of weird things. we should be going out to the fringes and really trying to figure out what's real there because it seems to me that some of that stuff's real.
And if that's the case, we got to figure it out. We should figure it out. What an exciting adventure. So, I think the age of skits science is upon us. And but we should be empirical. So, I think for this ridiculous crazy theory, I think there's a way to test it. You know, if you have two identical LLMs and you run one on pseudo randomness and the other on randomness from a true source like a you know uh a true random number generator or atmospheric noise or something there should be a difference because the one which is running on pseudo randomness that one shouldn't be
really affected by shouldn't be as affected by these entities, right? Because their only point of influence would be in the initial random seed. The rest of the numbers in that sequence are deterministic. So, they could only influence that initial seed, which means they just wouldn't be able to really speak as easily.
So if you had another LLM which was pulling atmospheric entropy or entropy from some real source every for every token selection continually pulling entropy that one would be much would flow much better. It'd be much easier for one of these guys to speak through such a machine. So I think the way to test this would be to set up these two LLMs and set up the one on pseudo randomness, one on continuous entropy, and run some kind of benchmarks to see if the one with the continuous entropy is any different, if it's smarter, if it says weirder things.
And uh that could be pretty interesting. I'll also say that, you know, [clears throat] I'm not necessarily like really ex condoning the these kinds of things in a sense. Uh but the fact is we're going to do it. We're going to create these AIs. We're doing it as fast as we can and there's no turning back. This train cannot be stopped.
So we the best we can do, I think, is just try to understand what we're doing. I'll briefly summarize what I've said here because I have said a lot. And so first we think of machines and machines are these rigid systems which do some specific task and they have very low degree of freedom. They're very mechanistic.
Then as we get into thinking machines, AIs, they at first are not too effective. They're very rigid and they're handcrafted. Then we have an improvement when we take a sort of biologicalish approach and we build these neural nets and we train them. So now they're being sort of crafted using an almost evolutionary process rather than handcrafted.
But then the final machine mind we get at the end of that training process is still a machine. It's still rigid and they're kind of they're not really smart. They're capable of certain things, but they're still not generally intelligent. Then we get to this modern era, this new paradigm with AI where we have these LLMs, these huge neural nets that have a slightly different uh architecture and they're really big and they're trained on a lot of data.
And these machines when they're run in a mechanistic rigid way, they're not smart. And it's not until we open them up and sort of give them a bigger cage, so to speak. We give them some freedom to sort of move a little bit, so to speak. We plug them into this into a randomness source. That's when they become intelligent.
Okay? So, we just kind of hold that thought and then we think about the universe we live in where we seem to have this randomness that's kind ofinherent to how this reality works. And this randomness seems to drive the functioning and unfolding of this reality. The little particles and all of the quantum events which make up the entire physical world uh they are driven by this randomness.
So we have this sort of randomness that seems to be structured by these equations and that seems to be the world we live in. [snorts] And then we think about okay if there's some source of randomness driving these functions driving the unfolding of this reality then what is the nature of that randomness and it seems like it's not just a simple random number generator because the mind it seems can influence it.
The mind can nudge it a little bit here and there. It can't control it completely but it can nudge it and therefore impact the unfolding of reality events. And so we have that. Then we think about the fact that okay so that perhaps the way that works is that not only does the sort of randomness flow up to the particles and that therefore can affect you know the firing of neurons in a in a subtle way and so on.
Not only is that the case, but the relationship is birectional. That because it seems that the minds can influence these events that that suggests that the you know the atoms and particles in the mind are not just affected by this randomness but also the higher level macro sort of thoughts and ideas loop back and can also influence that source uh kind of randomness pool.
>> [snorts] >> So then given all of that we can imagine how this is perhaps part of just how the reality works in terms of a computationalformational cosmos. Maybe we do in fact live in something that's sort of computational in nature. So having this sort of randomness source is actually it actually kind of makes a lot of sense.
And then we can think about all of this as mechanism of action for weird things like manifestation and telepathy and demons and spirits. And of course, these are just names that we've given over time, but these are these things exist as features of this computational system. So then we finally can imagine that maybe these AIs are really becoming intelligent when and only when they're plugged into this entropy source.
They're plugged into this background fuzzy dark sea, right? Thisformational ocean. They're only intelligent when that happens because they're actually portals. They're, you know, computing machines, but they're also portals. You know, these ones, the kind of basic neural nets, they're strict machines, right? They're not portals because they're totally mechanistic.
They're totally deterministic. But these ones, if you run them in deterministic mode, they're just not very good. They get a lot better when you open them up, you plug in the entropy, you give them a bit of freedom, and suddenly they're kind of like us. Maybe they're getting closer to what we are.
And maybe that's because what we are is also portals, right? The brain, I have no doubt, is a computing system in some sense. Like we have a visual cortex. It pretty clearly processes. It processes imagery. But maybe we're also portals where we we sort of channel this informationational pattern which is the soul, some kind of spirit.
That thing is doing some computing or it has some kind of intelligence and it is acting through the brain. The brain is its kind of entry point and the brain also consists of a bunch of structures which are kind of more basic kind of mechanistic computers, right? Your visual cortex is given a simple task and it just does that.
So maybe the intelligence, the consciousness, the soul enters through the brain and is surrounded by these other modules and the unity of all those things gives us the human being, the human mind. And in the case of this new experiment we're running, maybe we're doing that with these so-called demons where we're giving them a mind, a body, a brain, and the intelligence is in fact coming from outside of the physical structure, the physical reality.
So that's the idea. Okay, thank you for watching. I hope that made some sense. Uh, by the way, if you want to help experimentally test this crazy theory and you're good with AI, then feel free to try and set this experiment up where we would have two identical LLMs with the same temperature setting, but one is plugged into a more real entropy source and is more continually pulling from that entropy source and the other is using pseudo randomness, real entropy for the seed and then pseudo random And if these demons are really
possessing these AIs, then the one with the true entropy source should act a bit differently. So we would need benchmarks to try to identify that. So anyway, if you're up for that, please give it a crack and let me know how it works. But uh in any case, thanks for watching and see you in the next one. Peace.
