What I mean is, why is the form it's in not the form that it can most easily process and act upon? — Patterner

As the brain receives a sensory signal, it is just a signal, presumably without any qualitative content. The brain has to do the work so that the signal can be interpreted by (the conscious subset of) itself as qualitative.

The key insight is that the sensory manifold we experience is not accidental or epiphenomenal. It is a highly efficient way of organizing information that would otherwise overwhelm the nervous system.

Notice how the sensation has now become a symbol. — Banno

Not a new position. The smells are symbols. Smells exhibit the characteristic one way relation of symbols. The smell points to the event, brewing coffee, but you can analyze the coffee for a thousand years and you will never derive the smell.

Note that any suitable medium can function as symbols: roads use signage, books use glyphs, brains use qualia.

Suppose that your qual of coffee changes over time, but you do not notice; so that the way coffee smells for you now is utterly different to how it smelt to you as a child.

Now, what is it that is the same now as it was when you were a child? Well, it's not the qual, since that has changed. It's the language and behaviour around the smell of coffee that has stayed the same

The system functions entirely without the consistent private symbol. It is not required. — Banno

This does not follow.

It is theoretically possible that all of our private symbols vary over our lifetimes without our noticing. But for this to work, our memories have to vary in lockstep. If they do not, then not only would we notice, but the symbols would become useless. If what we remember as coffee yesterday smells like bacon today, then the symbol does not communicate anything.

It is as if you are arguing words are not required for language to function, because theoretically words might be varying without our noticing (along with our memories and all printed text).

Or, my preference, the smell of coffee just is those chemicals, under a different description. — Banno

This is very wrong. The smell is certainly not the chemical. This feels like naive realism.

If the smell were the chemical, an alien could analyze the chemical and derive the correct subjective smell. This is clearly impossible. The smell is not a chemical property. Smells are the end product of the conjunction of the chemical and the human sensory system.

Have we? Or have we co-ordinated our behaviour? — Banno

Both. We coordinated on the basis of internal sensations. That is part of the mechanism. Lacking this, you might say "I smell coffee", I might say "I smell bacon", one would be as groundless as the next.

What more, exactly, is there to "identifying my internal, private sensation as coffee" than is found in "I smell coffee" — Banno

Nothing. One just makes explicit what is already contained in the other.

Discourse functions even if our sensations differ... so what is the place of the sensation? — Banno

Since sensations are private, there is no need for them to be consistent between people. They only have to be consistent within an individual. Smell sensations are like a private, internal symbol table. We learn by consistently matching a public event, coffee brewing, with a private symbol, the smell of coffee. Then, when we later encounter the private symbol, the smell of coffee, we can infer the public event, coffee is brewing, is nearby.

How can this system function without the private symbol? And so how would smell discourse function without the system?

But I'm not seeing how a discreet internal event isn't as problematic as a "thing"; it remains that no one else can check that you are correct. — Banno

If you claim you smell coffee, I cannot look inside your head to verify. But I can attend to my sense of smell: do I experience the internal sensation I have leaned to associate with coffee, or don't I? I do. i can confuse that you are experiencing the same smell that I am. Or more sophisticated, that both of us are experiencing the internal event we associate with coffee (even if these are different).

That these sensations off coffee may be entirely different between us is ultimately only of philosophical significance, the discourse functions the same either way. But without these sensations, the discourse wouldn't happen at all.

That everyday discourse functions without the philosophical notion of qualia is not under dispute. But what relevance is that to us purported philosophers?

So if we agree to talk of a qual of coffee - even if here, now - what is it that is being agreed on? The term "the aroma of coffee" perhaps picks out a pattern of behaviour and report, coordinates shared expectations, and is indexical but public. But it does not pick out a thing. We may avoid the hypostatisation. — Banno

There is no hypostasisation. A quale is not a material thing, it is more an event. But it is concrete, not abstract. The "aroma of coffee" picks out the subjective character of the internal event that occurs when coffee is smelled. (While patterns of behavior, expectations, etc are all real, this is just not what "the aroma of coffee" refers to).

The privacy of qualia is philosophical, it doesn't necessarily have significance in linguistic practice. In practice, there is a cousin to naive realism, call it "naive intersubjectivism": the presumption that we all internally experience basically the same things. Even though philosophically, the subjective aroma of coffee is private, by presumption we can talk about it as if everyone is talking about the same thing, everyone experiences what I experience. Moreover, that which prompts the experience is public. And so the discourse can function, even if its philosophical presuppositions are suspect.

Even if naive intersubjectivism is abandoned, we can still discuss qualia. We just abandon the idea that we can understand concretely what others experience. "The aroma of coffee" becomes relative, that which the sniffer experiences when smelling coffee. There is no singular, concrete content, but rather a conceptual structure: that particular experience which each individual undergoes.

I think I agree with everything here.

Depression is a 'mental illness' but it is neurotic, rather than psychotic. I think it better to view neurotic 'mental illnesses' as skewing or limiting perspectives on reality, rather than breaking from reality, as with psychosis. — Jeremy Murray

Maybe not a break with a reality. But certainly a break with objectivity is assessments of one's life circumstance. How can a depressive evaluate this with any objectivity?

Like you, I wouldn't be here suicide were an easy option.

For depression, I've wants to say, "but there is always hope". But can we say this with confidence? Despite having crawled out of our own black holes? How do we know that others aren't much, much deeper, so deep they are doomed never to emerge?

I was haunted reading case reports from NL of assisted suicide granted to the depressed. Here is one example:

https://www.theguardian.com/society/article/2024/may/16/dutch-woman-euthanasia-approval-grounds-of-mental-suffering

If there was memory of the qualia, then its abrupt absence would be noticed. — noAxioms

I think what you have in mind is an incomplete absence of qualia. For instance, the idea of someone losing all five senses at once. Yet, they are blindsighted in all five senses, so they can still navigate the world as before, just without conscious awareness.

But keep two things in mind:

1. Memory is also qualitative. When we remember, we remember images, sounds, feelings. These are just as much qualia as external sensations. It is just that the brain is able to bookkeep these, marking them as internal (schizophrenia might be the failure of this bookkeeping mechanism).

Someone who lost all qualitative awareness would lose the qualitative aspect of memory as well. So, there would be no mismatch, memory (no qualia) would match current situation (no qualia).

2. Feelings are also qualitative. It is not just distress that would be lost, all feelingds would be lost.

No sensations, no feelings, no memory of either of these. If the sim lost qualia, it wouldn't notice a thing.

I am currently away visiting family for holidays, which is why replies are not always prompt. — noAxioms

No worries, happy holidays!

But why does the brain present what it already understands to itself in a different form? — Patterner

Think division of responsibility. Different parts of the brain are responsible for different functions. When receiving information from the world, one part of the brain translates that information into a form that can be easily processed and acted upon. Then the executive, the conscious part, uses that translated information to learn and to act.

we need to know why there's something it is like for anybody to see red — Patterner

FWIW, my take:

We, as conscious beings, live in the perspective of the brain's executive decision maker. All qualia are informational, they tell us about the world and about ourselves. Qualia are the way that the brain presents information to itself, in a form it can efficiently process and act upon. Sensory data, bodily sensations, feelings, all just information that guide us in making decisions and ultimately acting.

Qualia exist only in the context of the brain that produces them. We are, each of us, a machine within a machine. The brain creates for us a virtual world, and then lets us executive decision makers use and act on this information.

Qualia are as real as a computer animation. They exist, but only in terms of the larger framework which supports them. There is no exotic physics, nothing mystical. Just a system organizing information in a way it is able to effectively process and act on.

Form our perspective, qualia are elemental, mysterious, and apart from the material world. But that shouldn't be surprising for virtual machines who experience everything only through qualia.

If qualia are private, then how is it that you and these others agree about them? How do you know that, when you use the term "qualia", you are all talking about the same thing? — Banno

We don't know, and can never know, that the content of our qualia agree. What most of us do agree is that there is something that it is like to see an apple and smell ammonia. This shared fact structures our collective experience, and the concept can be communicated with language. While qualitative content can never be.

The concept is public, in that it refers abstractly to universal human features most of us understand. But a part of the concept is that its concrete instantiations as qualia we individually undergo are private.

In so far as they are private sensations, they are irrelevant — Banno

You may think the core feature of conscious experience is irrelevant. Others disagree.

in so far as they are public, we already have 'red" and "sour" to cover that use — Banno

These don't generally refer to qualia. Rather, to public features. We just happen to identify these features by a internal coding system, qualia. How each of our coding systems presents to us is not communicable by language or any other means, as there is no stable referent language can latch onto.

The logical or public irreducibility of first-person ascriptions is linguistic. Stop trying to explain them by positing private inner objects such as qualia. — Banno

I'm not. I think one has nothing to do with the other. You are the one trying to shoehorn your language games into a topic far afield from them.

quote="Banno;1026883"]Folk are using “What-it-is-like” accounts, which misunderstand the source of first-person irreducibility. They treat it as arising from private inner objects along the lines of qualia, but the actual source is grammatical: the first-person pronoun designates a role within communal language-games, not a metaphysical subject of experience.[/quote]


This feels silly. How on earth do you get first person perspective from mere grammar? That is quite the trick. You point out that people conflate subject-as-language-role and subject-as-experiencer, only to then somehow reduce one to the other?

Anything with the right cognitive skillset can play communal language games. ChatGpt, for instance. But this does not track at all with whether ChatGpt is an irreducible subject. It can be the (unlikely) case that it is, and the (likely) case that it isn't. Similarly, a dog, while communicative, lacks language entirely. Certainly they have no notion of indexicals. Yet, there is something it is like to be a dog, just as much as there is for me and you.

If the matrix product of these two concepts results in a 2x2 grid, where each cell is perfectly possible, how can one have their origin in the other?

Subject, Language User: Human
Subject, Non-Language User: Dog
Non-Subject, Language User: LLM
Non-Subject, Non-Language User: Tree


I honestly don't see how anyone can conflate these. I remembered and found this quote of yours (I remembered the thread title somehow).

Going over my own notes, I found an admission that I did not understand qualia - from 2012. In 2013, I said I do not think that there is worth in giving a name to the subjective experience of a colour or a smell. In 2014, I doubted the usefulness of differentiating a smell from the experience-of-that-smell. Never understood qualia. I still don't see their purpose. — Banno

Has your understanding improved in the intervening decade plus? Is it possible that you, and maybe some others here, have a cognitive architecture that makes it difficult to fix on qualia as a distinct concept? Or, is this just another philosophical difference?

An aside - the dependence here is on context, not on subjectivity. It's not that first-person statements are subjective - whatever that might mean - that is at issue, but how we account for the place of the context in first person statements and indexicals.

That's why the guff here about qualia is irrelevant. — Banno

To me the natural conclusion is that the guff here about indexicals is irrelevant.

I reject this fantasy. If my qualia valished abruptly, I would 1) notice, 2) not feel obligated to pretend otherwise as you imply, and 3) probably not even be able to express my distress since qualia is required for a human to do almost any voluntary thing like communicate coherently. — noAxioms

You say you will notice, but this already presumes that you have the capacity to notice. If the simulation is just state and processing, there will be no distress. A faithful simulation of the human brain will, somewhere in its workings, faithfully process all the state associated with a full qualitative experience. The agent will "experience" it's qualia, and report nothing unusual. There just may not be any actual qualia.

In the same way, your chip sim faithfully processes all the state associated with electrical flow. There just isn't any actual electricity.

Answers to the negative would break the simulation.
Why? The simulation just makes the chemicals and momentums do their things. It has no high level information that it's a human being simulated. It's just a bounded box with state, suitable for simulating a heap of decaying leaves as much as anything else without any change of code. — noAxioms

If you show a human an apple and ask them what color they are experiencing, they will say 'red'. This is a behavioral fact. And so, a simulated human agent should respond in the same way. If they do not, something is wrong with the simulation. But this says nothing about whether the agent is actually experiencing any qualia at all.

hat any of the values (the velocity of the moon relative to Earth say) is discreet, meaning it is impossible to express a typical real number. — noAxioms

Computers can't process infinite precision reals, but they can process numbers precise to an arbitrary degree. At some point the result will converge so that it is not measurably different.

Disagree with this one. Computation is used, sure, but most often the purpose is not to reproduce computational features. They simulate the weather a lot, but not the computational features of the weather at all. — noAxioms

I mean 'computational' in the broad scene, where one state of a weather system physically "computes" the next. The state and only the state is what is transformed, not the substance. And if you think about it, there must be a homology between this physical "computation" and the sort of computation a computer does, otherwise the computer wouldn't have the causal power to simulate it.

You're thinking like a model ship or something, not a model of physics, the latter of which does not reproduce physical features. — noAxioms

Right. Just trying to make my little taxonomy more complete.

Well, you ask the guy if his qualia is still there. I — noAxioms

But remember, this is a simulated human. Part of a human's behavior is to respond to questions about their qualia as if they had them. Answers to the negative would break the simulation.

Just as in your circuit example. The sim circuit faithfully reports so many amps at each point. Yet, there is no actual current flowing, so in a certain sense the sim circuit is lying, as is the sim human.

he zombie behaving identically without the qualia is either inconceivable or an assertion of epiphenomenal, which is identical to fiction. — noAxioms

Why? Can you quote or restate your argument?

It very much does simulate the current, at all points. — noAxioms

You are missing the point. It simulates the current. But there is no current, just numerical values representing current. That is what separates the simulation of a circuit from the circuit. The simulation reproduces computational features of the circuit, without reproducing the circuit.

Simulation: reproduces computational features
Model: reproduces (some) physical features
Imitation: reproduces behavioral features
Reproduction: reproduces all features

And so, Does the simulated guy have qualia? It would seem this can only be true if qualia were computational. And if so, you can't build a qualia detector, as you can't build a computation detector. A computation can be performed in a vast variety of physical media, in a vast variety of ways. No physical detector can operate at that level, and detect the computation no matter how it is instantiated. It would have to be a physical property for it to be detectable.

They can be, though. Elements aren't elemental, they can be further broken down into more basic particles. But discovering this more basic structure required tremendous intellectual work.

This is the same kind of work metaphysical philosophy attempts. But, as there is no standard of success, there is no real progress, unlike the sciences. We are more or less stuck with the same basic concepts we've used for millennia.

I seen no distinction here. The sim of the chip simulates a physical chip, and thus it exhibits all the relevant physical properties. If it didn't, it would be an invalid simulation. The chip cannot tell if it's simulated or not. — noAxioms

You are missing the distinction. The sim simulates a physical chip, simulating it's physical properties. But, it exhibits none of the physical properties of the chip: not the mass, not the current at any point, nothing. At best it can report values representing these physical quantities. Of course, any simulation simulates only a narrow subset of the physical properties deemed relevant.

If the simulation was aware, it suggests the awareness is a property of information/competition, not physics.

It isn't the Turing machine that's going to have feelings, it will be the simulated person. I said as much in the OP. So its that simulated guy that has the capacity, not the Turing machine. Neither the Turning machine nor the people running it will know what it's like to be the thing simulated. — noAxioms

How could this be compatible with physicalism? There is nothing physical about the simulated person outside of the Turing machine.

Simulations reproduce the informational, but not the physical, aspects of a system. If you are right, this implies that the 'what it's like' is informational, not physical (which I happen to believe).

But even if that were true, this still doesn't mean the simulation has qualia. It might be that only a certain type of informational processing manifest qualia. Ours have it, but the simulation lacks it. Would that imply qualia are epiphenomenal? Not necessarily. With the way our brains process information, qualia do work. The simulation might process in a completely different way that doesn't require qualia. The only requirement in a simulation is the informational behavior is reproduced. How it is reproduced is up to the simulation's designer.

I think you'd notice if your qualia suddenly vanished. — noAxioms

The zombie isn't aware, in the subjective sense, so it wouldn't notice anything.

Great, nice compromise.

Please don't remove them, @Jamal! I like that there is a permanent repository for the stories on the web. I might lose my copies someday, then they will be gone forever. I don't know if you still care, but I actually prefer that they are publicly available at this point. Easier to share with people.

That said, I'm personally pro euthanasia, and I do believe we should have the freedom to check out if that's the decision we come to. — Mijin

Euthanasia for the terminally Ill is one thing. For someone who is really depressed, or shaken by a loss that seems irrecoverable, that is quite another. I don't think it is ethical to make suicide a safe, available option for the depressed. If depression is a mental illness, then the person is out of their right mind, and does not have the competency to judge such a momentous decision for themselves.

My argument was that suicide harms everyone that valued the suicidal, including the suicidal's own future selves.

That suicide harms everyone who cared, sometimes devastatingly so, is obvious. What is less obvious is the future self argument, but I think a strong case can be made. In most contexts we treat future selves as moral agents, both unified with, and distinct from, present selves. If factory work causes you cancer that will kill you in 5 years, that is a terminal blow to your future self. Which is a blow your present self, as self-identity unifies past, present, and future selves. But I am quite different from my past selves, I have different beliefs, different motives, different goals, different abilities. Were I able to, I would bitterly resent a past self that killed me.

My rights as a present self are undisputable. But every present self is the future self of a past self. And so, if present selves have rights, future selves of present selves must have those same rights.

the branch of philosophy that deals with the first principles of things, including abstract concepts such as being, knowing, substance, cause, identity, time, and space.

So how can something be a "first principal"? Do you agree with google or not? — ProtagoranSocratist

First principle, in that these are concepts that are not simply compounds of other concepts. Diamond is a hard sparkly carbon substance, but substance itself, is just substance. We might look up diamond in a dictionary if we were unfamiliar with them, but we cannot look up substance. The definition only tries to codify our pre-existing intuitions of what substance is. If we lacked that intuition somehow, the definition would be meaningless to us.

This is what I take 'first principal' to mean. Not something that is necessarily ontologically basic. But something that is conceptually basic, the mental building blocks from which we build more complex conceptual structures, such as "wedding ring". Trivial seeming, but an intricate compound of the concepts 'marriage', 'diamond', 'ring', 'wealth', 'commitment', etc. And each of these are themselves compound. Because it is compound, discussion of "wedding ring" is not metaphysical, it is definitional, practical, cultural. Whereas, if you break these concepts down, you hit a kind of bedrock, where you find concepts like being, knowing, substance, cause, identity, time, and space.

These concepts cannot be broken down definitionally. They can only be philosophized, by creatively, artfully constructing a definition, which involves creating a deeper conceptual space into which these seemingly primordial concepts are placed. This is metaphysics.

And how can that happen just in neurobiological terms? Where is the neuroantomy? How is the human brain different from a chimp or even a Neanderthal? — apokrisis

Mainly different in it's language ability. Which allows it to think of a pink elephant, think about thinking about a pink elephant, and (sometimes) reliably report, "I am thinking of a pink elephant".

To introspect, as I conceive it, is not to think, feel, and experience, but to consider and potentially report the answers to the meta questions: "what am I thinking? What am I experiencing? What am I feeling?"

Therefore, you agree with the points of Epicurus and other philosophers who stated that pleasure is subjective. Since something (like opera, for instance) may be considered pleasure/non-pleasure at the same time by different perceivers, then music is dependent upon subjectiveness. — javi2541997

I think your wording threw me a bit. What brings one pleasure is subjective. Opera may be considered pleasurable or unpleasurable. I think most will agree with this, today. I tried to answer the slightly odd question, "Is pleasure [itself] subjective?", anyway.

And furthermore, are there insufferable experiences which are good? An appointment with the dentist, perhaps? — javi2541997

I fully agree that pleasure/pain and good/bad are independent axes. Whether good/bad is used in the advantageous sense, or the moral sense.

Where good/bad means advantageous/disadvantageous:

Good Pleasure:Success, hiking, social bonding
Bad Pleasure: Cigarettes, overeating, compulsive browsing/video games/etc
Good Pain: Dentists, surgery, workouts, study
Bad Pain: Illness, injury, depression

(TPF can occupy each of these!)

Where good/bad means moral/immoral:

Good Pleasure:Helping, reconciliation, activism, child rearing
Bad Pleasure: Sadism, exploitation, bullying, destruction
Good Pain: Self sacrifice, activism, child rearing
Bad Pain: Bitter arguing, war

What I consider a good pleasure, such as listening to opera, may be insufferable to you. According to this, pleasure seems to be a purely subjective concept. — javi2541997

Not necessarily. Opera is not itself pleasure, it is something that brings pleasure to you. If it is insufferable to me, it brings me no pleasure. The stimulus is not the response. Different stimuli may be needed to bring about the same pleasurable response in each of us.

What is and isn't pleasurable is subjective. But is pleasure itself subjective? On the one hand, pleasure, like other feelings, is a private sensation which can be experienced only by the one who feels it. On the other hand, pleasure universally attracts us to that which is pleasurable. Pleasure is a manifestation, made to a mind, of the body's instinct to do this thing, to seek this or that out. It is the carrot to pain's whip, and both work together to steer all the sentient animals. And so pleasure is an objective feature of the biology of everything with a mind.

Yes, I indeed think of introspection, or the idea of reflecting on the content and nature of our own mental states, on the model of self-analysis rather more than on the model or perception, as if we had an extra sense that turns inwards which I take to be a Cartesian confusion. — Pierre-Normand

I think we can indeed report our thoughts and feelings, as opposed to self-analyze. But of course we don't have a sense that turns inward.

We can report what we were thinking (in the sense of subvocalized words and images) only if the thought was salient enough to lodge itself in short term memory. If it has not, the thought is now inaccessible, and all we can do is try to reconstruct it based on context.

We can try to report what we are feeling, but it is tricky. There is the phenomenology of feeling, and there is its interpretive context. The same phenomenology might be interpreted differently depending on context (think excitement/anxiety). Then we have to choose the right conceptual bucket (aka word) to put this phenomenology/context into.

But what if introspection is a useful form of confabulation? Are you working with some science verified definition of introspection such that you could claim to make a genuine comparison between humans and LLMs? Or is the plausibility of both what humans say about themselves and what LLMs say about themselves the stiffest test that either must pass. — apokrisis

I'm not sure how science could verifiably define such a thing. I take it to mean here, the ability to reliably report inner state. How can introspection be confabulation, which is by definition an unreliable report?

In the paper, they test for introspection in a nifty way. "Concepts" are isolated by subtracting the inner state vector related to contextual understanding produced by two varying prompts. For instance, "HOW ARE YOU" - "how are you" represents "loudness". They then "inject" the concept back into the LLM. Then, they tell the LLM what they did, and ask it if it notices anything unusual. In the best model, Opus, it reports the correct concept about 20% of the time.

Don't you think a novelist who wrote their memoir would know much more about introspection than a cognitive scientist or a neuroscientist think they do? — Pierre-Normand

In the everyday/literary sense I understand it to mean something more like self-analysis, which is another thing entirely (contemplation/reasoning with the self as the object of scrutiny).

I could have read that paper carefully and made my own "chain of reasoning" response as is socially required – especially here on a "philosophy" forum trying to teach us to be more rational in a "present your full workings out" way.

But it was so much easier to back up my own gut response to just the quick description of the paper – where I dismissed it as likely yet again the same category error — apokrisis

You could have read the paper in the time it took you to write all that! Though to be fair you do seem to write quickly.

It really isn't the same category error. It describes some "LLM brain science" which does seem to demonstrate that LLMs are capable of introspection. Quite fascinating, imo.

There’s a research idea. Train an LLM on all available medieval texts and recreate the clever person of the 1400s. Have a conversation with your distant ancestor. — apokrisis

I do like this idea. The training set of top models is ridiculously vast though, including texts from even minor languages. They might already incorporate them.

You'd have to talk to the software developers to learn that. But right now I would expect that there is a lot of trade secrets which would not be readily revealed. — Metaphysician Undercover

The problem is, beyond the design of the llm "machinery" itself, they don't really know how it works either. LLM are in large respect black boxes, and a lot of effort is being put into figuring out what is actually going on.

But this is about it's ability to accurately introspect into it's own thought process (definitely check out the article I posted if you haven't yet). This is subject to confabulation. Or, to a kind of reenactment of the original thought process, but in an explanatory 'mode'.

But this doesn't give insight into what underlying method it actually uses to reason.

But aren't they just providing a reasonable confabulation of what a reasoning process might look like, based on their vast training data? — apokrisis

Maybe. But some kind of reasoning process must be at work, whether or not it's the human like chain of reasoning they offer as explanation of their thought process. Otherwise it is just not practical to simulate reasoning statistically. Imagine trying to do this even with simple math problems, the combinatorial explosion of possible inputs completely overwhelms mere statistics.

LLM research shows that that chains of reasoning aren't used to get to answers. They are just acceptable confabulations of what a chain of reasoning would look like. — apokrisis

My understanding of how the "reasoning" modes work is that they use a specially tuned model to produce text that represents what reflection on the users input might look like. Then so on, on the users text plus all the previous reasoning steps, until it is determined (somehow) that reasoning has proceeded far enough. Then the entire corpus of query plus intermediate texts produces the output.

But as for what happens in a single pass, I'm not sure even how much we understand at all about what is going on under the hood. How did research determine that chain of reasoning is not happening?

Notice the big difference though, human beings create the social norms, LLMs do not create the normative patterns they copy. — Metaphysician Undercover

We as individuals do not generally create social norms, we learn their rules and reproduce them, much as LLMs do. If there is creativity here, it is in the rare individual who is able to willfully move norms in a direction. But norms also shift in a more evolutionary way, without intentionality.

The LLM can imitate creativity but imitation is not creativity. — Metaphysician Undercover

Again, I would say that creativity is 95% imitation. We don't create art de novo, we learn genre rules and produce works adhering to them, perhaps even deviating a bit. Of course genre still affords a large scope for creativity. But, I'm not sure how you could argue that what LLMs produce is somehow uncreative, it also learns genre and produces works accordingly.

Consider the common question, "what are you thinking?". Or worse (for me), "What are you feeling"? To answer the question accurately does seem to require introspectively recalling and verbalizing your own cognitive or affective state. It is by no means a given that we are always able to do this. When I cannot and I know it, I will sometimes admit perplexity, and sometimes simply make up an answer. And I am certain there are times I can't but I don't know it, and I will confabulate something. Yet, the asker cannot reliably discern between these possibilities.

This is quite different from:

When you ask me to explain my reasoning, those same "voices"—the patterns encoding understanding of Husserl, Gibson, perception, affordances—speak again, now in explanatory mode rather than generative mode. — Pierre-Normand

Something like this undoubtedly happens when you ask a LLM, or a human, to explain its reasoning. But asking someone to "explain their reasoning" is not necessarily asking for introspection in the way that "What are you thinking/feeling" definitely does.

You can't just ask a LLM "What are you thinking today", obviously it will confabulate something. And if you could, you run into the same epistemic problem you have when you ask a human. Whereas, to ask it to explain its reasoning is not even a true introspective query. And so to demonstrate introspection in LLMs I think you have to do something like Anthopic did. By directly manipulating Claude's brain state, there is one right answer, and you know what it is

Similarly,

It's rather like claiming you've proven someone has introspective access to their neurochemistry because when you inject adrenaline into their bloodstream, they notice feeling jumpy and can report on it. — Pierre-Normand

Doesn't this indeed prove introspective access? Not exactly to neurochemistry per se, but to the affective states which correspond to it?

(BTW, IMO you thread the needle nicely in your use of AI on the site.)

Here's the TL;DR that you seem to require — Leontiskos

Not that I want to continue the discussion, but there was a good chunk of your reply I missed in my irritation, so apologies for that.

GTFO with your horseshit.

there are lots of LGBT individuals who agree with Bob, and who would find many who oppose him within this thread to be, "implying they are bad, immoral, and crazy." — Leontiskos

WTF. Bob has literally, explicitly, called multiple subsets of people bad, immoral, and/or crazy. This is quite different from simply opposing someone's wacky beliefs.

I literally gave you an example of bigotry. If you don't know by now that I think bigotry involves a mode of belief and not a material proposition, then you haven't read anything I wrote. — Leontiskos

Your misunderstanding of what bigotry is does not constitute an example. Again, you believe that any proposition, however odious and hateful it might seem on the surface, is not in itself bigoted. Such a proposition must be uttered by someone who we know is affectively obstinate, and we know in advance will never change their mind about it. Until we can somehow know that, we can never know if it is truly bigoted.

What is needed is a particular mode of belief, such as obstinacy (for example). — Leontiskos

So if I obstinately believe that the earth is round, that is bigotry by your reckoning?

I'll take that as a "yes," which contradicts what you just said. You say no one is personally attacking Bob and then you continue to personally attack Bob. That's the sort of gaslighting that Bob has been dealing with throughout, and it's not odd that he would defend himself. — Leontiskos

Nope, not a personal attack, except perhaps against his judgement. He might be doing this unwittingly, with the best intentions. But he is doing it regardless.

I've pointed out your error from the start, wherein you fail to understand that bigotry is a mode of behavior or belief, not an intrinsic quality of a proposition. — Leontiskos

And so your answer is "no". To you, no proposition or discourse can themselves be bigoted. And so if they are not, why should believing or promoting these propositions, no matter how obstinately, be bigoted either? Do you see how absurd this is?

Again, you have called KKK grand wizards "unbigoted", because the best ambassador to bigots we have ever seen were turned by him. And so whatever authority you may have had as to what bigotry is and isn't has already been ceded.

Er, but that has been a huge part of this thread, namely personal attacks and accusations on Bob. You yourself are arguing that someone who says what Bob is saying is bigoted, are you not? — Leontiskos

Bob is not only participating in, amplifying, and offering legitimatization of a larger homophobic and especially transphobic movement in this historical moment, especially in this country. But he has implicitly insulted forum members and their loved ones, implying they are bad, immoral, and crazy. So neither Bob or yourself are in any position to pearl clutch if he has received personal attacks in return.

When you take that pedantic route and erect curious and undefined terms like "definitional" and "substantive" you should expect similarly pedantic responses. — Leontiskos

Give me a fucking break. To attack these as curious and undefined is itself pedantic, fittingly as you are one of the most pedantic posters on here. If one were to take your pedantry seriously, a bigoted claim would simply be impossible.

Let's play a game. Make a claim that you believe is actually bigoted, if you think any exist.