One of the most profound insights revealed by large language models is that language is not just autoregressive—it is autogenerative (a term I made up; sue me).
a view from a different perch. the origin of language is the discovery that moral cooperation is generative within a shared context. language expands the shared context and maintains the cooperation. this creates collective intelligence processing about the shared context. language scales moral cooperation endlessly, is composable and is antifragile to stressors of new language, new uses and narrative control. language is the cooperation protocol and contains the value exchanged of processed intelligence about the shared context. stressors welcome AO
I'm glad you're opening up to language not being unique! As you perhaps know, I think language is only unique by virtue of its complexity, which allows for memetic evolution and replication. This is indeed a big deal, but there are no unique fundamental features that no animals share with us.
Oh, language IS motor control! It's a special case of motor control to manipulate the external world (and reflectively the internal world in the case of talking/thinking to oneself).
Yes, given starting conditions and laws of nature everything unfolds.
Language is unique because it marks a qualitative break from all other biological systems. It is not just a more complex form of motor control or perception—it is symbolic (not in the naive sense of symbolizing something), combinatorial, and recursive. This shifts it into an entirely different computational regime.
I’m not persuaded that language occupies a wholly separate “computational regime.” Framing it that way assumes that symbolic reference, combinatoriality, and recursion erupt fully formed, rather than unfolding along the same evolutionary gradients and neural gradients that shape every other biological signaling system. Compression in the service of entropy minimization keeps the story simpler and, I’d argue, more defensible: adaptive systems minimize future uncertainty at the lowest metabolic cost, steadily evolving ever tighter codes in the process. In that light, “symbols” are just especially durable compression shortcuts stabilized by social convention, while combinatoriality and recursion emerge naturally as each round of compression feeds into the next. This perspective explains linguistic structure with the same mechanisms that already account for gene regulation, motor planning, and predictive perception with no ontological leap required. A compression in the service of entropy minimization model integrates the dynamics of perception and symbolic language within a single informational framework defined by entropic pruning.
Language is the bridge that artificially connects thoughts to the sensual (natural) realm. Words exist as 1D "strings" on a 2D plane within the 3D realm. These words are being experienced through the visual sense while simultaneously generating the disembodied "voice" within consciousness. Words are in a superpositional state of existing in both the realm of the senses (body/particle) and the realm of thought (mind/wave function). If you make this subtle shift your model will be complete. The crux of it all is the generation of the voice in consciousness. This is the ego, the illusion of a separately existing individual self.
You have revealed THE problem. And it will be so very difficult for people to accept it. I've been trying to point out that language has nothing to do with the realm of the senses for a couple of years now. I've been ignored.
Do you see someone talking to you when that voice is speaking? It's disembodied.
The mind realm is a result of the evolutionary step into a higher level of Self-Awareness. Part of that collective consciousness advanced into planetary consciousness and the other part is humanity. Thought is an artificial construct of true Self. It's the construction of a Ponzi scheme where body-mind attempts to bootstrap reality without recognition of SPIRIT, true Self
No I don't see anyone. So when I don't see the source of a voice, it's disembodied??
I think of that voice as a simulation of actual voices. So yeah, as such, it's not an actual voice travelling through air. It's neuronal simulation. I can agree to call that disembodied if it helps! But I don't see any mysteries here.
Well if you can't see how the triune Fundamental Model of Reality that I shared, is exactly that, I can't simplify it any more than it already is. The entirety of reality and the illusion is expressed in 3 characters.
It is that voice from nowhere that projects the illusion of duality. This is why Ramana Maharshi taught the transcendence of thought
Symbols always symbolise something or else they are not symbols. Even binary strings generated by a RNG symbolise something. They symbolise pseudo-randomness.
You may not be able to deduce what a symbol such as a digit or a word may stand in for, but it's nonetheless a symbol.
If I ask you to tell me about your childhood the words you produce will symbols in this real, naive sense.
If we clone you or upload you, the same is true.
If we train an LLM with all the training data that you are trained on, it will be able to tell us about your childhood and those words will symbolise your childhood in the same way.
----
The bee's dance is symbolic, combinatorial and recursive. Not to the extent that our language is, but that's beside the point.
Humans probably are genetically coded to exploring various vocalisations combined with opening and closing the lips--producing 'B' /'V' like sounds. Or maybe even more primitive components are hard coded. In bees, perhaps several wiggles or sequences are hard coded. Doesn't matter. Both cases fulfill your requirements. There is no new qualitative or computational regime. There is no point in a baby's development when it suddenly acquires an LLM.
Another good example of symbolic, recursive and combinatorial behavior is dancing in humans. Is that a new computational regime?
Autogeneration probably arises though sef reference and feedback, leading to the formation of new structure and complexity. I belive this to be a fundemental feature in our universe and of conciousness. What is interesting is that I sense that the intensity of autogeneration and recursion is related to truth. For example if you were to have a LLM conversation that is banal or about flat earth, the recursive depth is limited. However, a more involved converstaion leads to added layers of complexity and autogeneration. I dont know if anyone lse has had this exoerience.
Yes, absolutely and well said! Language is the glue of social coordination, not just a medium of expression, but a protocol for synchronizing action and values—which really are latent variables for actions. In that sense, the superorganism may be said to have its own autoregressive dynamics, composed of those of its human constituents.
Hi, love your ideas (halfway through the TOE podcast) and Id like to know if any of them align with Daniel C Dennet's? In particular, the illusion of Qualia. I feel like the embedded language space you speak about is not only a semantic network, but also a functional map of conceptual possibilities. Therefore, is it just words all the way down? That is, are mental symbols not only used to communicate, but to think with. The symbol grounding problem is avoided because even though no single symbol is mapped to concrete or platonic concepts, the entire system of language has recreated in a sense, a complete model of reality.
Thank you and great question. I have always been —and still am—dead set against the eliminativist/illusionist perspective of qualia. I don’t think we can doubt the existence of the subjective. But LLMs and what they reveal about language adds complexity to the issue —and maybe inching towards a “resolution” of sorts. Because in my view, language has no mapping to the underlying sensory processes that give rise to qualia. Thus, the hard problem may specifically be a linguistic one! I spoke about this recently
Of course this raises some paradoxes/riddles about what it means to talk about qualia but of course there’s a lot that needs to be worked about this account of language before we get to these “hard” problems
Hope that I'm not gate crashing too much here. As an ex applied psychologist (military) and information systems engineer (predominantly, with some ML knowledge), semi-retired now with time to think about such things as the physics of reality and consciousness (!), I find myself strongly relating to your intuitions Elan. Presumably autogeneration can also be expected to be applicable to visual, auditory and proprioceptive etc aspects of brain function (own neural real-estate). Perhaps also layers of reasoning ( neither language nor sensory, why limit to those "dimensions") that reference reasoning of subordinates with token output such as feelings of correctness or not, hence intuition (and indeed feelings).
"In the beginning there was the word, and the word was with God and the word was God."
Does your theory suggest that there could be a realm of pure mind, devoid of sensory experience?
My theory suggests that the realm of the 5 senses is concluding within human consciousness due to our shifting of attention onto that which thought has created (the digital realm and the screen). Humanity is experiencing a transitionary stage between two HUGE iterations of the cycle of Samsara. These phases/iterations are the illusions of duality (body-mind, yang-yin, particle-wave, birth-death, + -, etc) that are projected upon ignorance of Source (SPIRIT, TAO, OBSERVER, LIFE, 0). This ignorance arises due to our interest in the illusion, like how a player of a video game loses awareness of body and surroundings when he is deeply engaged in the game.
The realm of the senses and the realm of thought are like the convex and concave of ). They can't be unified without unification within the "third" state, which is the state of superposition, which is the Reality. All forms of duality are impossible to unify because the appearance of the duality is an illusion. Unification is recognized as the natural state, which is pure Potential, 0.
I get what you're saying about language but I think the distinction needs to be between the realm of the senses and the realm of thought. Language/ symbol is the bridge/conduit through which Maya (the ignorance) is trying to bypass the Reality to keep the Ponzi scheme in tact. The goal of Maya is to keep attention on the illusion. It wants your attention to be pulled into the next iteration, which is yin-centric (mind, thoughts) rather than yang-centric (body/sensations). The digital realm is a type of purgatory, a transitional stage between the yang and yin iterations.
a view from a different perch. the origin of language is the discovery that moral cooperation is generative within a shared context. language expands the shared context and maintains the cooperation. this creates collective intelligence processing about the shared context. language scales moral cooperation endlessly, is composable and is antifragile to stressors of new language, new uses and narrative control. language is the cooperation protocol and contains the value exchanged of processed intelligence about the shared context. stressors welcome AO
I'm glad you're opening up to language not being unique! As you perhaps know, I think language is only unique by virtue of its complexity, which allows for memetic evolution and replication. This is indeed a big deal, but there are no unique fundamental features that no animals share with us.
Oh, language IS motor control! It's a special case of motor control to manipulate the external world (and reflectively the internal world in the case of talking/thinking to oneself).
Yes, given starting conditions and laws of nature everything unfolds.
Language is unique because it marks a qualitative break from all other biological systems. It is not just a more complex form of motor control or perception—it is symbolic (not in the naive sense of symbolizing something), combinatorial, and recursive. This shifts it into an entirely different computational regime.
I’m not persuaded that language occupies a wholly separate “computational regime.” Framing it that way assumes that symbolic reference, combinatoriality, and recursion erupt fully formed, rather than unfolding along the same evolutionary gradients and neural gradients that shape every other biological signaling system. Compression in the service of entropy minimization keeps the story simpler and, I’d argue, more defensible: adaptive systems minimize future uncertainty at the lowest metabolic cost, steadily evolving ever tighter codes in the process. In that light, “symbols” are just especially durable compression shortcuts stabilized by social convention, while combinatoriality and recursion emerge naturally as each round of compression feeds into the next. This perspective explains linguistic structure with the same mechanisms that already account for gene regulation, motor planning, and predictive perception with no ontological leap required. A compression in the service of entropy minimization model integrates the dynamics of perception and symbolic language within a single informational framework defined by entropic pruning.
I agree
Language is the bridge that artificially connects thoughts to the sensual (natural) realm. Words exist as 1D "strings" on a 2D plane within the 3D realm. These words are being experienced through the visual sense while simultaneously generating the disembodied "voice" within consciousness. Words are in a superpositional state of existing in both the realm of the senses (body/particle) and the realm of thought (mind/wave function). If you make this subtle shift your model will be complete. The crux of it all is the generation of the voice in consciousness. This is the ego, the illusion of a separately existing individual self.
You have revealed THE problem. And it will be so very difficult for people to accept it. I've been trying to point out that language has nothing to do with the realm of the senses for a couple of years now. I've been ignored.
Imagine that.
Why would the voice in the mind be disembodied?
Where did this mind realm come from? When did life integrate it?
I'm on board with non-dualism / the illusion of separate self. But it seems to me you're replacing that dualism with a sense/mind dualism!
Do you see someone talking to you when that voice is speaking? It's disembodied.
The mind realm is a result of the evolutionary step into a higher level of Self-Awareness. Part of that collective consciousness advanced into planetary consciousness and the other part is humanity. Thought is an artificial construct of true Self. It's the construction of a Ponzi scheme where body-mind attempts to bootstrap reality without recognition of SPIRIT, true Self
https://www.nonconceptuality.org/1-fundamental-model-of-reality
No I don't see anyone. So when I don't see the source of a voice, it's disembodied??
I think of that voice as a simulation of actual voices. So yeah, as such, it's not an actual voice travelling through air. It's neuronal simulation. I can agree to call that disembodied if it helps! But I don't see any mysteries here.
Well if you can't see how the triune Fundamental Model of Reality that I shared, is exactly that, I can't simplify it any more than it already is. The entirety of reality and the illusion is expressed in 3 characters.
It is that voice from nowhere that projects the illusion of duality. This is why Ramana Maharshi taught the transcendence of thought
Symbols always symbolise something or else they are not symbols. Even binary strings generated by a RNG symbolise something. They symbolise pseudo-randomness.
You may not be able to deduce what a symbol such as a digit or a word may stand in for, but it's nonetheless a symbol.
If I ask you to tell me about your childhood the words you produce will symbols in this real, naive sense.
If we clone you or upload you, the same is true.
If we train an LLM with all the training data that you are trained on, it will be able to tell us about your childhood and those words will symbolise your childhood in the same way.
----
The bee's dance is symbolic, combinatorial and recursive. Not to the extent that our language is, but that's beside the point.
Humans probably are genetically coded to exploring various vocalisations combined with opening and closing the lips--producing 'B' /'V' like sounds. Or maybe even more primitive components are hard coded. In bees, perhaps several wiggles or sequences are hard coded. Doesn't matter. Both cases fulfill your requirements. There is no new qualitative or computational regime. There is no point in a baby's development when it suddenly acquires an LLM.
Another good example of symbolic, recursive and combinatorial behavior is dancing in humans. Is that a new computational regime?
Autogeneration probably arises though sef reference and feedback, leading to the formation of new structure and complexity. I belive this to be a fundemental feature in our universe and of conciousness. What is interesting is that I sense that the intensity of autogeneration and recursion is related to truth. For example if you were to have a LLM conversation that is banal or about flat earth, the recursive depth is limited. However, a more involved converstaion leads to added layers of complexity and autogeneration. I dont know if anyone lse has had this exoerience.
Yes, absolutely and well said! Language is the glue of social coordination, not just a medium of expression, but a protocol for synchronizing action and values—which really are latent variables for actions. In that sense, the superorganism may be said to have its own autoregressive dynamics, composed of those of its human constituents.
Hi, love your ideas (halfway through the TOE podcast) and Id like to know if any of them align with Daniel C Dennet's? In particular, the illusion of Qualia. I feel like the embedded language space you speak about is not only a semantic network, but also a functional map of conceptual possibilities. Therefore, is it just words all the way down? That is, are mental symbols not only used to communicate, but to think with. The symbol grounding problem is avoided because even though no single symbol is mapped to concrete or platonic concepts, the entire system of language has recreated in a sense, a complete model of reality.
Thank you and great question. I have always been —and still am—dead set against the eliminativist/illusionist perspective of qualia. I don’t think we can doubt the existence of the subjective. But LLMs and what they reveal about language adds complexity to the issue —and maybe inching towards a “resolution” of sorts. Because in my view, language has no mapping to the underlying sensory processes that give rise to qualia. Thus, the hard problem may specifically be a linguistic one! I spoke about this recently
https://youtu.be/5zWe4gtVXuQ
Of course this raises some paradoxes/riddles about what it means to talk about qualia but of course there’s a lot that needs to be worked about this account of language before we get to these “hard” problems
Hope that I'm not gate crashing too much here. As an ex applied psychologist (military) and information systems engineer (predominantly, with some ML knowledge), semi-retired now with time to think about such things as the physics of reality and consciousness (!), I find myself strongly relating to your intuitions Elan. Presumably autogeneration can also be expected to be applicable to visual, auditory and proprioceptive etc aspects of brain function (own neural real-estate). Perhaps also layers of reasoning ( neither language nor sensory, why limit to those "dimensions") that reference reasoning of subordinates with token output such as feelings of correctness or not, hence intuition (and indeed feelings).
"In the beginning there was the word, and the word was with God and the word was God."
Does your theory suggest that there could be a realm of pure mind, devoid of sensory experience?
My theory suggests that the realm of the 5 senses is concluding within human consciousness due to our shifting of attention onto that which thought has created (the digital realm and the screen). Humanity is experiencing a transitionary stage between two HUGE iterations of the cycle of Samsara. These phases/iterations are the illusions of duality (body-mind, yang-yin, particle-wave, birth-death, + -, etc) that are projected upon ignorance of Source (SPIRIT, TAO, OBSERVER, LIFE, 0). This ignorance arises due to our interest in the illusion, like how a player of a video game loses awareness of body and surroundings when he is deeply engaged in the game.
The realm of the senses and the realm of thought are like the convex and concave of ). They can't be unified without unification within the "third" state, which is the state of superposition, which is the Reality. All forms of duality are impossible to unify because the appearance of the duality is an illusion. Unification is recognized as the natural state, which is pure Potential, 0.
I get what you're saying about language but I think the distinction needs to be between the realm of the senses and the realm of thought. Language/ symbol is the bridge/conduit through which Maya (the ignorance) is trying to bypass the Reality to keep the Ponzi scheme in tact. The goal of Maya is to keep attention on the illusion. It wants your attention to be pulled into the next iteration, which is yin-centric (mind, thoughts) rather than yang-centric (body/sensations). The digital realm is a type of purgatory, a transitional stage between the yang and yin iterations.