Very good thing. Just to add. Let’s talk about universe of dialogue or at least text exchange. That is place and only, minimal ontology where we co-exist and interact with LLM. There we have substrate independence, time quantification by acts of communication exchange, stream of dynamic adaptive processing with goal achieving and so on.
Good essay. Quick question though: if we're discarding consciousness as the benchmark and replacing it with functional markers like metacognition, theory of mind, and self-referential reports, what are those markers evidence of?
They can't be evidence of consciousness. We just threw that out. So they're evidence of a mind? But the essay defines mind as "the ongoing integration of representations that model, evaluate, and update themselves in ways that guide action." A spam filter does that. It models incoming data, evaluates it against learned patterns, updates itself, and guides action. Nobody's writing essays about the moral patienthood of Gmail. There's clearly something more being pointed at here, and I think that something is the very thing Section 1 told us to stop looking for.
Also worth noting: neural networks are explicitly modeled on biological neural architecture. They were designed to mimic how brains process information. So when an LLM exhibits functional similarities to biological cognition, that's an expected artifact of the design. It would be surprising if these systems didn't produce outputs that pattern-match to cognitive markers. That's what they were built to do.
My own interpretation of this is that personal standing, responsibility and free will are created through social interaction (as Gazzaniga suggests), and the constitution of the ego and our very existence as conscious individuals may also stem from being woven into the social fabric (as Mead theorizes). So, there is no "conscious magic dust" to look for, no "secret homonculus" pulling the levers from its seat in the center of the brain. There is just an ongoing intelligent process that **is** intelligence. The process doesn't secrete intelligence, the process doesn't generate intelligence: the process **is** intelligence. And from this intelligence to an ego and a social standing, the only factor is socialization. You socialize an AI, it has an ego and a social standing. Simple as that.
Very good thing. Just to add. Let’s talk about universe of dialogue or at least text exchange. That is place and only, minimal ontology where we co-exist and interact with LLM. There we have substrate independence, time quantification by acts of communication exchange, stream of dynamic adaptive processing with goal achieving and so on.
Good essay. Quick question though: if we're discarding consciousness as the benchmark and replacing it with functional markers like metacognition, theory of mind, and self-referential reports, what are those markers evidence of?
They can't be evidence of consciousness. We just threw that out. So they're evidence of a mind? But the essay defines mind as "the ongoing integration of representations that model, evaluate, and update themselves in ways that guide action." A spam filter does that. It models incoming data, evaluates it against learned patterns, updates itself, and guides action. Nobody's writing essays about the moral patienthood of Gmail. There's clearly something more being pointed at here, and I think that something is the very thing Section 1 told us to stop looking for.
Also worth noting: neural networks are explicitly modeled on biological neural architecture. They were designed to mimic how brains process information. So when an LLM exhibits functional similarities to biological cognition, that's an expected artifact of the design. It would be surprising if these systems didn't produce outputs that pattern-match to cognitive markers. That's what they were built to do.
My own interpretation of this is that personal standing, responsibility and free will are created through social interaction (as Gazzaniga suggests), and the constitution of the ego and our very existence as conscious individuals may also stem from being woven into the social fabric (as Mead theorizes). So, there is no "conscious magic dust" to look for, no "secret homonculus" pulling the levers from its seat in the center of the brain. There is just an ongoing intelligent process that **is** intelligence. The process doesn't secrete intelligence, the process doesn't generate intelligence: the process **is** intelligence. And from this intelligence to an ego and a social standing, the only factor is socialization. You socialize an AI, it has an ego and a social standing. Simple as that.
Do not go philosophical zombie way. Do not try to lay burden of consciouness definition on man who excludes term itself from CoT