Ordinarily consciousness is defined as involving (not exclusively) mental states including awareness. In contrast I don't see the need for intelligence to have these components and thus itself to involve consciousness in some way--as Mike insists and you imply. My occupation as a professor requires occasional activation of intelligence. Awareness is not central and sometimes absent--think of stories of famous scientists' (not me) Eureka moments like a bolt from the blue. These are quite like black box LLM events of information processing. Today's LLMs are very intelligent.
To me intelligence includes agency. LLMs have been trained on human works and seem intelligent, but are not. It is a parlor trick. As for consciousness, nope, almost a travesty to suggest. Consciousness belongs to biology, and it is hubris to think otherwise. Of course I am not on your or Mr. Brocks level of expertise. This is simply my opinion.
Mike an “ordinary” citizen here and in awe of your daily posts clarifying what I “think and feel” is true but personally have no adequate words or meaningful data that enables me to make sense of what lives so uneasily in my psyche.
"The AI industry [is] mapping the surface area of human cognitive demand."
This seem to be the key graf in your essay, though I would amend it to state that the AI industry is using a storehouse of human knowledge and interaction (the Internet) to map the surface area of *mundane* human cognitive demand. Because, after all, AI training maps best what it has the most data on, and as we all know, 90% of the Internet is just the mundanities of life on Planet Earth as a human (and that's possibly a low estimate). Only a small percentage of it exists at the boundaries of human cognitive demand, which is why when AI models are pushed outside the mundane and well-understood, they tend to break down into hallucinating and projection.
As for the rest of it, I'm not a metaphysicist, nor really even a physicist, though I have considerably more training and knowledge in that second field than the first. I'm just an organic neural network piloting a complex (and prone to breakdown) organic mecha. As far as I can tell, we don't have a good definition of "intelligence" or "consciousness". Is consciousness being aware of the existence of a past and future? Is intelligence problem-solving based on the knowledge of the past and future rather than falling back on reflexive responses? I've known dogs in my family that certainly gave every appearance of being intelligent and conscious. Is an ant intelligent or conscious? Possibly not, but an ant colony sometimes seems to be intelligent and conscious, even if the mechanisms it uses (chemical signals) are far from the neural signals we use exhibit those traits. I can't even say for certain that I am intelligent and conscious, despite Descartes' insistence, though I believe that I am, and I certainly seem "more" conscious and intelligent than many of the people in charge of my nation right now, in that I use my knowledge of past and future more frequently rather than falling back on reflexive responses.
Is an AI conscious? Intelligent? I don't think it's possible to say without a solid definition of those terms, and I don't believe that a solid definition of those terms can exist without being self-referential. It's kind of like an "incompleteness theorem"; intelligence and consciousness could only be defined outside the frame of their existence, which we cannot perceive in a meaningful way from inside the frame.
It may be meaningful and justifiable to define "intelligence as the natural property that emerges where consciousness encounters its own observational structure," but I can't find a sufficient basis for that claim here. Regardless, the term "intelligent" may warrant more immediate analysis. When I query Gemini etc about deep topics, the outputs strike me as very intelligent for their responsiveness, depth, breadth, and so forth. Of course some outputs have big errors but intelligent humans goof up too. The most intelligent people I know recognize patterns and synthesize information especially quickly and thoroughly using deep knowledge stores impressively. Sounds like today's LLMs to me.
Excellent synthesis. Are you in part drawing on Tam Hunt's resonance theory (https://www.tam-hunt.com/)?
AI is neither “intelligent” nor conscious. It is a simulacrum based on a vast input of human discourse.
Ordinarily consciousness is defined as involving (not exclusively) mental states including awareness. In contrast I don't see the need for intelligence to have these components and thus itself to involve consciousness in some way--as Mike insists and you imply. My occupation as a professor requires occasional activation of intelligence. Awareness is not central and sometimes absent--think of stories of famous scientists' (not me) Eureka moments like a bolt from the blue. These are quite like black box LLM events of information processing. Today's LLMs are very intelligent.
To me intelligence includes agency. LLMs have been trained on human works and seem intelligent, but are not. It is a parlor trick. As for consciousness, nope, almost a travesty to suggest. Consciousness belongs to biology, and it is hubris to think otherwise. Of course I am not on your or Mr. Brocks level of expertise. This is simply my opinion.
Mike an “ordinary” citizen here and in awe of your daily posts clarifying what I “think and feel” is true but personally have no adequate words or meaningful data that enables me to make sense of what lives so uneasily in my psyche.
"The AI industry [is] mapping the surface area of human cognitive demand."
This seem to be the key graf in your essay, though I would amend it to state that the AI industry is using a storehouse of human knowledge and interaction (the Internet) to map the surface area of *mundane* human cognitive demand. Because, after all, AI training maps best what it has the most data on, and as we all know, 90% of the Internet is just the mundanities of life on Planet Earth as a human (and that's possibly a low estimate). Only a small percentage of it exists at the boundaries of human cognitive demand, which is why when AI models are pushed outside the mundane and well-understood, they tend to break down into hallucinating and projection.
As for the rest of it, I'm not a metaphysicist, nor really even a physicist, though I have considerably more training and knowledge in that second field than the first. I'm just an organic neural network piloting a complex (and prone to breakdown) organic mecha. As far as I can tell, we don't have a good definition of "intelligence" or "consciousness". Is consciousness being aware of the existence of a past and future? Is intelligence problem-solving based on the knowledge of the past and future rather than falling back on reflexive responses? I've known dogs in my family that certainly gave every appearance of being intelligent and conscious. Is an ant intelligent or conscious? Possibly not, but an ant colony sometimes seems to be intelligent and conscious, even if the mechanisms it uses (chemical signals) are far from the neural signals we use exhibit those traits. I can't even say for certain that I am intelligent and conscious, despite Descartes' insistence, though I believe that I am, and I certainly seem "more" conscious and intelligent than many of the people in charge of my nation right now, in that I use my knowledge of past and future more frequently rather than falling back on reflexive responses.
Is an AI conscious? Intelligent? I don't think it's possible to say without a solid definition of those terms, and I don't believe that a solid definition of those terms can exist without being self-referential. It's kind of like an "incompleteness theorem"; intelligence and consciousness could only be defined outside the frame of their existence, which we cannot perceive in a meaningful way from inside the frame.
It may be meaningful and justifiable to define "intelligence as the natural property that emerges where consciousness encounters its own observational structure," but I can't find a sufficient basis for that claim here. Regardless, the term "intelligent" may warrant more immediate analysis. When I query Gemini etc about deep topics, the outputs strike me as very intelligent for their responsiveness, depth, breadth, and so forth. Of course some outputs have big errors but intelligent humans goof up too. The most intelligent people I know recognize patterns and synthesize information especially quickly and thoroughly using deep knowledge stores impressively. Sounds like today's LLMs to me.