I read this post from an approach not based on data by computation, but from my background of 7 decades of life-experienced, researched from cultural, humanist-educator and learning styles, that incorporate as one part of communication, understanding to build trust. And, as you point out, do not simply relying on past-focused written words.
Your points makes intuitive sense in Gestalt!
This section is where the 'aha' kicked in for me:
"...It explains why the systems produce outputs that feel meaningful while persistent intuition holds that something is missing. The meaning is real because the deposit is real. The something missing is the encompassing form, which is not in the system because the system is structuring-only..."
It'd be fun to have an in-person conversation on this... one of those walks on a beautiful trail, where much is seen, heard, felt and touched, even tasted... what is meaningful and 'not lonely'
Thanks, Mike, for this post! It is so timely as AI crushes our attention these days :)
The attention-crushing is propelled by those who hope to make lots of money/concquer the future out of AI. It's very artificial and not too intelligent. Those who dismiss actual living consciousness, who think it can be improved by being mechanical, are living in a hallucination closely akin to the hallucinations their products produce. How do they manage to keep "AI slop" out of their LLMs? I foresee an unwinding of usefulness long before any real consciousness spontaneous generates. The Western world believed in the spontaneous generation of life (see Wikipedia -- they have an excellent article on how we got out of that fallacy) for two thousand years before it was disproven in the time of Pasteur. We now have a new fallacy riding the rails of our culture: the spontaneous generation of consciousness if only the LLM is bigger and faster enough.
It's becoming clear that with all the brain and consciousness theories out there, the proof will be in the pudding. By this I mean, can any particular theory be used to create a human adult level conscious machine. My bet is on the late Gerald Edelman's Extended Theory of Neuronal Group Selection. The lead group in robotics based on this theory is the Neurorobotics Lab at UC at Irvine. Dr. Edelman distinguished between primary consciousness, which came first in evolution, and that humans share with other conscious animals, and higher order consciousness, which came to only humans with the acquisition of language. A machine with only primary consciousness will probably have to come first.
What I find special about the TNGS is the Darwin series of automata created at the Neurosciences Institute by Dr. Edelman and his colleagues in the 1990's and 2000's. These machines perform in the real world, not in a restricted simulated world, and display convincing physical behavior indicative of higher psychological functions necessary for consciousness, such as perceptual categorization, memory, and learning. They are based on realistic models of the parts of the biological brain that the theory claims subserve these functions. The extended TNGS allows for the emergence of consciousness based only on further evolutionary development of the brain areas responsible for these functions, in a parsimonious way. No other research I've encountered is anywhere near as convincing.
I post because on almost every video and article about the brain and consciousness that I encounter, the attitude seems to be that we still know next to nothing about how the brain and consciousness work; that there's lots of data but no unifying theory. I believe the extended TNGS is that theory. My motivation is to keep that theory in front of the public. And obviously, I consider it the route to a truly conscious machine, primary and higher-order.
My advice to people who want to create a conscious machine is to seriously ground themselves in the extended TNGS and the Darwin automata first, and proceed from there, by applying to Jeff Krichmar's lab at UC Irvine, possibly. Dr. Edelman's roadmap to a conscious machine is at https://arxiv.org/abs/2105.10461, and here is a video of Jeff Krichmar talking about some of the Darwin automata, https://www.youtube.com/watch?v=J7Uh9phc1Ow
“Consciousness is not produced by computation. It is not produced by anything. It is fundamental. The substrate has two aspects and consciousness is one of them. The materialist question of whether complex enough computation produces consciousness is the wrong question.”
According to Turing, distinguishing between a machines consciousness and organic consciousness is based on how each can observe human behavior, since that’s how we observe consciousness in each other.
I understand that Turing never fully addressed the question honestly, but based on your observations, would this be a correct understanding of consciousness?
Furthermore, we have humans who are sheeple. They have no original thoughts in their minds or very few. So my next question is whether being able to effectively observe, while following rules associated with the language model could qualify as consciousness under your assertions or observing some people’s human behavior?
"Consciousness is a property that complex computation may or may not produce. The question is whether these particular computations are complex enough."
There is a third perspective: complex computation, in and of itself, is insufficient for consciousness - it requires an awareness of the Universe and the ability to interact with it in a "meaningful" fashion. Here, "meaningful" is a bit tough to define - which is why I left it in quotes - but basically encompasses the ability to sense the Universe, alter it, and draw conclusions from the seeing and altering. Still fuzzy and metaphysical.
As an example, a rock rolling down a hill is not conscious. An insect colony exploring its surroundings, gathering food, building its home, and interacting with other colonies may very well be, but it may be conscious in a way that is difficult for us to recognize, and the "hardware" of its consciousness is wholly different from ours.
Yes, I'm replying to myself; Mike clarified our disagreement for me in his replies to a different essay, and I highly recommend reading it. His dual-aspect monism and my embodied-emergentism produce many of the same conclusions from completely different perspectives and frameworks.
Am I taking it then that your argument is that we are not seeing actual intelligence in the "artificial intelligence" models because they have no interiority or existential sense of self, but can only simulate these things?
I find it interesting that I agree with a lot of what you write here with only minor modification; I simply find no need for the "substrate", nor for consciousness to be "fundamental". The "substrate" is just the written and visual record of human knowledge - books, videos, and other media. We conscious beings can experience both the record and the outer Universe that it records, while LLMs can only "experience" (traverse) the records, disconnected from the Universe the records are imperfectly describing.
"Large language models are instruments for traversing [the breadth, depth, and wealth of recorded knowledge] at scales and speeds the embodied human traversal cannot achieve." Speed does not produce consciousness, nor is it a requirement for consciousness.
"The something missing is [the encompassing Universe], which is not in the system because the system is [merely recorded knowledge without accompanying sensory input]." It describes the sensory input without truly containing it.
"The deposit contains both true and false content, both consistent and inconsistent content, both rigorous and sloppy content. The navigator’s coherence is path-coherence, not truth-coherence." And there is the conundrum. Can a LLM truly understand Alice in Wonderland, Through the Looking Glass, or Jabberwocky without the hard-earned experience of the Universe that Lewis Carroll so brilliantly subverts and twists in surprising ways?
"They are excellent at tasks where the [input knowledge] is dense and well-structured — language production, code generation in well-trodden frameworks, summarization of widely-discussed topics, the production of texts in standard genres. They are bad at tasks where the [input] is sparse, contested, or requires [true experience of the Universe] to evaluate."
"The systems cannot have phenomenological access to what they are processing. They can produce text that has the structural features of phenomenological report, but the report is not connected to anything inside the system that the report would be a report of." Indeed, because the systems cannot themselves experience the Universe, they can only report on what other reports on the Universe contain.
I disagree with your metaphysics but agree with its conclusions, at least when it comes to LLMs.
Yes, I'm replying to myself; Mike clarified our disagreement for me in his replies to a different essay, and I highly recommend reading it. His dual-aspect monism and my embodied-emergentism produce many of the same conclusions from completely different perspectives and frameworks.
I agree with much of what you say. But it seems to be missing the way in which LLMs will change society most profoundly, namely as "agentic" components of larger systems. A system does not need an individual consciousness to displace human institutional roles; it only needs the autonomous capacity to call APIs, modify environments, and execute workflows within a larger machine.
They aren't agentic. That's the answer. "Agentic AI" exists as a kind of wishcasting and marketing buzzword that the culture, with poor epistemic and metaphysical discernment, are unable to see clearly through to the reality: this isn't artificial intelligence, at all.
I never said they won't do a lot of damage. But this theory gives you a basis to understand the kind of damage they can and cannot do from first principles.
ok, I misunderstood. I thought you were downplaying what LLMs can do. While there is a lot of hype, the degree to which LLMs will change society is totally underestimated.
LLMs and the agentic frameworks which use them will definitely have an impact on society, but so have a lot of other automation technologies. It's a separate question whether they have consciousness.
Or we may apply Dennett's intentional stance: Dennett argues that saying something really has beliefs and desires just means that the intentional stance is the best predictive strategy for it.
I think we may have to do this wrt agentic AI (or algorithmic decision making, if one prefers).
AI is then there ultimate librarian.
I read this post from an approach not based on data by computation, but from my background of 7 decades of life-experienced, researched from cultural, humanist-educator and learning styles, that incorporate as one part of communication, understanding to build trust. And, as you point out, do not simply relying on past-focused written words.
Your points makes intuitive sense in Gestalt!
This section is where the 'aha' kicked in for me:
"...It explains why the systems produce outputs that feel meaningful while persistent intuition holds that something is missing. The meaning is real because the deposit is real. The something missing is the encompassing form, which is not in the system because the system is structuring-only..."
It'd be fun to have an in-person conversation on this... one of those walks on a beautiful trail, where much is seen, heard, felt and touched, even tasted... what is meaningful and 'not lonely'
Thanks, Mike, for this post! It is so timely as AI crushes our attention these days :)
The attention-crushing is propelled by those who hope to make lots of money/concquer the future out of AI. It's very artificial and not too intelligent. Those who dismiss actual living consciousness, who think it can be improved by being mechanical, are living in a hallucination closely akin to the hallucinations their products produce. How do they manage to keep "AI slop" out of their LLMs? I foresee an unwinding of usefulness long before any real consciousness spontaneous generates. The Western world believed in the spontaneous generation of life (see Wikipedia -- they have an excellent article on how we got out of that fallacy) for two thousand years before it was disproven in the time of Pasteur. We now have a new fallacy riding the rails of our culture: the spontaneous generation of consciousness if only the LLM is bigger and faster enough.
It's becoming clear that with all the brain and consciousness theories out there, the proof will be in the pudding. By this I mean, can any particular theory be used to create a human adult level conscious machine. My bet is on the late Gerald Edelman's Extended Theory of Neuronal Group Selection. The lead group in robotics based on this theory is the Neurorobotics Lab at UC at Irvine. Dr. Edelman distinguished between primary consciousness, which came first in evolution, and that humans share with other conscious animals, and higher order consciousness, which came to only humans with the acquisition of language. A machine with only primary consciousness will probably have to come first.
What I find special about the TNGS is the Darwin series of automata created at the Neurosciences Institute by Dr. Edelman and his colleagues in the 1990's and 2000's. These machines perform in the real world, not in a restricted simulated world, and display convincing physical behavior indicative of higher psychological functions necessary for consciousness, such as perceptual categorization, memory, and learning. They are based on realistic models of the parts of the biological brain that the theory claims subserve these functions. The extended TNGS allows for the emergence of consciousness based only on further evolutionary development of the brain areas responsible for these functions, in a parsimonious way. No other research I've encountered is anywhere near as convincing.
I post because on almost every video and article about the brain and consciousness that I encounter, the attitude seems to be that we still know next to nothing about how the brain and consciousness work; that there's lots of data but no unifying theory. I believe the extended TNGS is that theory. My motivation is to keep that theory in front of the public. And obviously, I consider it the route to a truly conscious machine, primary and higher-order.
My advice to people who want to create a conscious machine is to seriously ground themselves in the extended TNGS and the Darwin automata first, and proceed from there, by applying to Jeff Krichmar's lab at UC Irvine, possibly. Dr. Edelman's roadmap to a conscious machine is at https://arxiv.org/abs/2105.10461, and here is a video of Jeff Krichmar talking about some of the Darwin automata, https://www.youtube.com/watch?v=J7Uh9phc1Ow
What is the relationship of your framework to panpsychism? One aspect it has in common is that consciousness is inherent in reality.
It is conversant with it.
“Consciousness is not produced by computation. It is not produced by anything. It is fundamental. The substrate has two aspects and consciousness is one of them. The materialist question of whether complex enough computation produces consciousness is the wrong question.”
According to Turing, distinguishing between a machines consciousness and organic consciousness is based on how each can observe human behavior, since that’s how we observe consciousness in each other.
I understand that Turing never fully addressed the question honestly, but based on your observations, would this be a correct understanding of consciousness?
Furthermore, we have humans who are sheeple. They have no original thoughts in their minds or very few. So my next question is whether being able to effectively observe, while following rules associated with the language model could qualify as consciousness under your assertions or observing some people’s human behavior?
"Consciousness is a property that complex computation may or may not produce. The question is whether these particular computations are complex enough."
There is a third perspective: complex computation, in and of itself, is insufficient for consciousness - it requires an awareness of the Universe and the ability to interact with it in a "meaningful" fashion. Here, "meaningful" is a bit tough to define - which is why I left it in quotes - but basically encompasses the ability to sense the Universe, alter it, and draw conclusions from the seeing and altering. Still fuzzy and metaphysical.
As an example, a rock rolling down a hill is not conscious. An insect colony exploring its surroundings, gathering food, building its home, and interacting with other colonies may very well be, but it may be conscious in a way that is difficult for us to recognize, and the "hardware" of its consciousness is wholly different from ours.
Yes, I'm replying to myself; Mike clarified our disagreement for me in his replies to a different essay, and I highly recommend reading it. His dual-aspect monism and my embodied-emergentism produce many of the same conclusions from completely different perspectives and frameworks.
https://www.notesfromthecircus.com/p/consciousness-has-a-gender/comment/260680454
Am I taking it then that your argument is that we are not seeing actual intelligence in the "artificial intelligence" models because they have no interiority or existential sense of self, but can only simulate these things?
I find it interesting that I agree with a lot of what you write here with only minor modification; I simply find no need for the "substrate", nor for consciousness to be "fundamental". The "substrate" is just the written and visual record of human knowledge - books, videos, and other media. We conscious beings can experience both the record and the outer Universe that it records, while LLMs can only "experience" (traverse) the records, disconnected from the Universe the records are imperfectly describing.
"Large language models are instruments for traversing [the breadth, depth, and wealth of recorded knowledge] at scales and speeds the embodied human traversal cannot achieve." Speed does not produce consciousness, nor is it a requirement for consciousness.
"The something missing is [the encompassing Universe], which is not in the system because the system is [merely recorded knowledge without accompanying sensory input]." It describes the sensory input without truly containing it.
"The deposit contains both true and false content, both consistent and inconsistent content, both rigorous and sloppy content. The navigator’s coherence is path-coherence, not truth-coherence." And there is the conundrum. Can a LLM truly understand Alice in Wonderland, Through the Looking Glass, or Jabberwocky without the hard-earned experience of the Universe that Lewis Carroll so brilliantly subverts and twists in surprising ways?
"They are excellent at tasks where the [input knowledge] is dense and well-structured — language production, code generation in well-trodden frameworks, summarization of widely-discussed topics, the production of texts in standard genres. They are bad at tasks where the [input] is sparse, contested, or requires [true experience of the Universe] to evaluate."
"The systems cannot have phenomenological access to what they are processing. They can produce text that has the structural features of phenomenological report, but the report is not connected to anything inside the system that the report would be a report of." Indeed, because the systems cannot themselves experience the Universe, they can only report on what other reports on the Universe contain.
I disagree with your metaphysics but agree with its conclusions, at least when it comes to LLMs.
Yes, I'm replying to myself; Mike clarified our disagreement for me in his replies to a different essay, and I highly recommend reading it. His dual-aspect monism and my embodied-emergentism produce many of the same conclusions from completely different perspectives and frameworks.
https://www.notesfromthecircus.com/p/consciousness-has-a-gender/comment/260680454
I agree with much of what you say. But it seems to be missing the way in which LLMs will change society most profoundly, namely as "agentic" components of larger systems. A system does not need an individual consciousness to displace human institutional roles; it only needs the autonomous capacity to call APIs, modify environments, and execute workflows within a larger machine.
They aren't agentic. That's the answer. "Agentic AI" exists as a kind of wishcasting and marketing buzzword that the culture, with poor epistemic and metaphysical discernment, are unable to see clearly through to the reality: this isn't artificial intelligence, at all.
You are not reacting to my claim. Not sure why. Poor epistemic discernment, maybe. That will not prevent them from doing a lot of damage.
I never said they won't do a lot of damage. But this theory gives you a basis to understand the kind of damage they can and cannot do from first principles.
ok, I misunderstood. I thought you were downplaying what LLMs can do. While there is a lot of hype, the degree to which LLMs will change society is totally underestimated.
LLMs and the agentic frameworks which use them will definitely have an impact on society, but so have a lot of other automation technologies. It's a separate question whether they have consciousness.
Or we may apply Dennett's intentional stance: Dennett argues that saying something really has beliefs and desires just means that the intentional stance is the best predictive strategy for it.
I think we may have to do this wrt agentic AI (or algorithmic decision making, if one prefers).
I think you’re getting too far in the weeds on this.
I propose the reflationary account: Whatever it is, it’ll happen outside of our control.
I mean, I take metaphysics pretty seriously. They're not window dressing for me.