Really liked the post. Current ML tentatively seems like it could capture intelligence (LLM's) and agency/desires (reinforcement learning), but qualia remains a mystery. I feel like these 3 things, all of which are pretty fundamental to our intuitive understanding of sentience, could be logically independent of one another. It would be very interesting to study qualia using tools from math/physics.
Also, I read the description of your personal situation linked from your profile, and I'm sorry about what you're going through. Hope you can find a way to stay safe, and that everything goes well for you.
If your model doesn’t “logically exclude any other model”,
It doesnt have to be logical. You can exclude other universes and wotnot on the empirical basis that they are not observed.
The same logic may not apply to experience: if you want to explain subjective experience, you need to explain any possible subjective experience, any experience that could be constructed, observed or not.
Here's lc's opinion:
It is both absurd, and intolerably infuriating, just how many people on this forum think it's acceptable to claim they have figured out how qualia/consciousness works, and also not explain how one would go about making my laptop experience an emotion like 'nostalgia', or present their framework for enumerating the set of all possible qualitative experiences.
(Including the ones not experienced by humans naturally, and/or only accessible via narcotics, and/or involve senses humans do not have or have just happened not to be produced in the animal kingdom)
I think people severely underestimate the importance of studying subjective experience (qualia).
In my opinion, people think a lot about "the mystery of qualia" (the hard problem of consciousness), but not enough about "the miracle of qualia". I.e. why qualia is cool and interesting. It's not as trivial as it seems. Subjective experience is always with us, so it's hard to notice how strange it is (even without any "hard problems"). Down below is the list of reasons why qualia deserves way more attention than people give it. For convenience, I split the post into 4 parts.
Epistemic status: I haven't read a lot of relevant literature. Some claims are based on indirect evidence: we clearly don't live in a culture were subjective experience is valued as much as I think it should. Discussing this topic may be an exercise in prioritization.
Subjective experience (qualia) tells us that there are differences between objects which feel very simple, but lay beyond any current knowledge. That is to say we don't have any model of those differences. We don't know the (mathematical) space of our experiences.
Subjective experience is like a big carrot hanging right before your eyes. Telling "hey, I'm hiding some new knowledge you're completely missing out". If I were a mathematician or an AI researcher or an artist, I would be going insane trying to guess what could resemble the workings of subjective experience.
Some examples of differences which feel simple, but don't have an explanation/a model:
And much more. Any experience teases you about the knowledge you lack. Every second. And yet a lot of people don't react to this "teasing" or don't even feel it. Don't try to advance their understanding of experience.
Imagine your favorite piece of art or your favorite place. Your favorite story, fictional or real. Your favorite fictional character. Your memories about personalities of your friends. All of this is about qualia.
Anything that is meaningful or interesting in life is created by the subjective differences between bits of information. The difference between people's personalities, which is easy to perceive and impossible to express, is a particular example of this and another miracle of qualia.
What topic could be more fascinating and important than this? And yet nobody seems too interested.
Also, there's another connection between qualia and values: in ethics, the ability to have subjective experience often determines "moral worth".
There's a weird way to evaluate the importance of qualia. But it may serve as an intuition pump for some ideas.
Imagine a bunch supernatural claims being true: e.g. the God exists, immortal souls exist, magic exists. The topic of subjective experience would still be at least equally important as all those claims combined. In a way, the topic of qualia would become equivalent to all those claims: the experience of all beings = the souls of all beings (if souls existed) = every experience that God can create (if God existed). This equivalence is sometimes reflected in philosophy and religion, for example in Gottfried Leibniz's Monadology, George Berkeley's subjective idealism and Buddhism.
Qualia is as important as anything that can possibly exist. This is a fascinating property. Qualia is stranger than any fiction.
Imagine if there was a secret of creating any possible universe, encoded in your brain. This is what qualia is. It creates the universe of your experience. I don't think people are interested enough in qualia.
...
No, really, imagine that tomorrow you learn that God exists and immortal souls exist too. What would your emotions be?
If your emotions about qualia are weaker, then you're irrational. (If my arguments are correct.)
Qualia signals missed knowledge, but it also signals missed ideas. It's hard to even think about any model of qualia.
Imagine you say "the qualia of tastes can be modeled by a three dimensional vector space". The immediate questions follow: can qualia exist in a four dimensional vector space? can qualia exist in something that's not a vector space? If the answer is "yes", then you still need to explain qualia, you've only scratched the surface of the phenomenon.
To give a model of qualia you need to explain why qualia can't logically be described by any other model.
This is a stronger requirement for an "explanation" than in Physics. Physics can be considered solved if someone comes up with a model of the Universe which predicts everything correctly. Other possible Universes can be ignored. However, if you want to explain qualia, then you can't ignore any hypothetical subjective experience. You may call it "the hard problem of modeling qualia".
"The hard problem of modeling qualia" may serve as an intuition pump for why it's so easy to end up with panpsychism while trying to solve the hard problem of consciousness. If you provide a model of human consciousness without making all other models logically impossible, then you end up with conscious rocks.
A similar thing happens when you model the Universe. If your model doesn't "logically exclude any other model", then you can't be sure if your calculator isn't one of the possible universes. This idea is known as the mathematical universe hypothesis or Tegmark's Level 4 multiverse: "any mathematical structure corresponds to a universe". It is Physics' analogy of panapsychism.
You can draw this parallel:
Intelligence and qualia are two sides of the same coin (for humans). Both intelligence and experience can expand and develop by absorbing information.
A priori, the urge to explain experience should be at least equal to the scientific urge. But some factors make qualia even more important than intelligence (for humans):
Science is a world-wide movement, an institution. Is there at least a single small movement about explaining subjective experience? No. We miss at least a half of our culture.
The basis of our society is knowledge related to intelligence. "Knowledge is power".
90% of your experience is qualia, not intelligence.
90% of your experience doesn't matter to other people.
You may suffer your entire life and it's gonna matter less than a couple of mathematical theorems. Because knowledge is power. And your suffering is not "knowledge" enough. Most of the things you care about are not "knowledge" enough.
To establish qualia as a source of knowledge means a chance to change this fact, to change the basis of our society. The scientific revolution also meant a revolution in power. And there is a chance for another revolution. The chance may be very small, but it doesn't matter given how important the topic is anyway. Those are the stakes of studying qualia.
Any knowledge, if it's meta enough, becomes "meta knowledge about qualia".
For example, let's analyze Physics in the style of Rene Descart (Cogito, ergo sum):
Level 4 turns "philosophy of physics and math" into "philosophy of experience". Even if we can be deluded about 4, this knowledge is going to die the last.
So, the knowledge about qualia is the most fundamental type of knowledge. (More fundamental than physics and math.) It is another reason why experience is more important than intelligence for humans.
Some theories of why humans have moral values:
I think those theories are true on their respective levels of reality. They are aspects of the truth. But there's another very big aspect of the truth:
Loving a sentient being feels fundamentally different from eating a sandwich. Those are objectively different experiences, even if values are subjective. This difference between experiences may give a strong bias towards valuing love more than a sandwich.
Moral sentimentalism is the closest idea to this I know of. But it goes somewhere in the wrong direction: instead of analyzing experiences and their connection to morality it talks about just a single type of experience, the "feeling of wrongness".
For moral sentimentalists, our emotions and desires play a leading role in the anatomy of morality. Some believe moral thoughts are fundamentally sentimental, others that moral facts are related to our sentimental responses, or that emotions are the primary source of moral knowledge.
If you want to know how experience may connect to morality, why do you focus on a single type of experience? Why do you talk about "emotions" instead of experiences in general?
I think we should be precommitted to trying to model qualia. Even if we don't know how to approach it, we should try. The same way people become precommitted to trying to solve certain problems. The same way we are precommitted to trying to solve Alignment. Even if the solution may lay beyond the scope of our current possibilities and knowledge.
Moreover, there is a chance that studying qualia will help to solve Alignment. I think the chance is not that small, given that (1) qualia is core to "value" and "meaning" (both things are key for Alignment) and (2) we should study qualia anyway.
Some people do study experience, of course:
Neural correlates of consciousness
The neural correlates of consciousness (NCC) constitute the minimal set of neuronal events and mechanisms sufficient for a specific conscious percept. Neuroscientists use empirical approaches to discover neural correlates of subjective phenomena; that is, neural changes which necessarily and regularly correlate with a specific experience.
But all of my points still stand. Neuroscientists shouldn't be the only people who think about experience. And it seems that so far neuroscientists don't have enough knowledge to make promising/useful enough models of subjective experience. Otherwise their discoveries would already have massive implications for AI research.
Qualia isn't necessary for an AI. Still, the topic of AI is indirectly related to qualia. AI research studies effective ways to operate with concepts and gives indirect evidence about the way humans may operate with concepts.
However, it seems right now AI research doesn't give a lot of ideas about qualia and human concepts:
In philosophy, there's supposed to be an entire field about studying qualia, called "phenomenology".
The discipline of phenomenology may be defined initially as the study of structures of experience, or consciousness. Literally, phenomenology is the study of “phenomena”: appearances of things, or things as they appear in our experience, or the ways we experience things, thus the meanings things have in our experience. Phenomenology studies conscious experience as experienced from the subjective or first person point of view.
Sounds good, but:
I haven't read enough, so don't take my word for it. I don't know how much of this is true. But from my point of view, the cultural impact of phenomenology amounted to zero anyway: people don't care about qualia enough.
The Fun Theory Sequence by Eliezer Yudkowsky
Fun Theory is the field of knowledge that deals in questions such as "How much fun is there in the universe?", "Will we ever run out of fun?", "Are we having fun yet?" and "Could we be having more fun?"
I think the Fun Theory suffers from considering subjective experience only in the context of (infinite) fun/boredom. Qualia is interesting enough by itself.