But I am only convinced of this fact because I have experienced it independently from any other agent. I know for sure I exist in some sense, I believe very strongly that the physical world exists, and I believe to a lesser degree that other agents have similar experiences to mine.

In other words, I am more certain of the fact that your comment exists than I am of the fact that your subjective experience exists. If I found out the comment I am responding to was in fact GPT-3 output, I would be much less shocked than if I found out my own senses or memory deceived me.

So, the physical world is a much more believable explanation than a world of conscious discussion; and exploring the physical world in the aspects we ourselves can observe about it alone is a much more convincing argument than trying to discuss it with other conscious agents.

It's much easier to convince someone else that I can pass through walls than it is for me to actually pass through a wall.

>In other words, I am more certain of the fact that your comment exists

>So, the physical world is a much more believable explanation than

Phrasing all of these in terms of what you believe and what you are convinced/certain of, instead of in terms of what absolutely is, makes my point for me. You didn't even realize you were talking in terms of you getting me to agree over what to believe rather than some objective nature of things indifferent to my opinion. That's how deep seated this is in the way brains and consciousness work.

>I would be much less shocked than if I found out my own senses or memory deceived me.

You even phrased the discussion of your own sensory perception in terms of messages. In this case messages from your own senses. And you personified them too, treating them as conscious enough to lie to you.

>I know for sure I exist in some sense... and I believe to a lesser degree that other agents have similar experiences to mine.

You're already agreeing with my axioms. I know I exist. And making an argument about what consciousness is wouldn't mean much if there wasn't a "you" that was convinced by it. So lets just take it as axioms that any compelling argument has to exist in words, and in any written thought there's "you", "me", and the words themselves. In a way you're being more restrictive than I am. I'm already acquiescing in the axioms that you exist as the same sort of consciousness entity as myself and others in the domain of conversation. If you're having this conversation you're a consciousness and that's all the definition of consciousness we need to make a convincing argument about consciousness.

All I'm saying is, do I need your steadfast belief in the underlying reality as axiom, or can we construct it from the parts we're already agreeing on?

> Starting with just the existence of the arguers and their arguments, can they argue for the existence of anything more? If not, what else do they need to assume?

No, you can't argue the world into being from these axioms. They could probably invent logic and mathematics, but nothing of the natural sciences can be discovered without the senses.

It seems contradictory to admit they could invent math and logic, but then reject that they could go one step further and rig that understanding of math into and understanding of physics. I'll admit without some corpus of data to understand the motivation for constructing physics is tenuous. But the question was about the principle of if they could and if you're granting me math I don't see any obstacle left.

>Phrasing is irrelevant. My senses are not little people talking to me about what they perceive

Sure maybe your senses aren't actually little people, but if you've already evolved a social reasoning / grunting system that only knows how to talk about people, why not convert that into a system for reasoning / grunting about everything by imagining everything as little people. It might not be literally true, but its a useful fiction. Its a fiction that lets us hack "conversational reality consciousness" into "physical reality consciousness". The difficulty in twisting the phrase differently is supportive of this hypothesis.

The problem with math is that math can describe any possible universe, and there is no way to choose until you confront it with the real world. Nothing in math prevents the world from having 1 dimension of time and 1 of space, for example. Nothing in math prevents the electron from being much larger than the proton, or the existence of solitary quarks or anything else.

So I would grant you that the world of talking agents could describe our physical world through math, but they could also imagine any other physical world, and they would have no way to choose one.

My point was that we can't find out anything about the phsycial world by just discussing it. We each have to experience it ourselves using our own senses. We can of course later discuss to devise new ways of understanding what the world is and how it works, but even then, we need to put any theories we come up with to the test to check if they actually hold up.

Superficially yes but actually no. This fictitious admin character doesn't have any actual powers, I'm simply choosing a particular "consciousness" to be a fixed and agreed upon meter stick of objectivity. All measurement systems are arbitrary. We could choose anyones POV to fix as the "objective" truth. If we both agree on the same "definer entity" then we can be in agreement about other things objectively defined relative to them. But crucially this is all still nothing more than us reaching agreement. All that really exists is still just you me and the words.

So how about it. Do you believe in Bob?

The joke of course is quantum mechanics also works even if you don't believe in it (or so they say). If someone rejects quantum physics and embraces magic horseshoes, will anything punish them for being objectively wrong? Is there anything you can do to force them to believe the objective truth? If the answer is no to both, then what makes quantum physics the objective truth and magic horseshoes utterly wrong? Maybe there is an objective truth, but all we can have is belief about which things are objective truths. I can be contrarian and say flat earthers are correct, and there's nothing you can do to force me to agree otherwise. So if nothing eventually forces agreement on that matter, what makes one side objectively true?

For the record, I do actually believe in reality. I'm just interested to see what happens when we turn the problem of consciousness on its head. Instead of assuming reality and understanding consciousness within it, assume consciousness and try to paint a picture of reality within it. This is all academic exercise.

You cite the primacy of sensation, but what more is sensation than a message from the sensory organ to the brain. Perhaps encoded in the brains internal language rather than the plain text we are used to, but messages none the less. The concept of existence-as-messages is thus not contradicted by the experience of messages from your own body.

Irrespective of any underlying physical reality, I can't escape the fact that I can never test reality itself, only my perception of it. We can get a lot of mileage out of the physicalist approach, accepting as an axiom that reality is real and we're just living in it. But in that ground up approach we've had terrible difficulty deriving our ourselves from physical first principles. I'm not saying its impossible or wrong to take the path from atoms on upwards to consciousness and perception. But no one has been able to make it to the end of that path. What happens if we go the other way, starting from the known conclusion "I'm here and conscious enough to converse" and working our way down to an inanimate reality governed by objective physical principles? Can a system only designed to represent/reason about people talking to and about people be hacked into a system for representing and reasoning in general?

If we start from purely language models like GPT3 and continue to teach it the "social reasoning" of saying things we want to hear, will the language model eventually become capable of non-social reasoning as well? In the process of figuring out what we want to hear well enough to describe a non-contradictory scene to us, does GPT3 have to actually learn the rules of 3d euclidean space governing the scene? Is there any possible way to avoid scene contradictions without a full understanding of the underlying physical reality it is supposed to describe to us?

This is false, it's a version of the homunculus fallacy. A sensory organ is something that connects the brain to the physical world. Even if you chose to model it as a an agent that sends messages to the brain, it is an agent of a different nature. The sensory agent doesn't receive messages from other agents, it receives raw input from the outside world (photons, electrical fields, chemical reactions etc).

In contrast, if you were a brain in a vat with no ability to directly perceive the world or interact with it in any way, it would be impossible for you to know that "I can pass through walls" is fundamentally impossible.

Even for your GPT-3 thought experiment - ultimately it is the effects of the physical world perceived directly by humans sense organs that shape what GPT-3 would utter. That is, even if it can learn what the world is like simply by talking to us, it's still learning about the real world from someone's direct experience with it. If we were all GPT-3s, with no cameras and pressure sensors and motors etc, we would be unable to reason in any sense about the world itself.

We could perhaps come to agree upon some imagined world, but that could change arbitrarily much from one day to the next on a whim.

Where do you draw the distinction between "messages" and "information" (raw or cooked). Information theory was contrived to model messages sent in a noisy channel, but it applies just as well to data streams that have no communicative intent or origin. Its a distinction without a difference. You may as well treat all information as messages in a channel, even if the sender is nature herself. Alternatively, you may as well treat all messages as just information, and view "senders" with "intent" as just another physical process in a world of physical computation. As the cliche saying goes, "information is physical".

>We could perhaps come to agree upon some imagined world, but that could change arbitrarily much from one day to the next on a whim.

Ok, this is a fine basis to work with. How about this. "Reality is the set of beliefs which, if you disagree with them too much and for too long, you are eventually and permanently removed from the conversation." For example, quite recently, large swaths of people held an exquisite referendum on the existence of covid. Needless to say, rather than covid disappearing on their whim, a great deal of them are now permanently no longer participants in this conversation.

Notice this isn't far off from my original postulate. "Whatever we agree upon is our reality, and we'll argue with/fight anyone who disagrees until they agree with us or us with them." Allowing for some personification I could phrase this scenario as "they disagreed with the virus and the virus won."

So perhaps I do need to add one thing to the postulates. One thing which remains objectively true even in the conversational model of reality.

"You can die. Dying means never being heard from in this conversation again."

You're right there's no mechanism for any choice of words to win over any other choice of words without an objective consequence to losing. The things that exist so far are "you", "me", "this" and "death". Not where I was expecting this to go but good point.