Symmetry Theory of Valence “Explain Like I’m 5” edition

Hello! I have some thoughts about qualia and the binding problem. I don’t know where should I put them, so I will leave it here. I hope it will help somehow.

In my oppinion the problem of qualia generally consists of two equally important issues:
1) The quality problem – what really makes any two experiences different (like blue and red or good and bad or even pink and C note)?
2) The binding problem – how is it possible that we experience multiple sensations at the same time?

The quality problem is hard, but it seems to be solvable. At least it doesn’t reqiere us to change radically our current understanding of the phisical world. We already have neural networks that can roughly visualise a picture what a given person sees using only the data about its brain activity ( https://singularityhub.com/2018/01/14/this-neural-network-built-by-japanese-researchers-can-read-minds/#sm.001518jxz10eaf1lq272rd4fe30hc ). So now we can look at what actually those neural networks have learned, which patterns did they extract from the data, and it will allow us to link qualia to certain behaviour of the neurons.
It’s harder then to extrapolate correctly those principles to non-human systems. It is impossible to test for example wether a rock has any qualia. However when a theory is ready, we could build a non-biological brain implant to add some new feeling (a new color, for example). Then we could add it to a volunteer and ask him wether he feels that new qualia. If he does, it would be a good evidence that feeling doesn’t require a biological substrat.
However we need to be careful here to distinguish wether a brain feels by that implant or just takes some information from that implant and learns to feel it on his own. There already is a real cyborg who managed to extend his qualia range. Neil Harbisson was born with an extreme form of colour blindness and because he wanted to correct that, he mounted an antenne into his head to gain a new feeling of colours, ( https://en.m.wikipedia.org/wiki/Neil_Harbisson ). In Neil’s case the implant probably don’t generate a feeling itself – it just passes information to Neil’s brain. The same way we can suppose that our arms and legs doesn’t feel pain themselves. They probably just pass the information about pain to a brain, where the feeling is being born. To distinguish situations of information transmission and feeling generation we would have to look at what time does it require for new feeling to arise. I suppose that if brain just gets new information, it requires some time to learn how to interpret that. But if an implant generates a feeling, there should be some immideate effect.
Of course we still need some mathematical models of qualia. For example we need to explain why RGB input of an eye gives us such distinct colours like yellow, magenta, cyan, black, white. Why don’t I perceive yellow as a greenish red or redish green? It looks completely different! Maybe it’s a deal of cultural influennce, I don’t know. English speakers tend to mix cyan and blue into one category. It’s strange, but despite the fact that I am a native Russian speaker and we have clear distinction between cyan and blue, I kinda can agree that cyan feels a little bit bluish and doesn’t feel greenish at all. However on a color wheel cyan feels completely distinct. Maybe that all happens the way it happens because a lot of things like sky and water tend to be somewhere between cyan and blue, and that experience influences our associations between colours.
However building models of qualitive distinctions between qualia doesn’t seem to be something unrealistic. Realtions inside one category seem to be at least intuitively understandable. Yes, it is timeconsuming, it requires a lot of introspection and some smart euristics. To understand realtions between different categories is of course harder. But at least we can think of them using programming analogy: feelings are data and categories are data types. And data types are just some data about data. Or maybe categories are some specific data structures. At least we can try to think of them in that way. We have some terms, some basic notions to think of them.
The binding problem feels to be much harder exactly because we don’t even know how to think about it. For centuries reductionist view on the universe has been prooving its consistency. You don’t need to believe that an apple is something whole to understand how it falls. You can consider it a bunch of atoms where each of them interacts with others completely independantly, and you will get the same result (or even a more precise one). Our experience is the first thing where we have to believe into wholyness. Why should different neurons representing different pixels of my experience contribute to a single observer? Why can’t each of them have it’s own independent one-pixel experience? Even if we imagine that all the informotion is somehow transmitted to a single neuron in my brain, why shouldn’t its separate parts have their own one-pixel experiences? What gives a rise to a single picture that I see? Even the idea of a single neuron would’t solve the problem (moreover it isn’t supported by the data that we have from neurology).
It feels necessary to invent some glue that would link those one-pixel experiences into something whole. And it’s really hard not to fall into any kind of misticism here.
IIT gives us an ingenious idea to consider causal relations to be such glue. And it seems indeed promising since causality really exists, otherwise physics wouldn’t work. The problem that I see in IIT is that it relies on such notions as computing element, logical state, moment. And those are abstractions. Maybe I am ignorant and dont’t understand ITT well enough, then I will appreciate if you explain me where I am wrong. But let me explain my point of view.
Any logical element (neuron or logical gate) requires some time to react. Let it be 100 ms for simplicity. If we consider it to be the smallest step, then all works fine: each neuron’s state most of the time depends on its neighbors states’ on previous step. So there is constantly some amount of consciousness. But if we take a smaller step, lets say 1 ms, then for 99 steps each neuron’s state depends only on its own previous state. So for the most part of the time there is no conciousness. Then we need some explanation about what produces the feeling that our experience is constinuous. Moreover if consciousness appears only for a single moment for each 100 ms, then a slightest asynchronisation between two subsystems will damage consciousness dramatically, because those asynchronised subsystems will contribute to consciousness in different time frames.
Lets go further. Lets suggest that a neuron is a simplest element in a system. Then IIT works great. Lets now suppose then a neuron consists of dendrites and an axon. Maybe I am wrong, but it seems to me, that its own dendrites will define its next state much better than any of its neighboring neurons, then according to the exclusion principle all the neighboring neurons should be excluded, shouldn’t they?
Don’t get me wrong. I consider IIT a great idea. But it still rises a lot of questions. Maybe those questions above are some logical falacies that I’m falling into, but I have found one simple phenomena that make me doubt.
Giulio Tononi claims that according to IIT a simulation of conciousness won’t be conscious the same way, as simulation of fire won’t be a real fire. It made me think a lot wether there is some way to check that out. Conciousness, is it a state of hardware or a state of software? Is it a phisical state or is it some kind of pure information? How could we test that? And I came to an idea of a test that is as simple as a finger in front of your nose. Look, our inner image of the external world is being combined from the information that comes from two eyes. And the most part of it is doubled, because eyes are looking at the same direction. So if they were looking at different directions, they would get more unique information in total. And it would require larger area of subjective image. It’s hard to make your eyes look in opposite directions, however it’s easy to do the opposite: put your finger in front you and watch it bringing it closer to your nose. And you can notice that the world that you see is becoming smaller. But the most important is that you don’t feel anything beyong its border. Not blackness nor anything less. It means that the window of your consciousness literaly becomes smaller at this moment.
What does it mean after all? Imagine that there is some grid on which we transmit some information, and that grid feels it. If consciousness is a state of the grid, than each part of it should feel at least something even when it doesn’t receive any signal. The same way when we close our eyes, we do not perceive nothing. We still perceive something. We feel blackness. When we sleep we don’t feel even that. My point is that consciousness isn’t a state of the grid. Otherwise it couldn’t become smaller or larger. Consciousness is the informaition contained in the grid. And since there can be less information, it can become smaller or larger. So I claim not only that simulation can be conscious, but that simulation is the only thing that can be conscious.
I can’t say that my argument contradicts IIT. Actually IIT allows some phisical parts to be excluded or included depending on their state, so it just requiers some specific logic to exclude exactly those elements, that are not percieving information from the eye. However I find one more pleasant thing in the point of view that consciousness is information. You see, information isn’t something purely physical. However it is something that seems to exist. Maybe that is the root of mind-body dualism. It is possible to make some sorts of things with information that are impossible with physical objects. You can combine information together. Imagine that there is a bulb that turns green when a person standing in front of it is an old lady with a blue bag and turns red if it is somebode else. It’s just one bit of information, however if you see that the bulb is green, you will know 3 distinct facts about the last person that stood in front of it: it was a lady, she was old and she had a blue bag. Information may bind other information together. If you know that a sheet of paper can serve as a container for information, that knowledge will make you percieve its content as something whole and colinked. This property allows information to be distributed in space and time.
The problem with the notion of information is that it is abstract and depends on a subject. Does a book contain any information? Well, it seems to depend on whether there exists somebody who is able to read it. So it’s not the kind of information that would produce consciousness. In order to produce an observer our notion of information must itself have some objective nature.
Probably causality is indeed a good basis to define such natural information. However I see no good reasons why should we look at causality only in two neighboring moments of time. We could consider a content of a book to be information that is frozen in time until it produce some causal effect. So a book has consciousness, but that consciousness doesn’t experience time. But it becomes a part of our consciousness when we read the book (so some causal interaction happens).
OK, but we still haven’t solved the problem of binding. We can bind information for computers or for humans because there already is some actor (computer or human) that “knows” how to read that information. A human knows that he has to read the first page first and the second page second. Computer “knows” that it has to read all the commands one by one and to change the order only if it has readen a kind of “goto” command. And exactly those preexisting rules of reading allow us manipulate readers behavior to bind some information into some large entities. For example if we want to allow a person to move immideatly to an interesting part of our book and we know that he will start reading from the first page, we can put here a table of contents. That’s how we bind some pieces of information into larger objects – chapters. We know where a reader will start reading and we put here some information, that describes (influences thème process of reading of) next information.
If we speak about natural information, then the reader is probably the universe itself. And where does it start reading? Well, probably everywhere at once. It may sound confusing, but at least we know the order of reading: things that are closer to each other will infliunce each other faster than things at a larger distance. We know for sure that information can’t travel faster than the speed of light. Now if there is something that can manage the propagation of causal influence, make it faster or slower, change its direction, then it is a mechanism that allows to bind information into some larger entities. And that “something” are probably the laws of physics.
I can’t say that my thoughts should change something dramatically in the IIT or in the theory of valence present abow. But at least I have tried to look at the problem at some more fundamental level to make some gueses about how the reductionist world which we live in could lead to something so irreduceble as consciousness.

2 thoughts on “Symmetry Theory of Valence “Explain Like I’m 5” edition”

Pingback: The ecosystem of unifying ontologies | Opentheory.net
Count of Darkness says:

January 8, 2019 at 10:41 am

Hello! I have some thoughts about qualia and the binding problem. I don’t know where should I put them, so I will leave it here. I hope it will help somehow.

In my oppinion the problem of qualia generally consists of two equally important issues:
1) The quality problem – what really makes any two experiences different (like blue and red or good and bad or even pink and C note)?
2) The binding problem – how is it possible that we experience multiple sensations at the same time?

The quality problem is hard, but it seems to be solvable. At least it doesn’t reqiere us to change radically our current understanding of the phisical world. We already have neural networks that can roughly visualise a picture what a given person sees using only the data about its brain activity ( https://singularityhub.com/2018/01/14/this-neural-network-built-by-japanese-researchers-can-read-minds/#sm.001518jxz10eaf1lq272rd4fe30hc ). So now we can look at what actually those neural networks have learned, which patterns did they extract from the data, and it will allow us to link qualia to certain behaviour of the neurons.
It’s harder then to extrapolate correctly those principles to non-human systems. It is impossible to test for example wether a rock has any qualia. However when a theory is ready, we could build a non-biological brain implant to add some new feeling (a new color, for example). Then we could add it to a volunteer and ask him wether he feels that new qualia. If he does, it would be a good evidence that feeling doesn’t require a biological substrat.
However we need to be careful here to distinguish wether a brain feels by that implant or just takes some information from that implant and learns to feel it on his own. There already is a real cyborg who managed to extend his qualia range. Neil Harbisson was born with an extreme form of colour blindness and because he wanted to correct that, he mounted an antenne into his head to gain a new feeling of colours, ( https://en.m.wikipedia.org/wiki/Neil_Harbisson ). In Neil’s case the implant probably don’t generate a feeling itself – it just passes information to Neil’s brain. The same way we can suppose that our arms and legs doesn’t feel pain themselves. They probably just pass the information about pain to a brain, where the feeling is being born. To distinguish situations of information transmission and feeling generation we would have to look at what time does it require for new feeling to arise. I suppose that if brain just gets new information, it requires some time to learn how to interpret that. But if an implant generates a feeling, there should be some immideate effect.
Of course we still need some mathematical models of qualia. For example we need to explain why RGB input of an eye gives us such distinct colours like yellow, magenta, cyan, black, white. Why don’t I perceive yellow as a greenish red or redish green? It looks completely different! Maybe it’s a deal of cultural influennce, I don’t know. English speakers tend to mix cyan and blue into one category. It’s strange, but despite the fact that I am a native Russian speaker and we have clear distinction between cyan and blue, I kinda can agree that cyan feels a little bit bluish and doesn’t feel greenish at all. However on a color wheel cyan feels completely distinct. Maybe that all happens the way it happens because a lot of things like sky and water tend to be somewhere between cyan and blue, and that experience influences our associations between colours.
However building models of qualitive distinctions between qualia doesn’t seem to be something unrealistic. Realtions inside one category seem to be at least intuitively understandable. Yes, it is timeconsuming, it requires a lot of introspection and some smart euristics. To understand realtions between different categories is of course harder. But at least we can think of them using programming analogy: feelings are data and categories are data types. And data types are just some data about data. Or maybe categories are some specific data structures. At least we can try to think of them in that way. We have some terms, some basic notions to think of them.
The binding problem feels to be much harder exactly because we don’t even know how to think about it. For centuries reductionist view on the universe has been prooving its consistency. You don’t need to believe that an apple is something whole to understand how it falls. You can consider it a bunch of atoms where each of them interacts with others completely independantly, and you will get the same result (or even a more precise one). Our experience is the first thing where we have to believe into wholyness. Why should different neurons representing different pixels of my experience contribute to a single observer? Why can’t each of them have it’s own independent one-pixel experience? Even if we imagine that all the informotion is somehow transmitted to a single neuron in my brain, why shouldn’t its separate parts have their own one-pixel experiences? What gives a rise to a single picture that I see? Even the idea of a single neuron would’t solve the problem (moreover it isn’t supported by the data that we have from neurology).
It feels necessary to invent some glue that would link those one-pixel experiences into something whole. And it’s really hard not to fall into any kind of misticism here.
IIT gives us an ingenious idea to consider causal relations to be such glue. And it seems indeed promising since causality really exists, otherwise physics wouldn’t work. The problem that I see in IIT is that it relies on such notions as computing element, logical state, moment. And those are abstractions. Maybe I am ignorant and dont’t understand ITT well enough, then I will appreciate if you explain me where I am wrong. But let me explain my point of view.
Any logical element (neuron or logical gate) requires some time to react. Let it be 100 ms for simplicity. If we consider it to be the smallest step, then all works fine: each neuron’s state most of the time depends on its neighbors states’ on previous step. So there is constantly some amount of consciousness. But if we take a smaller step, lets say 1 ms, then for 99 steps each neuron’s state depends only on its own previous state. So for the most part of the time there is no conciousness. Then we need some explanation about what produces the feeling that our experience is constinuous. Moreover if consciousness appears only for a single moment for each 100 ms, then a slightest asynchronisation between two subsystems will damage consciousness dramatically, because those asynchronised subsystems will contribute to consciousness in different time frames.
Lets go further. Lets suggest that a neuron is a simplest element in a system. Then IIT works great. Lets now suppose then a neuron consists of dendrites and an axon. Maybe I am wrong, but it seems to me, that its own dendrites will define its next state much better than any of its neighboring neurons, then according to the exclusion principle all the neighboring neurons should be excluded, shouldn’t they?
Don’t get me wrong. I consider IIT a great idea. But it still rises a lot of questions. Maybe those questions above are some logical falacies that I’m falling into, but I have found one simple phenomena that make me doubt.
Giulio Tononi claims that according to IIT a simulation of conciousness won’t be conscious the same way, as simulation of fire won’t be a real fire. It made me think a lot wether there is some way to check that out. Conciousness, is it a state of hardware or a state of software? Is it a phisical state or is it some kind of pure information? How could we test that? And I came to an idea of a test that is as simple as a finger in front of your nose. Look, our inner image of the external world is being combined from the information that comes from two eyes. And the most part of it is doubled, because eyes are looking at the same direction. So if they were looking at different directions, they would get more unique information in total. And it would require larger area of subjective image. It’s hard to make your eyes look in opposite directions, however it’s easy to do the opposite: put your finger in front you and watch it bringing it closer to your nose. And you can notice that the world that you see is becoming smaller. But the most important is that you don’t feel anything beyong its border. Not blackness nor anything less. It means that the window of your consciousness literaly becomes smaller at this moment.
What does it mean after all? Imagine that there is some grid on which we transmit some information, and that grid feels it. If consciousness is a state of the grid, than each part of it should feel at least something even when it doesn’t receive any signal. The same way when we close our eyes, we do not perceive nothing. We still perceive something. We feel blackness. When we sleep we don’t feel even that. My point is that consciousness isn’t a state of the grid. Otherwise it couldn’t become smaller or larger. Consciousness is the informaition contained in the grid. And since there can be less information, it can become smaller or larger. So I claim not only that simulation can be conscious, but that simulation is the only thing that can be conscious.
I can’t say that my argument contradicts IIT. Actually IIT allows some phisical parts to be excluded or included depending on their state, so it just requiers some specific logic to exclude exactly those elements, that are not percieving information from the eye. However I find one more pleasant thing in the point of view that consciousness is information. You see, information isn’t something purely physical. However it is something that seems to exist. Maybe that is the root of mind-body dualism. It is possible to make some sorts of things with information that are impossible with physical objects. You can combine information together. Imagine that there is a bulb that turns green when a person standing in front of it is an old lady with a blue bag and turns red if it is somebode else. It’s just one bit of information, however if you see that the bulb is green, you will know 3 distinct facts about the last person that stood in front of it: it was a lady, she was old and she had a blue bag. Information may bind other information together. If you know that a sheet of paper can serve as a container for information, that knowledge will make you percieve its content as something whole and colinked. This property allows information to be distributed in space and time.
The problem with the notion of information is that it is abstract and depends on a subject. Does a book contain any information? Well, it seems to depend on whether there exists somebody who is able to read it. So it’s not the kind of information that would produce consciousness. In order to produce an observer our notion of information must itself have some objective nature.
Probably causality is indeed a good basis to define such natural information. However I see no good reasons why should we look at causality only in two neighboring moments of time. We could consider a content of a book to be information that is frozen in time until it produce some causal effect. So a book has consciousness, but that consciousness doesn’t experience time. But it becomes a part of our consciousness when we read the book (so some causal interaction happens).
OK, but we still haven’t solved the problem of binding. We can bind information for computers or for humans because there already is some actor (computer or human) that “knows” how to read that information. A human knows that he has to read the first page first and the second page second. Computer “knows” that it has to read all the commands one by one and to change the order only if it has readen a kind of “goto” command. And exactly those preexisting rules of reading allow us manipulate readers behavior to bind some information into some large entities. For example if we want to allow a person to move immideatly to an interesting part of our book and we know that he will start reading from the first page, we can put here a table of contents. That’s how we bind some pieces of information into larger objects – chapters. We know where a reader will start reading and we put here some information, that describes (influences thème process of reading of) next information.
If we speak about natural information, then the reader is probably the universe itself. And where does it start reading? Well, probably everywhere at once. It may sound confusing, but at least we know the order of reading: things that are closer to each other will infliunce each other faster than things at a larger distance. We know for sure that information can’t travel faster than the speed of light. Now if there is something that can manage the propagation of causal influence, make it faster or slower, change its direction, then it is a mechanism that allows to bind information into some larger entities. And that “something” are probably the laws of physics.
I can’t say that my thoughts should change something dramatically in the IIT or in the theory of valence present abow. But at least I have tried to look at the problem at some more fundamental level to make some gueses about how the reductionist world which we live in could lead to something so irreduceble as consciousness.

Comments are closed.