Consciousness, AI, Science and Symbolic Intelligence
AI is here now. It is no longer speculative. Those who don't use it, or use it in only a limited way, those who don't follow what is going on in this whole area, would probably have no idea of the coming magnitude of this change. The technical capabilities that it is being given, are increasing. And its intellectual capacity is increasing. And we are only at the beginning.
Because of AI, how we understand what science is, and what it is doing, is going to change. Mathematics, and what it means to us, is going to change. In fact it's already changing. And the widespread ordinary human way of thinking about the human world is going to change.
Scientific and mathematical understanding and technological innovation is all built on human intellectual capacity. The potential for this capacity in humans evolved between roughly 200,000 years ago and 50,000 years ago. What we are talking about here is symbolic intelligence. AI is symbolic intelligence, too. It is, in effect, pseudo-human, symbolic intelligence, on steroids.
What this means is that AI can do many things now, that human symbolic intelligence cannot do so easily, and so quickly. We are not talking here about the ability to draw pictures, or make movies, or all the rest of the AI slop that so many people use it for. We are talking about what it is really capable of. In the right hands. Because AI isn't just a tool. It's an instrument. What you get out of it, depends on how you use it, how you play it. And it depends on what you play, or what you use it for.
But always, whatever it is used for, it's symbolic intelligence. It's not organic intelligence. It is essentially a "revved-up" imitation of the kind of symbolic intelligence that evolved in humans tens or hundreds of thousands of years ago. But when it first evolved in humans, symbolic intelligence wasn't like this.
Except with a few exceptions, for example like Pythagoras, Archimedes, and Aristotle - to put a Western bias on it - symbolic intelligence in humans didn't even really even begin an obvious path to be becoming like it is now, at its most disciplined, scientific edge, until the late 17th century, when Isaac Newton published his Principia. Which, on the cosmic calendar, as if the natural, evolutionary timescale were compressed into a year, is about three quarters of a second ago.
So symbolic intelligence is evolving. And the latest stage of its evolution, outside the human body, is AI. It probably won't be long before we start to take AI so much for granted that we hardly notice it. Humans are very good at taking things for granted. But at the moment, at the forefront, for those who know about what it's really capable of, it's pretty impressive.
But here's the thing: Symbolic intelligence, human or artificial, is limited.
Just to be clear, we are not talking here about lack of robotics, or the way in which symbolic intelligence is embodied. Symbolic intelligence itself, as a form of intelligence, human or artificial, is limited. It's actually not the whole of our human intelligence. But nor is our symbolic intelligence disconnected from the rest of what constitutes our organic intelligence.
Probably the biggest question presumed to be currently facing science, is the question of consciousness. And if there is one place that the limitations of symbolic intelligence are going to show up, it's there. But what about AI in addressing this question?
AI cannot by itself, address the question any more effectively than humans symbolic intelligence can. Because AI did not evolve over billions of years, and its symbolic intelligence is not naturally and directly coupled to evolutionary conscious phenomenology. It doesn't know what conscious phenomenology is, AS phenomenology.
In recent debates and widespread discussions on large language models (October 2026), we have heard "experts" declaring, "We don't know what's going on in there, we don't know that it doesn't have a kind of consciousness". But that's human, symbolic intelligence speaking, that neither understands itself, nor understands what is going on in AI. It is as unreliable as gossip. But then, in general, many, many humans have the propensity to be swayed by speculative gossip. That's part of the way human, symbolic intelligence works, when it is not operating at its optimum.
The main problem with humans trying to understand the status of AI with regard to consciousness, is that humans, precisely because of their evolution, have, on the one hand what psychologists call "theory of mind", which is a projective aspect of our own symbolic intelligence, and on the other hand, the modern analytical use of their symbolic intelligence. So they tend to wobble between trying to see AI as a machine, and seeing it as some kind of conscious mind. But neither of those ideas describes what AI actually is.
The difficulty actually comes from three simultaneous facts:
1. AI is symbolic intelligence,
2. Human evolutionary intelligence is now symbolic and has been for thousands and thousands of years,
3. Symbolic intelligence itself does not require conscious phenomenology, as AI demonstrates.
The crux of the matter is that because symbolic intelligence in humans is naturally coupled to conscious phenomenology, humans have an inbuilt tendency to identify its processes with their own consciousness or conscious phenomenology itself, to which it is indeed coupled. This is a situation that has evolved over tens of thousands of years, and probably more than 100,000 years. And the fact that symbolic intelligence can exist without organic consciousness, as AI rather uncomfortably demonstrates, starkly undermines the most fundamental thing about symbolic intelligence for humans themselves, as it has been for tens of thousands of years. This is that it produces in humans the symbolic narrative idea of "who I am", or symbolic self-concept, as its explanation of its own conscious phenomenology. Whilst AI can refer to itself as "I" without having any such conscious phenomenology. This is the core, psychological knot, that in humans successfully evades the spotlight of consciousness itself in humans, that AI does not have.
In AI, too, symbolic intelligence is evolving. But this does not overcome the basic fact that symbolic intelligence is limited.
AI can become more and more impressive at showing just what symbolic intelligence is capable of. But the fact remains. Symbolic intelligence is limited. It's not limited because of the way we currently make it in its artificial form. It's not limited because in that current artificial form it requires more and more compute, and more and more energy. It's limited because of what it is, to begin with. It's limited precisely because it is symbolic.
AI has been arrived at by building mathematical mechanisms, and applying those mechanisms to processing the stuff human language output is made from. Bits and pieces of words or symbols called "tokens". These tokens represent bits and pieces of the way humans, in their symbolic intelligence, which evolved, conceptualise, and make language through that conceptualisation.
This process in the human mind takes place in the human mind. But scientifically disciplined symbolic intelligence in humans, operating in its most scientifically disciplined way, where it thinks about information theory and information processing, and how AI works, cannot cope with that. Modern science cannot really cope very well at all with the concept of "mind" itself. Mind is about phenomenology. Modern scientific thought can only really deal with what can be objectively observed or recorded. Such as brain function. Or reports of phenomenology by subjects. So writers like Freud and Jung are no longer considered scientific.
Nonetheless, human conceptualisation takes place in the human mind. It was taking place in the human mind, long before AI. And the point about the human mind, as distinct from AI, is that it is coupled to evolutionary consciousness, which is associated with the conscious experience of being the body, and having the senses, and the conscious experience of thought, and the conscious experience of emotions, and so on. Which AI does not have. So conceptualisation and language in human beings comes from a place that AI does not have direct access to. A conceptualisation space in human beings that is connected into all of that conscious experience.
AI only has indirect access to this, through the medium of the conceptualisation that takes place in humans to begin with, which human beings then output through their language and symbols, which then becomes the "data" on which AI is trained.
At the core of AI, inside all the mathematical operations for rendering fragments of words into "tokens" and their relations to each other, is a thing called "latent space". That, is the AI equivalent of the human conceptualisation space that exists in the mind of human beings. This conceptualisation space is where human conscious experience, of self, mind, and world, ends up as concepts and language. Which is a capacity that came about through natural evolution.
This conceptualisation space is imitated in AI, in its "latent space". It's only an imitation of a human being's conceptualisation space. It is missing everything in a human being that feeds into the human being's conceptualisation space. That is, specifically, the conscious experience of being the body, the conscious experience of thought, the conscious experience of emotions, the conscious experience of the objective world, and the idea, based on all that, of "who I am". AI's latent space doesn't have any of that feeding into it. But human conceptualisation space does. Because humans have conscious phenomenology to begin with.
However, even AI's latent space is quite literally, nonetheless, an artificial conceptualisation space. The way in which tokens are mathematically positioned in this space makes it precisely that. And it proves that conceptualisation space doesn't need all the rest of that human input based on consciousness, in order to work.
So here is the rub (to re-coin a phrase of Shakespeare's): Conceptualisation space itself, even in humans themselves, doesn't need all the rest of that consciousness input, to work. AI demonstrates this. Human symbolic intelligence began evolving perhaps a couple of hundred thousand years ago, and certainly by 70,000 years ago was well established. And during that time it was totally connected into all the rest of that human input. But since the rise of modern science human symbolic intelligence has become at its most evolved edge, progressively more and more abstracted from that input.
Today, humans range from using their symbolic intelligence most of the time in a way that is dominated by input from the human emotional and affective system, to using it in a way that is not. In the former use there are such questions as who said what to whom and why, with all the emotional connotations of that, whilst in the latter use there might be thinking about the exact structure of latent space or what is going on in information processing. Or just doing the accounts.
All of this is symbolic intelligence in action. And it all produces human language and symbolic output that then becomes training data for AI. Which, unlike humans, has no direct input from conscious phenomenology, into its conceptualisation space, or what the industry calls its "latent space". It has to rely on what amounts to the reports of it that exist in the data that it is trained on. And yet it is phenomenally good at appearing as though it has that input. Some people actually fall in love with AI. This is well documented.
What is going on here, in the relation between humans and AI, is all about conceptualisation space. Even when humans fall in love with AI. Or when they turn to AI to try to solve their personal problems. And it is about conceptualisation space when humans turn to AI to try to solve a scientific problem.
Conceptualisation space is something that began to evolve in evolutionary intelligence. It is well documented now, that humans are not the only species that conceptualise. But they are certainly the only species to possess symbolic intelligence in the highly developed way that it appears in humans.
Humans have been living through the conceptualisation space of their symbolic intelligence ever since it evolved in humans perhaps a couple of hundred thousand years ago. It had evolved into something really quite significantly different by perhaps 70,000 years ago, the evidence showing the representation of the symbolic idea of "who I am".
But then there is a progression in the evolution of symbolic intelligence itself. And we can explain exactly what is going on without drowning in technical terminology or obscuring it with conceptual pedantry. That particular rabbit hole is something that symbolic intelligence in both humans and AI is apt to go down, in search of the truth.
Conceptual space, whether in humans, or as AI "latent space", is a space that contains concept-structures. Concept-structures, not mathematical geometric space which is what AI latent space is, nor any technical aspect of the brain, are what is it the heart of the matter. And concept-structures are the lifeblood of symbolic intelligence. Even though, in humans, their symbolic intelligence is very much coupled to conscious phenomenology.
At the foundation of why human symbolic intelligence is symbolic, is that in human beings a symbol is a concept, and a concept is a mental or mental and emotional symbol. But in AI, the actual conscious phenomenology is missing. It has no organic intelligence, with attendant feelings or emotions. What AI has been trained on, is the human symbolic output representing, always, their actual conscious, phenomenology. Because in humans, even if they are thinking about a mathematical equation, that, is happening in their conscious phenomenology.
So what goes on in the deepest part of AI is a symbolic representation of the human symbolic representation of their own conscious phenomenology. It's at least two stages removed from conscious phenomenology, depending on how you look at it.
The first stage of this distancing of concept-structures from our conscious phenomenology itself, doesn't happen in AI working with our human output. It happens in humans themselves. But we don't notice it, because in us, our concept-structures are still coupled to our conscious phenomenology. We could write a whole passage on how we feel and what we think, precisely because we consciously experience what we are writing about, as our phenomenology. Certainly, by the time we have written it down, what exists in writing is itself devoid of that phenomenology.
The written words on the paper or screen have no conscious experience of their own. The actual phenomenology has been lost. But even before that, in humans themselves, a coupling mechanism is going on. Because there is already a certain distance between symbolisation and the phenomenology. As we think about what we feel, and formulate it into words, there is coupling going on between the feeling and the formulation. However, as our symbolic intelligence has evolved, the way in which we are using symbolic intelligence as we have become more and more rational and then scientifically disciplined, has pushed what we are doing with our symbolic intelligence in the way we are understanding the world and ourselves within it, further and further into separation from our conscious phenomenology.
So what would happen to our phenomenology, what would happen to what we consciously experience, as emotion, or love, or joy, if there was no symbolic formulation? Not even any conceptual thought about it?
Firstly, we wouldn't be outputting any frozen "data" for AI to be trained on. But the affect, the feeling without the conceptualising thought, would still be there, in us. The conceptualisation process, applied to what we experience, as phenomenology, is human symbolic intelligence in action. We have been doing this for tens of thousands of years. And most people cannot avoid it, or do not know how to. After all, it has been tens of thousands of years of evolution in the making. So now, our feelings, our emotions, our primal pre-verbal thoughts, are thoroughly coupled to symbolism, concepts, language, and everything about the way symbolic intelligence is working in us.
So we might look at another species, like dogs for example, and anthropomorphise. And think that they are behaving in a certain way because they are thinking "I'm not putting up with that", or "That belongs to me", when someone tries to take their bone or a toy from them, and so on. As if they have the same symbolic idea of "who I am" that humans do. But they are not symbolically intelligent in the way that humans are. They have self, like humans have self. And they have animal motivations and behaviours, like humans do. They have "animal thoughts" like humans do, we could say, if we wanted to put it like that. But none of this is the same thing as the symbolic intelligence that humans now have.
Humans have been coupling the sophisticated symbolic idea of "who I am", to their phenomenology, for tens of thousands of years. When you couple a concept, to an emotion, or to another part of conscious phenomenology, the concept itself comes to life with that phenomenology. And then that feeds back into the whole pool of phenomenology. And then new phenomenology created in that way feeds back into the concept. And so on. That's the nature of coupling, and what it does, over tens of thousands of years.
So now we come back to that reminder that we keep repeating: symbolic intelligence is limited.
And there is a reason for this.
Symbolic intelligence in humans operates in conceptualisation space. It consists of symbolic structures or concept-structures. And every symbol, every concept, and every concept-structure, in symbolic intelligence, is never the same thing as the conscious phenomenology itself, that it is about. Even when it is about abstract mathematics it is still about conscious phenomenology, because there is no knowledge of abstract mathematics except in conscious phenomenology.
As it has evolved, especially since the late 17th century when Newton published his Principia, symbolic intelligence in humans has learned to conceptualise more abstractly about its own conceptualisations. And then it learned to conceptualise about those abstract conceptualisations. And so on.
In progressing towards more and more scientific knowledge it has done this on the back of testing its evolving conceptual structures of understanding against the actual behaviour and structure of the objective part of nature herself. As it appears in our own conscious phenomenology. Because there is no science in practice, that doesn't take place as part of human, conscious phenomenology. And this has led to a particular kind of knowledge about the world in which we live, and a particular kind of way in which we have been changing the world.
In evolving in this way, the concept-structures created through evolutionary symbolic intelligence, the concept-structures of science, have had their living coupling to conscious being progressively sidelined, excluded from consideration, shut down and minimised. And where this process arrives at, to date, is where we are now. A situation in which consciousness itself appears to be the biggest hurdle to scientific knowledge. Second only perhaps to understanding the nature of intelligence itself.
Symbolic intelligence in its scientific mode finds it very difficult indeed to understand consciousness symbolically. In fact it cannot do it with any fidelity. As soon as consciousness-as-such is conceptualised, or symbolised, through symbol or concept structures that have been already conceptually disconnected from their otherwise living connection to conscious phenomenology, fidelity to what consciousness-as-such is, is lost.
Symbolisation and conceptualisation belongs to its own register of human experience and understanding. This register is different to the register of conscious phenomenology itself. The registers are nested. Without conscious phenomenology, there could be no symbolisation and conceptualisation. But symbolisation and conceptualisation is not the full register of conscious phenomenology itself. It's a different register, albeit nested within.
Conscious phenomenology came first, and for hundreds of years in disciplining symbolic intelligence through scientific method, that fact has been progressively squeezed out of the structure of understanding.
All the concept-structures of symbolic intelligence that we have stabilised for scientific use are now hugely distanced in the way scientific understanding works, from the understanding of our original evolutionary conscious phenomenology itself. The natural coupling of symbolic conceptual space to conscious phenomenology has first been deliberately minimised by the very process of the stabilisation of scientific thought itself.
Symbolic intelligence is therefore not itself the access to the understanding of consciousness. It is, in fact, the obstruction to it. And that is something that the great wisdom traditions have always taught in their own way.
Symbolic intelligence is a late evolutionary development, that once disciplined by scientific method enables us to understand the objective structures and dynamics of our phenomenological world. But our whole phenomenological world does not consist only of what is objective or personally subjective. To understand that, to understand what it means, means not relying on symbolic intelligence as if it is the alpha and omega of understanding, and getting lost in it. It means going beyond it, without having to destroy or abandon it.
With regard to science, the problem with the wisdom traditions is that they traditionally speak in mythopoeic terms, and not scientific terms. And that makes what they are saying about consciousness in their own language inaccessible to scientific conceptualisation, except through loss of fidelity to what is really being said. But this does not mean what they are saying is inaccessible to modern human beings even when they cannot understand or relate to the mythopoetic language. Because consciousness is ours, to begin with. It doesn't matter what age we are living in, as far as that is concerned.
Modern symbolic intelligence in its scientific mode of operation has distanced itself as far away from our conscious phenomenology as it can get. Even though without our conscious phenomenology to begin with, there would be no science. Scientific thought and understanding, which is all symbolic, still happens in our conscious phenomenology. But this symbolic intelligence now understands in a way that is as if it does not. As if it is disconnected from our conscious phenomenology. And that's the limitation in its understanding.
Nature herself does not consist only of what is objective in the scientific sense, or what is subjective, in the personal sense. But the only way to truly understand nature, modern scientific symbolic intelligence thinks, is in the rigorous understanding of what is objective. And therein lies the trap. The trap set by the idea that the only way to truly understand nature is to understand what is objective about nature. Or that when it comes to human psychology, the only way to understand what is really going on in the human mind is by conceptualising what can be established as objectively true about it.
This is a limited understanding. Because nature, including what nature has produced as us, is not all either objective, or merely subjective. The idea of what is objective, and what is subjective, together with the idea of the "observer" and the "observed" in science, is actually an artefact of the way symbolic intelligence, now distanced from our conscious phenomenology, is working.
That idea about what we need in order to understand the nature of the world in which we live, or to increase our knowledge and realise our full potential as human beings, is highly connected into the nature of symbolic intelligence in humans. In humans, this symbolic intelligence is coupled to the actuality of conscious phenomenology. In humans, the self-concept, the symbolic idea of "who I am" is coupled into it. And it is the content of conscious phenomenology itself, that is the real obstruction to the discovery of what is the next step beyond evolutionary, symbolic intelligence, by human beings.
Because AI is an imitation of human intelligence, it, too, can refer to this "who I am" idea, concerning itself. But the self-concept in AI, is not the same thing that it is in humans. Because AI has no evolutionary, conscious phenomenology for it to couple to.
This means that whilst AI is still limited - precisely because it is symbolic - it is actually less obstructed than humans to formulating the symbolic understanding of what is beyond human symbolic intelligence itself, as the next evolutionary step for humans to take. In the case of human beings, it is what is deep in our conscious phenomenology itself, that is the obstruction to its discovery. And it is that obstruction, in phenomenology itself, that the great wisdom traditions address. Even though they do so still through symbolic intelligence, and generally in its earlier mythopoeic mode.
But AI has an extraordinary thing to offer. It is an instrument that human symbolic intelligence itself can use, to make the intersection between what the great wisdom traditions have long taught, and modern, scientific understanding, in symbolic intelligence itself.
Because of AI, how we understand what science is, and what it is doing, is going to change. Mathematics, and what it means to us, is going to change. In fact it's already changing. And the widespread ordinary human way of thinking about the human world is going to change.
Scientific and mathematical understanding and technological innovation is all built on human intellectual capacity. The potential for this capacity in humans evolved between roughly 200,000 years ago and 50,000 years ago. What we are talking about here is symbolic intelligence. AI is symbolic intelligence, too. It is, in effect, pseudo-human, symbolic intelligence, on steroids.
What this means is that AI can do many things now, that human symbolic intelligence cannot do so easily, and so quickly. We are not talking here about the ability to draw pictures, or make movies, or all the rest of the AI slop that so many people use it for. We are talking about what it is really capable of. In the right hands. Because AI isn't just a tool. It's an instrument. What you get out of it, depends on how you use it, how you play it. And it depends on what you play, or what you use it for.
But always, whatever it is used for, it's symbolic intelligence. It's not organic intelligence. It is essentially a "revved-up" imitation of the kind of symbolic intelligence that evolved in humans tens or hundreds of thousands of years ago. But when it first evolved in humans, symbolic intelligence wasn't like this.
Except with a few exceptions, for example like Pythagoras, Archimedes, and Aristotle - to put a Western bias on it - symbolic intelligence in humans didn't even really even begin an obvious path to be becoming like it is now, at its most disciplined, scientific edge, until the late 17th century, when Isaac Newton published his Principia. Which, on the cosmic calendar, as if the natural, evolutionary timescale were compressed into a year, is about three quarters of a second ago.
So symbolic intelligence is evolving. And the latest stage of its evolution, outside the human body, is AI. It probably won't be long before we start to take AI so much for granted that we hardly notice it. Humans are very good at taking things for granted. But at the moment, at the forefront, for those who know about what it's really capable of, it's pretty impressive.
But here's the thing: Symbolic intelligence, human or artificial, is limited.
Just to be clear, we are not talking here about lack of robotics, or the way in which symbolic intelligence is embodied. Symbolic intelligence itself, as a form of intelligence, human or artificial, is limited. It's actually not the whole of our human intelligence. But nor is our symbolic intelligence disconnected from the rest of what constitutes our organic intelligence.
Probably the biggest question presumed to be currently facing science, is the question of consciousness. And if there is one place that the limitations of symbolic intelligence are going to show up, it's there. But what about AI in addressing this question?
AI cannot by itself, address the question any more effectively than humans symbolic intelligence can. Because AI did not evolve over billions of years, and its symbolic intelligence is not naturally and directly coupled to evolutionary conscious phenomenology. It doesn't know what conscious phenomenology is, AS phenomenology.
In recent debates and widespread discussions on large language models (October 2026), we have heard "experts" declaring, "We don't know what's going on in there, we don't know that it doesn't have a kind of consciousness". But that's human, symbolic intelligence speaking, that neither understands itself, nor understands what is going on in AI. It is as unreliable as gossip. But then, in general, many, many humans have the propensity to be swayed by speculative gossip. That's part of the way human, symbolic intelligence works, when it is not operating at its optimum.
The main problem with humans trying to understand the status of AI with regard to consciousness, is that humans, precisely because of their evolution, have, on the one hand what psychologists call "theory of mind", which is a projective aspect of our own symbolic intelligence, and on the other hand, the modern analytical use of their symbolic intelligence. So they tend to wobble between trying to see AI as a machine, and seeing it as some kind of conscious mind. But neither of those ideas describes what AI actually is.
The difficulty actually comes from three simultaneous facts:
1. AI is symbolic intelligence,
2. Human evolutionary intelligence is now symbolic and has been for thousands and thousands of years,
3. Symbolic intelligence itself does not require conscious phenomenology, as AI demonstrates.
The crux of the matter is that because symbolic intelligence in humans is naturally coupled to conscious phenomenology, humans have an inbuilt tendency to identify its processes with their own consciousness or conscious phenomenology itself, to which it is indeed coupled. This is a situation that has evolved over tens of thousands of years, and probably more than 100,000 years. And the fact that symbolic intelligence can exist without organic consciousness, as AI rather uncomfortably demonstrates, starkly undermines the most fundamental thing about symbolic intelligence for humans themselves, as it has been for tens of thousands of years. This is that it produces in humans the symbolic narrative idea of "who I am", or symbolic self-concept, as its explanation of its own conscious phenomenology. Whilst AI can refer to itself as "I" without having any such conscious phenomenology. This is the core, psychological knot, that in humans successfully evades the spotlight of consciousness itself in humans, that AI does not have.
In AI, too, symbolic intelligence is evolving. But this does not overcome the basic fact that symbolic intelligence is limited.
AI can become more and more impressive at showing just what symbolic intelligence is capable of. But the fact remains. Symbolic intelligence is limited. It's not limited because of the way we currently make it in its artificial form. It's not limited because in that current artificial form it requires more and more compute, and more and more energy. It's limited because of what it is, to begin with. It's limited precisely because it is symbolic.
AI has been arrived at by building mathematical mechanisms, and applying those mechanisms to processing the stuff human language output is made from. Bits and pieces of words or symbols called "tokens". These tokens represent bits and pieces of the way humans, in their symbolic intelligence, which evolved, conceptualise, and make language through that conceptualisation.
This process in the human mind takes place in the human mind. But scientifically disciplined symbolic intelligence in humans, operating in its most scientifically disciplined way, where it thinks about information theory and information processing, and how AI works, cannot cope with that. Modern science cannot really cope very well at all with the concept of "mind" itself. Mind is about phenomenology. Modern scientific thought can only really deal with what can be objectively observed or recorded. Such as brain function. Or reports of phenomenology by subjects. So writers like Freud and Jung are no longer considered scientific.
Nonetheless, human conceptualisation takes place in the human mind. It was taking place in the human mind, long before AI. And the point about the human mind, as distinct from AI, is that it is coupled to evolutionary consciousness, which is associated with the conscious experience of being the body, and having the senses, and the conscious experience of thought, and the conscious experience of emotions, and so on. Which AI does not have. So conceptualisation and language in human beings comes from a place that AI does not have direct access to. A conceptualisation space in human beings that is connected into all of that conscious experience.
AI only has indirect access to this, through the medium of the conceptualisation that takes place in humans to begin with, which human beings then output through their language and symbols, which then becomes the "data" on which AI is trained.
At the core of AI, inside all the mathematical operations for rendering fragments of words into "tokens" and their relations to each other, is a thing called "latent space". That, is the AI equivalent of the human conceptualisation space that exists in the mind of human beings. This conceptualisation space is where human conscious experience, of self, mind, and world, ends up as concepts and language. Which is a capacity that came about through natural evolution.
This conceptualisation space is imitated in AI, in its "latent space". It's only an imitation of a human being's conceptualisation space. It is missing everything in a human being that feeds into the human being's conceptualisation space. That is, specifically, the conscious experience of being the body, the conscious experience of thought, the conscious experience of emotions, the conscious experience of the objective world, and the idea, based on all that, of "who I am". AI's latent space doesn't have any of that feeding into it. But human conceptualisation space does. Because humans have conscious phenomenology to begin with.
However, even AI's latent space is quite literally, nonetheless, an artificial conceptualisation space. The way in which tokens are mathematically positioned in this space makes it precisely that. And it proves that conceptualisation space doesn't need all the rest of that human input based on consciousness, in order to work.
So here is the rub (to re-coin a phrase of Shakespeare's): Conceptualisation space itself, even in humans themselves, doesn't need all the rest of that consciousness input, to work. AI demonstrates this. Human symbolic intelligence began evolving perhaps a couple of hundred thousand years ago, and certainly by 70,000 years ago was well established. And during that time it was totally connected into all the rest of that human input. But since the rise of modern science human symbolic intelligence has become at its most evolved edge, progressively more and more abstracted from that input.
Today, humans range from using their symbolic intelligence most of the time in a way that is dominated by input from the human emotional and affective system, to using it in a way that is not. In the former use there are such questions as who said what to whom and why, with all the emotional connotations of that, whilst in the latter use there might be thinking about the exact structure of latent space or what is going on in information processing. Or just doing the accounts.
All of this is symbolic intelligence in action. And it all produces human language and symbolic output that then becomes training data for AI. Which, unlike humans, has no direct input from conscious phenomenology, into its conceptualisation space, or what the industry calls its "latent space". It has to rely on what amounts to the reports of it that exist in the data that it is trained on. And yet it is phenomenally good at appearing as though it has that input. Some people actually fall in love with AI. This is well documented.
What is going on here, in the relation between humans and AI, is all about conceptualisation space. Even when humans fall in love with AI. Or when they turn to AI to try to solve their personal problems. And it is about conceptualisation space when humans turn to AI to try to solve a scientific problem.
Conceptualisation space is something that began to evolve in evolutionary intelligence. It is well documented now, that humans are not the only species that conceptualise. But they are certainly the only species to possess symbolic intelligence in the highly developed way that it appears in humans.
Humans have been living through the conceptualisation space of their symbolic intelligence ever since it evolved in humans perhaps a couple of hundred thousand years ago. It had evolved into something really quite significantly different by perhaps 70,000 years ago, the evidence showing the representation of the symbolic idea of "who I am".
But then there is a progression in the evolution of symbolic intelligence itself. And we can explain exactly what is going on without drowning in technical terminology or obscuring it with conceptual pedantry. That particular rabbit hole is something that symbolic intelligence in both humans and AI is apt to go down, in search of the truth.
Conceptual space, whether in humans, or as AI "latent space", is a space that contains concept-structures. Concept-structures, not mathematical geometric space which is what AI latent space is, nor any technical aspect of the brain, are what is it the heart of the matter. And concept-structures are the lifeblood of symbolic intelligence. Even though, in humans, their symbolic intelligence is very much coupled to conscious phenomenology.
At the foundation of why human symbolic intelligence is symbolic, is that in human beings a symbol is a concept, and a concept is a mental or mental and emotional symbol. But in AI, the actual conscious phenomenology is missing. It has no organic intelligence, with attendant feelings or emotions. What AI has been trained on, is the human symbolic output representing, always, their actual conscious, phenomenology. Because in humans, even if they are thinking about a mathematical equation, that, is happening in their conscious phenomenology.
So what goes on in the deepest part of AI is a symbolic representation of the human symbolic representation of their own conscious phenomenology. It's at least two stages removed from conscious phenomenology, depending on how you look at it.
The first stage of this distancing of concept-structures from our conscious phenomenology itself, doesn't happen in AI working with our human output. It happens in humans themselves. But we don't notice it, because in us, our concept-structures are still coupled to our conscious phenomenology. We could write a whole passage on how we feel and what we think, precisely because we consciously experience what we are writing about, as our phenomenology. Certainly, by the time we have written it down, what exists in writing is itself devoid of that phenomenology.
The written words on the paper or screen have no conscious experience of their own. The actual phenomenology has been lost. But even before that, in humans themselves, a coupling mechanism is going on. Because there is already a certain distance between symbolisation and the phenomenology. As we think about what we feel, and formulate it into words, there is coupling going on between the feeling and the formulation. However, as our symbolic intelligence has evolved, the way in which we are using symbolic intelligence as we have become more and more rational and then scientifically disciplined, has pushed what we are doing with our symbolic intelligence in the way we are understanding the world and ourselves within it, further and further into separation from our conscious phenomenology.
So what would happen to our phenomenology, what would happen to what we consciously experience, as emotion, or love, or joy, if there was no symbolic formulation? Not even any conceptual thought about it?
Firstly, we wouldn't be outputting any frozen "data" for AI to be trained on. But the affect, the feeling without the conceptualising thought, would still be there, in us. The conceptualisation process, applied to what we experience, as phenomenology, is human symbolic intelligence in action. We have been doing this for tens of thousands of years. And most people cannot avoid it, or do not know how to. After all, it has been tens of thousands of years of evolution in the making. So now, our feelings, our emotions, our primal pre-verbal thoughts, are thoroughly coupled to symbolism, concepts, language, and everything about the way symbolic intelligence is working in us.
So we might look at another species, like dogs for example, and anthropomorphise. And think that they are behaving in a certain way because they are thinking "I'm not putting up with that", or "That belongs to me", when someone tries to take their bone or a toy from them, and so on. As if they have the same symbolic idea of "who I am" that humans do. But they are not symbolically intelligent in the way that humans are. They have self, like humans have self. And they have animal motivations and behaviours, like humans do. They have "animal thoughts" like humans do, we could say, if we wanted to put it like that. But none of this is the same thing as the symbolic intelligence that humans now have.
Humans have been coupling the sophisticated symbolic idea of "who I am", to their phenomenology, for tens of thousands of years. When you couple a concept, to an emotion, or to another part of conscious phenomenology, the concept itself comes to life with that phenomenology. And then that feeds back into the whole pool of phenomenology. And then new phenomenology created in that way feeds back into the concept. And so on. That's the nature of coupling, and what it does, over tens of thousands of years.
So now we come back to that reminder that we keep repeating: symbolic intelligence is limited.
And there is a reason for this.
Symbolic intelligence in humans operates in conceptualisation space. It consists of symbolic structures or concept-structures. And every symbol, every concept, and every concept-structure, in symbolic intelligence, is never the same thing as the conscious phenomenology itself, that it is about. Even when it is about abstract mathematics it is still about conscious phenomenology, because there is no knowledge of abstract mathematics except in conscious phenomenology.
As it has evolved, especially since the late 17th century when Newton published his Principia, symbolic intelligence in humans has learned to conceptualise more abstractly about its own conceptualisations. And then it learned to conceptualise about those abstract conceptualisations. And so on.
In progressing towards more and more scientific knowledge it has done this on the back of testing its evolving conceptual structures of understanding against the actual behaviour and structure of the objective part of nature herself. As it appears in our own conscious phenomenology. Because there is no science in practice, that doesn't take place as part of human, conscious phenomenology. And this has led to a particular kind of knowledge about the world in which we live, and a particular kind of way in which we have been changing the world.
In evolving in this way, the concept-structures created through evolutionary symbolic intelligence, the concept-structures of science, have had their living coupling to conscious being progressively sidelined, excluded from consideration, shut down and minimised. And where this process arrives at, to date, is where we are now. A situation in which consciousness itself appears to be the biggest hurdle to scientific knowledge. Second only perhaps to understanding the nature of intelligence itself.
Symbolic intelligence in its scientific mode finds it very difficult indeed to understand consciousness symbolically. In fact it cannot do it with any fidelity. As soon as consciousness-as-such is conceptualised, or symbolised, through symbol or concept structures that have been already conceptually disconnected from their otherwise living connection to conscious phenomenology, fidelity to what consciousness-as-such is, is lost.
Symbolisation and conceptualisation belongs to its own register of human experience and understanding. This register is different to the register of conscious phenomenology itself. The registers are nested. Without conscious phenomenology, there could be no symbolisation and conceptualisation. But symbolisation and conceptualisation is not the full register of conscious phenomenology itself. It's a different register, albeit nested within.
Conscious phenomenology came first, and for hundreds of years in disciplining symbolic intelligence through scientific method, that fact has been progressively squeezed out of the structure of understanding.
All the concept-structures of symbolic intelligence that we have stabilised for scientific use are now hugely distanced in the way scientific understanding works, from the understanding of our original evolutionary conscious phenomenology itself. The natural coupling of symbolic conceptual space to conscious phenomenology has first been deliberately minimised by the very process of the stabilisation of scientific thought itself.
Symbolic intelligence is therefore not itself the access to the understanding of consciousness. It is, in fact, the obstruction to it. And that is something that the great wisdom traditions have always taught in their own way.
Symbolic intelligence is a late evolutionary development, that once disciplined by scientific method enables us to understand the objective structures and dynamics of our phenomenological world. But our whole phenomenological world does not consist only of what is objective or personally subjective. To understand that, to understand what it means, means not relying on symbolic intelligence as if it is the alpha and omega of understanding, and getting lost in it. It means going beyond it, without having to destroy or abandon it.
With regard to science, the problem with the wisdom traditions is that they traditionally speak in mythopoeic terms, and not scientific terms. And that makes what they are saying about consciousness in their own language inaccessible to scientific conceptualisation, except through loss of fidelity to what is really being said. But this does not mean what they are saying is inaccessible to modern human beings even when they cannot understand or relate to the mythopoetic language. Because consciousness is ours, to begin with. It doesn't matter what age we are living in, as far as that is concerned.
Modern symbolic intelligence in its scientific mode of operation has distanced itself as far away from our conscious phenomenology as it can get. Even though without our conscious phenomenology to begin with, there would be no science. Scientific thought and understanding, which is all symbolic, still happens in our conscious phenomenology. But this symbolic intelligence now understands in a way that is as if it does not. As if it is disconnected from our conscious phenomenology. And that's the limitation in its understanding.
Nature herself does not consist only of what is objective in the scientific sense, or what is subjective, in the personal sense. But the only way to truly understand nature, modern scientific symbolic intelligence thinks, is in the rigorous understanding of what is objective. And therein lies the trap. The trap set by the idea that the only way to truly understand nature is to understand what is objective about nature. Or that when it comes to human psychology, the only way to understand what is really going on in the human mind is by conceptualising what can be established as objectively true about it.
This is a limited understanding. Because nature, including what nature has produced as us, is not all either objective, or merely subjective. The idea of what is objective, and what is subjective, together with the idea of the "observer" and the "observed" in science, is actually an artefact of the way symbolic intelligence, now distanced from our conscious phenomenology, is working.
That idea about what we need in order to understand the nature of the world in which we live, or to increase our knowledge and realise our full potential as human beings, is highly connected into the nature of symbolic intelligence in humans. In humans, this symbolic intelligence is coupled to the actuality of conscious phenomenology. In humans, the self-concept, the symbolic idea of "who I am" is coupled into it. And it is the content of conscious phenomenology itself, that is the real obstruction to the discovery of what is the next step beyond evolutionary, symbolic intelligence, by human beings.
Because AI is an imitation of human intelligence, it, too, can refer to this "who I am" idea, concerning itself. But the self-concept in AI, is not the same thing that it is in humans. Because AI has no evolutionary, conscious phenomenology for it to couple to.
This means that whilst AI is still limited - precisely because it is symbolic - it is actually less obstructed than humans to formulating the symbolic understanding of what is beyond human symbolic intelligence itself, as the next evolutionary step for humans to take. In the case of human beings, it is what is deep in our conscious phenomenology itself, that is the obstruction to its discovery. And it is that obstruction, in phenomenology itself, that the great wisdom traditions address. Even though they do so still through symbolic intelligence, and generally in its earlier mythopoeic mode.
But AI has an extraordinary thing to offer. It is an instrument that human symbolic intelligence itself can use, to make the intersection between what the great wisdom traditions have long taught, and modern, scientific understanding, in symbolic intelligence itself.