Clearly, 4) should not lead to consciousness, but then again, if the "replay" is perfect, there should be no difference on the biological neural circuitry side between 4) and 3), being injected currents to fire. Then you start to wonder whether the H-H mechanism is critical (2), or dendritic-somatic computation is critical (1), or even whether axonal spike propagation is critical (0).

What is missing here?

For philosophical zombies, I am thinking 3) is sufficient.

Or is it that, as long as the causality loop is intact regardless of the medium (3), we have qualia? I guess this is basically what Fading Qualia claims,,,

A question to philosophers:

Does Chalmers' "Fading Qualia" have anything to say about how far the silicon neurons needs to go in terms of mimicking biological neurons? Is it sufficient if other neurons cannot tell the difference, or does it need to emulate beyond that?

Hi Roman,

Thanks for the reply.

I totally understand that people use the word consciousness with various definitions and probably mine are quite narrow.  

So what if I simply say qualia? 

And what would be your take on Chalmers' Fading Qualia? From your view point, I am thinking that Chalmers and I are in the same category in terms of using the word consciousness.

My own interpretation of the proposed thought experiment is that, I am breaking down Chalmers' silicon neurons into multiple functional stages, and instead of replacing neurons one by one, I am replacing functions one by one.

Hi Roman,

Which of your work would you suggest? Your book sounds interesting, but can you suggest something that is freely available?

My usual work is electrophysiology, fMRI and psychophysics under illusory conditions where the stimulus is rendered invisible. If you can take a look at some of my work, you will see what I mean by "consciousness"

Article Adaptation to invisible motion results in low-level but not ...

Article Attention But Not Awareness Modulates the BOLD Signal in the...

Conference Paper Visual backward masking in rats: a behavioral task for study...

Hi Matasaka, please read this speculative paper of mine: Thoughts on Qualia for Machines.. it was just my humble attempt to understand this elusive concept. I'm neither a neuroscientist nor a philosopher; it was just my intuitive thought on the topic.

Qualia - the raw senses - I think is the lowest, the most primitive form of consciousness. No living animal is without it. All other forms of consciousness - integration of information, attention, behavior, etc - build on top of it. Please see more in the paper and throw me your thoughts. Thanks.

http://vixra.org/abs/1505.0146

Hi Nordin,

Thanks for your support on the definition!

Let me read your paper. Meanwhile, I have another Q&A thread that might be of interest to you. 

https://www.researchgate.net/post/How_can_we_test_machine_consciousness_How_can_we_test_various_theories_of_consciousness

Hi Nordin,

I went through your paper. Although I take different positions on how qualia may be generated by neural circuits, it was an interesting read! 

So what would be your opinion on the four conditions that I propose as minimum requirement for generating qualia? Let's assume for now that the emulation part is done by today's Neumann type computers on silicon. Please introduce new functions to neurons if necessary (quantum mechanisms?).

Hi Roman,

Thanks, I will be looking forward to your reply.

One purpose of the thought experiment was to build a platform that compares different hypotheses on the neural (and non-neural!) mechanisms of consciousness (e.g. the quantum hypotheses would predict condition 2?).

So I am curious to know the essence of your hypothesis and what it predicts.

Also it would be great if anyone can give me feedback to the thought experiment, so that we can make it a better "platform".

(Well honestly, "platform" is kind of an excuse, because I myself do not have a convincing answer at this moment, and therefore does not really function as a thought experiment!) 

Hi Matasaka,

Thanks for reading the paper.

About your proposed experiment, what exactly will be the intent?  What much can one learn about qualia from the experiments?

Imagine this: an alien or some other being trying to learn about what is the "soul", whatever one calls it, the life that makes a human alive. The alien breaks the arm - the human remains alive. Break the legs, nope human still alive. Pull off the ear, nope. Break the neck, human is dead now...  Should the alien conclude that the soul is in the neck?   

The thing is.. the brain is more than just neurons.. Each neuron is already a complex machinary.. what if qualia is somewhere deep inside the neuron?

Hi Roman,

Thanks for the review.  3 main points in your comments:

1) reference to "machine consciousness" without identifying the "internal" physical processes so reference to machine consciousness does not belong there.

My intent in the paper is to show the overall scheme, the overall working mechanism, a rough solution... just like before one work out a maths or programming problem in detail,  one might sketch first a rough solution.

I did mention that Hameroff's quantum process might be the "internal" physical process, but that remains just one possibility.

2) Panpsychism is also mentioned yet it is not clear why this metaphysics is called for when the author clearly points towards physicalism.

Nope, I did not say that. It cannot be just physicalism.  No matter how complex the machinery, however esoteric the physical theory that one have, eventually, one has to deal with the question of how does it all result in a mental state. One has to accept panpsychism. But as stated in the paper, I do no accept current interpretation of panpsychism. Rocks do not feel, light bulbs have no qualia, your car has no sense of its own.  

3) The major hurdle the author faces is trying to explain how his mental units "qualia packet"  combine to form a mental state. If the author does not explain this the manuscript remains a fairy tale.

Yes, I cannot explain in detail yet how it happen. But as said, its just a "rough solution".

Hi Masataka,

You wiill probably know my answer. The sort of qualia we talk about will disappear with 2. Qualia are in the relation between PSPs and an acoustic mode in the dendritic tree as I see it. 

Interestingly at the recent Towards a Science of Consciousness at Helsinki there were really only two types of session of note. One dealt with issues of physicalism and philosophy of mind and seemed to move towards a rather balanced position based on structuralism and the need to include qualia within such a framework. The other dealt with interventional biology directed at pinning down qualia mechanisms. This involved discussion of anaesthetics and how they may de-tune fine temporal co-ordination between neurons by influencing electromechanical modes in microtubules through binding to tubulins. The science is not all yet in place but it seems to remain the only viable avenue and moreover it seems to be getting a lot of new support recently after a rather long period of being a bit fringe - probably because the physics was not quite right.

As a result of this I have moved my position from focusing on electromechanical mode coupling in the cell membrane to coupling in microtubules. I am now persuaded that this makes a lot more sense since the acoustic parameters of dendrites are likely to be mostly determined by cytoskeleton, not membrane and Huxley had warned me that membrane modes did not look very plausible.

So it is number 2 for me -deep inside the neuron as Nordin says.

Just two comments in relation to Nordin's response.

Like Galen Strawson I think physicalism, to be coherent, must entail panpsychism, but it may be of a subtle sort, as Nordin suggests.

I see no need for qualia packets to combine. In fact I think combination is a non-starter. Togetherness has to be, as James said, a 'togetherness of relation to something' and you cannot then start adding togetherness to togetherness. The basic experiential unit must be the only sort of unit, rich at the fundamental level. That is absolutely no problem in modern condensed matter physics, with dynamic units spanning macroscopic structures, so I think the combining thing can be laid to rest. We just need to find dynamic units in neurons that can receive a rich enough pattern of potentials. I think we are nearly there.

I had a look at the paper on qualia and machines, Nordin. You have a  lot of good ideas there. I would just repeat the point that qualia packets cannot combine. A quale is already a relation between something and 'the way the world is to it'. You cannot add that to the way the world is to something else. The only combination one can have is that of all the aspects of the total way the world is to some one thing - a rich scene or experience for that thing. That is ready to hand in quantum theory since all the potentials encountered by a quantised mode are co-available to it in an indivisible way. But potentials encountered by other modes cannot be 'combined' with this. At the neural level there are quite enough inputs to a single dendritic tree to give the richness we need so there are no real obstacles other than pinning down the mode that encounters the potentials (presumably PSPs).

HI Jonathan,

Nice to hear back from you. Yes, I actually made condition 2), thinking about your hypothesis. I am a little exhausted today after surgery so let me try to respond in the following days.

Hi Nordin,

>What much can one learn about qualia from the experiments?

Jonathan and, from what I understand from below,  you would also need to say condition 2). 

>The thing is.. the brain is more than just neurons.. Each neuron is already a complex >machinary.. what if qualia is somewhere deep inside the neuron?

I had discussions with my good friend, who is an expert in the topic and has seminal papers and theories, says 4). I am guessing that Chalmers and believers of IIT would also claim 4).

If I was forced to make a choice, I would also say 4), but I have some doubts, since the division between 3) and 4) can be quite close (Let me explain in the following days)

So already, I don't think its totally useless as you claim in your alien arguments.

Hi Roman,

Thanks... Is there any paper that summarize the book, or can you produce one? I would love to read it, but the ringgit is not doing so good against the US currency, and USD100 is a big amount for me here :-).  

If it is 4 Masataka, then my question would be 'what is conscious of what' in 3?

There is a problem with the question, however, which I think is giving rise to the crux of all these discussions - that the meaning of 'what is conscious' is scale dependent and people do not take this into account. What I mean by that is that the question has to radically shift meaning if we are engaged in discussions about brains from the meaning it has in daily life. It is quite reasonable to say when discussing a field trip that the PhD students went into the forest and observed ant behaviour. At that scale it is fine to relate terms like observe or see or perceive or be aware to whole organisms or persons. But if we are interested in mechanisms of observation per se, such as how the ant is seen or how biased records might be for various reasons, we need to throw away the everyday meanings and work at a sub-organism level. SO we have to ask 'what sub-organism component is "conscious" in some more fine grained physical sense such as there being something that it is like for this component to sense the presence of an ant'.

As an illustration: A man is lost in New York. He is in TImes Square and wants to get to Central Park. He phones his girlfriend who knows New York and asks 'how do I get to Central Park'. She says 'Oh, all you need to do is go to New York and you will find it right there'. This is what almost everyone discussing consciousness is doing and so everyone is baffled. But at a finer grained level there is no need to be baffled if we ask 'but which street do I take from Times Square?'.

In other words to say that a person is conscious and a rock is not conscious is to open the discussion with false premises derived from the wrong scale of analysis. Until people see that the discussion of Angels on pinheads will continue to eternity!!! 

Hi Jonathan,

You are raising an interesting point, but unfortunately, I cannot think of a good reply as of now (I hope others can!).

So let me try to directly answer your question.

> If it is 4 Masataka, then my question would be 'what is conscious of what' in 3?

I hypothesize that "algorithms are conscious of its processed contents".

So if biological neurons are engaged in a single algorithm (e.g. predictive coding, generative model), it becomes aware of its calculated contents (e.g. topdown generation/prediction of the sensory environment).

In 3), the algorithm gets split into two realms, the biological and the artificial part, but still remains as a single algorithm.

In 4), the algorithm is there, a very simple one that only sequentially outputs a predefined variable (previously recorded firing patterns that are being replayed), but it does not have the necessary processed contents.

Regarding the above, I only have Japanese publications as of now, but the idea precedes the chaotic fluctuation thing that I was promoting in the other Q&A thread. I will test it with my split brain rewiring experiment.

http://redwood.berkeley.edu/seminar-info.php?id=253

https://archive.org/details/Redwood_Center_2014_04_30_Masataka_Watanabe

I finally got to start the experiments!

That would then be a very dualist position, Masatka, I think. You invent a new law of nature totally outside and unlike anything in physics - that algorithms experience. One might then ask why a gavotte does not experience - a similar rule based pattern of events.

And I am unclear what an algorithm would be. Is it the abstract rule set of if.. then.. etc. I don't think it can be because this is not even an instantiated dynamic event. If it is the instantiation then it will be an algorithm acted out in one particular way in response to inputs - like a gavotte danced in a particular variation. But I see all sorts of infinite regresses and baulking problems arising in relation to the boundary of this algorithm's domain and at what point in time the experience occurs. How does the algorithm know when it is ready to experience? Or when it has ceased to be an algorithm?

Hi Jonathan,

Yes it basically comes down to this point.

I consider myself as bad/good as Chalmers claiming that all information has its subjective side (consciousness).

And from my viewpoint, although you won't agree, claiming that any quantum mechanism to be conscious is no better.

We all need to introduce a new law of nature, "objective X is subjective consciousness".

Being an experimental neuroscientist, I put emphasis on the experimental testability of the hypothesis, especially at this pre-Newtonian stage of science. A testable hypothesis is always better than a non-testable one, because at least it tells us something, even if it is falsified.

I won't agree, Masataka, you are right. I do not need to introduce a new law because the existence of the observer's experience is already required by physics. But even if I do at least need to fill in new bridging details they are consistent with physics in the crucial matter of localisty. In fact not only are my suggestions consist with the rather complex requirements of locality in modern physics they are actually derived from those requirements. In contrast the event of an algorithm experiencing would be non-local and therefore in direct violation of physics - requiring indeed a new dualistic law to apply in some special circumstances. Special circumstances are best avoided.

I agree that testability is crucial. I doubt you will be able to test the idea of an algorithm experiencing precisely because the posit is non-local and therefore contingent on events distant in spacetime and baulkable. Almost by definition you cannot test ideas like this.

On the other hand my theory has proven testable. When I arrived at a model in 2005 I predicted that dendritic membranes would support an electromechanically coupled resonant acoustic mode of 'sliding-cuff' form at 1-10 megaHerz which would influence the opening of stretch sensitive ion channels and thereby mediate post synaptic integration as well as 'experiencing the PSPs. The acoustic mode was put in the membrane but I recognised that the acoustic parameters of the mode would be dependent on cytoskeleton, just as the acoustic parameters of a mode in a violin string are dependent on the structure of the violin. 

Recent work has shown that microtubules support unusual electrical conductance modes that can resonate at 3megaHertz. To do that it is assumed that they are somehow electromechanically coupled, so there will be acoustic modes with group velocities at that frequency almost certainly. Moreover, microtubules are anchored to ion channels and receptors in the membrane. Intriguingly, although my hypothesis looks to have been wrong in placing the coupled dipoles in the membrane, an acoustic mode in the cytoskeleton would in fact produce exactly the same 'sliding-cuff' effect as I proposed (just with the skeleton sliding under the membrane rather than vice versa) with the same effect because of channel attachments. 

So it looks as if the test I proposed for my model 10 years ago is close to being fulfilled, even if with a slight adjustment to the premises. Moreover, anaesthetics appear to interact with tubulins in a way that might well change the resonance properties of the system and there is a lot of evidence for anaesthetics working through 'detuning' of neuronal co-operation rather than actually blocking action potentials.

Jonathan,

About the point you raised: "I would just repeat the point that qualia packets cannot combine. A quale is already a relation between something and 'the way the world is to it'. You cannot add that to the way the world is to something else. The only combination one can have is that of all the aspects of the total way the world is to some one thing - a rich scene or experience for that thing."

I sense a confusion.. A qualia packet is not the same as qualia, the mental state itself. One do not experience a qualia packet itself. I guess, its due to my own confusion between the plural and singular form of the word itself. I should perhaps rephrase it all as follow, using the word "quale" rather than qualia:

A quale packet is not the same as quale, the mental state itself. One do not experience a quale packet. A quale packet is supposed to be generated by a single neuron.

Why must it be in packets? Suppose a single neuron generates a quale packet qi, and suppose a set of neurons of size n+1 must coordinate together to sustain a quale Q1. Then, if a quale is not decomposible, the following is true:

     q0 = q1 = ... = qn = Q1............ (1)

But consider "neural reuse" - as noted in a number of papers, neurons are typically reused, which means that some neurons will be used for other quale Q2, for example:

    qm = qn = qo = qp = Q2........ (2) 

which contradicts equa (1) above.

Hence, logically, it should be:

q0 + q1 + ... + qn = Q1

and 

qm + qn + qo + qp = Q2........ (2) 

That is, quale units add up to form a complete quale.

Matasaka,

Sorry, I didn't explicitly choose between the options, as I didn't quite understand the experiment. But now I do, and yes, I would go for (2) too.

About your statement: 

" I hypothesize that "algorithms are conscious of its processed contents". "

So if biological neurons are engaged in a single algorithm (e.g. predictive coding, generative model), it becomes aware of its calculated contents (e.g. topdown generation/prediction of the sensory environment). "

Do you mean that neuronal network is like a network of computers, and that the firing from one neuron to another is more than just an electrical signal, but it is a data encoding, one that contains certain "necessary processed contents"?

What sort of content would this be? It sounds an intriguing idea, but is it based on some observations, experiments... You mention some papers in Japanese - can you translate the gist of that? Thanks.

By the way, Matasaka, I do not know yet what is your opinion on how qualia are generated, which explains  why initially I did not quite understand the purpose of your experiment.  

Dear Nordin,

I don't think the plurality matters much. I am really just restating the combination problem, since you are using a panpsychist framework (sort of).

If a quale packet is not itself a quale the presumably it is some biophysical event. I think I might understand what you are proposing better if I knew what biophysical event you are proposing it is and what experiencing biophysical entity is getting the benefit of these packets in the form of a quale or aspect of its experience.

Another point that may be worth mentioning is that neuronal spikes almost certainly have propositional meanings rather than 'wordlike' meanings. So a signal would not mean 'red', but rather 'item q is red'. The literature of feature binding got very confused because van der Malsburg suggested that spikes might mean 'yellow' and 'triangle' . He actually called these propositions but of course they are not. And a computing system (in the broadest sense) can only draw inferences from juxtapositions of propositions, not just words. There is reason to think that the brain can cope with something like 4-7 items at one time and presumably it has a code for tagging their identity, as in 'item q' or 'item p'. It can then 'bind' 'q is red' and 'q is round' to infer that q is red and round and might be a tomato.

On this basis it is very unsurprising that cells get used for different purposes since in sensory pathways all sorts of complex comparative inferences are going on to do with light intensity and ocular saccades. So firing of a neuron may well mean 'p is left of q', which could be used under a very wide range of circumstances.

Hi Jonathan,

"what biophysical event"   -  I try not to think in terms of specific physical event, at least not at this stage,  preferring instead to think of what logically should be the case, how would nature (or God) logically engineer this thing.. Hence, I would just think of this event as some physical event somewhere within the brain cell..

Suppose, it is some microtubules in there engaging in a quantum process, and as part of its "inner" property, generates a certain qualia unit. Now, its hard for me to imagine  a single event generating a qualia such as the taste of spices or chocolate which is simply too subjectively complex, too rich. Visually, I imagine the "wave" associated with such a rich qualia to have a  complex pattern.. I just think that it would be natural to think that nature would compose such pattern from simple, unit patterns, and logically each unit pattern would be produced by a single neuron. Neurons "orchestrate" together, their wave pattern interacting, superimposing, to form a single coherent qualia that we can "feel".

"neuronal spikes almost certainly have propositional meanings rather than 'wordlike' meanings"....  I have read some neuroscience papers correlating neuronal activation with colors, emotion, etc... But I've not come across your assertion.. Maybe, not being the guy from the Brain department, I might have missed out some new developments... Could you please kindly cite some references for me to read up on.. thanks.

Hi Jonathan and Nordin,

I openly admit that how qualia is generated from algorithms is simply a new law of nature, as in the "law of gravity". I cannot provide any further explanations. I just believe that qualia rides on brain mechanisms that is used for the remainder of functions (e.g. sensorimotor transformation), which is based on axonal spikes, PSPs, somatic PSP integration, neuronal firing and most importantly, requiring a massive number of interconnected neurons.

I see in Jonathan's theory that you try to do the same thing until the final stage where you introduce quantum effects in a dendritic tree within a single neuron.

The problem I have is that, I just cannot believe that a single neuron can take over our whole experience. I understand it as a modern version of William James' Pontifical cells.

'William James in 1890 proposed a related idea of a pontifical cell.[20] The pontifical cell is defined as a putative, and implausible, cell which had all our experiences. It is in this different from a concept specific cell in that it is the site of experience of sense data. James's 1890 pontifical cell was instead a cell "to which the rest of the brain provided a representation" of a grandmother. The experience of grandmother occurred in this cell.' (Wikipedia: https://en.wikipedia.org/wiki/Grandmother_cell)

Is this true?

Well then, putting aside the above problem for now, what do you exactly mean by

" I do not need to introduce a new law because the existence of the observer's experience is already required by physics."?

Are you referring to some of the extreme interpretations of quantum physics that requires a conscious observer? I thought they are basically falsified by today. And anyhow, even if you are taking them for granted, isn't requiring a conscious observer and being the conscious observer itself a totally different thing?

Hi Jonathan and Nordin,

But I do appreciate Jonathan's attempts to avoid the other problem that other quantum theories seem to have.

Let's say consciousness is integrated over multiple neurons with some quantum effects. How does any quantum effect select which neurons to integrate and which not? In general, neuronal representation is not organized in a physical distance based manner. There are some exceptions like functional columns, but then again, much wider span of neurons needs to be integrated retinotopically and hierchically. As an example, how do we perceive a close up red rose that extends bilaterally? How are only the "red" neurons integrated, but not the "green" and "blue", when they can be equally activated by leaves and the sky?

In contrast neuronal spike communication based on hard-wired axons and synapses  has the selectivity and flexibility to get the job done.

HI Nordin,

"Do you mean that neuronal network is like a network of computers, and that the firing from one neuron to another is more than just an electrical signal, but it is a data encoding, one that contains certain "necessary processed contents"?"

No, single neurons for me are just single neurons, nothing more, nothing less. Simple integrate and fire neurons are good enough for me.

The algorithm is implemented on the neural network as a whole, as in the usual generative model.

Nordin,

I realise that one can propose a model without being precise about the biophysical events doing certain jobs. However, one needs to define the events in terms that can actually do those jobs. Qualia are the features of a dynamic relation - between some biophysical unit and its immediate biophysical environment in the brain - i.e. they are features of a dynamic relation not of a 'state' or indeed an event describe without indicating what is relating to what. So a somatic spike in a cell is not going to be any good to us because there is no indication what this spike will 'be like something' to. This is where in a sense I am with Masataka in suggesting that a quale must go with some 'causal passing on' (although Masataka wants to take that further to an 'input-output relation' and I will come back to why I have trouble with that).

So if a single cell is generating a 'quale packet', which I think is a very good suggestion, then the event that is the quale packet will need to be something that influences something that could 'feel' that packet, taken together with a lot of others , as a quale. The only communicative events that cells engage in are sending signals to other cells. And most of the events involved in signal sending are just the ionic fluxes along the way. They will not be the quale packets because quale are not passed on to further events - we never get quale passed on from events outside or from other people. So we need to look at the event that actually IMPINGES on the next cell - which I think has to be either the arrival of neurotransmitter from our original qualia packet cell or the PSP that is generated. There I think is a promising option.

My apologies. I am having trouble with the software so I will curtail this message here and reboot. 

You worry about small things being too simple, Nordin. But remember Feynman has pointed out that at the small scale everything is MORE complex. Simplicity comes with aggregation. Individual events like a photon connecting a light to a screen are unbelievably complex. 

The next point is that there are no 'complex waves' spreading out in brains. There are no waves travelling across brains except the rather dull individual action potentials which are all identical in form (more or less). 

And the next point is that neurons do not 'orchestrate'. Nothing 'glues the action of one neuron to another - I think Roman is confusing the issue here. Action potentials in individual nerves are separate events that are not in any sense combined. The only combination we have is the CO-ARRIVAL of many post synaptic potentials at a neuronal dendritic tree. All other ideas about combining are inconsistent with physics and I think bogus. Moreover, there is no need for them, because co-arrival is what we need to explain rich inputs or experiences.

The stuff about propositional meanings for spikes is something you may find hard to track down. The literature on this is fragmentary and mostly unhelpful. The errors made by von der Malsburg have heavily skewed the literature. You may find useful papers by Shadlen and Movshon. It may be useful to start with the Wikipedia entry on the binding problem which I wrote some years back. It may have got changed but I tried to lay out some of the problems there. Some of my own views are given in an essay on my UCL webpage (Google UCL, Jonathan Edwards) called 'reply to Poeppel and Embick'.

Masataka,

You are a bright guy, and diligent in discussion, so please if you want to discuss my model get it right!!!!

My model has nothing to do with James's pontifical cell. Moreover, the wikipedia entry is garbled. This is a very confused area. The pontifical cell has nothing to do with grandmother cells. People need to think this through.

I am not for a minute suggestion that there is some special cell that experiences things and arranges responses. I am suggesting that there are millions of individual cells each experiencing the same content in parallel. My written accounts of my model make this abundantly clear time after time.

So there is no suggestion of one cell controlling our actions. There is no 'control centre' in a brain. That is a misconception based on ideas like 'purpose' and 'agency' that we need to abandon. Lots of cells mediate responses. But if they are to do so usefully they need to have a specific input that relates to a world scene. The most recent figure I have seen for the average number of synaptic inputs for cortical pyramidal cells is 40,000 although this may be a bit generous. That is absoutely complex enough to account for what we know about the richness of scenes. There simply is no problem. You may feel emotionally that it is not what you are expecting but in science we need to go with the figures!!!

When I say:

" I do not need to introduce a new law because the existence of the observer's experience is already required by physics."?

I do not mean any WIgner type nonsense about wave functions. I simply mean what Descartes and Newton and Galileo and Leibniz and all the other founders of physics knew - that physics is a set of rules for predicting WHAT WILL BE THE CONTENT OF EXPERIENCE when we observe under certain specified conditions. If you think about it everything in physics is calibrated ultimately by its disposition to engender certain sorts of experience in us. To use a physics textbook you have to have experiences to match your predictions with . It is so basic everyone forgets it. There is nothing 'physical' about physical events that we can recognise independent of our experiences so 'physical' is ultimately defined by a disposition to engender certain experiences. And that being so experience is required by physics. 

Hi Jonathan,

Sorry I cut one part out of your theory that your experiencing cells are highly redundant. But in terms of my point that, its very less likely for all information necessary for generating our entire subjective experience converging onto a single cell,

"million of cells experiencing the same content"

does not matter. It has the exact same problem of information convergence, only that you have a lot of backup.

Hi Jonathan,

Even if we assume a very optimistic number of single cell synaptic inputs, I don't see an area in the brain where all necessary information may converge. But ok, although its quite unlikely, we cannot completely disprove it either, at this point of anatomical knowledge.

So let's move on for now.

I guess I am allowed to claim that the firstmost key to your hypothesis, which differentiates it from others is that, millions of cells are experiencing the same full content, and what "I" am experiencing originates from one of them, which resembles a little the quantum multiverse arguments.

How are you going to experimentally test this aspect of yours? I am aware of the Arizona attempts you've mentioned before in regard to microtubules, but if I understand correctly, that would be only testing the more generic quantum aspects (which I heard having serious non-quantum artifacts,,, ).

Hi Jonathan,

I agree with Matasaka.. in the same way info is distributed over cells, the same should be the same for qualia. It does not make engineering sense to have "a lot of backup".

As for the wave interaction, I think its not something classical like action potential or EM. If neurons are quantum devices  (I always say "if" as I do not want to commit to the nature of the physical processes yet), then their wave functions can interact as in a quantum network. It makes logical sense (to me at least) even if we are not at a stage where we can actually see these things in action, due to lack of apparatus or due to our lack of understanding of quantum phenomenon.

And by neurons 'orchestrating', I do not mean that they are in any way tightly coupled or centrally synchronized to one another. I mean actually co-occurrences of neuronal activities.

Hi Masataka,

"No, single neurons for me are just single neurons, nothing more, nothing less. Simple integrate and fire neurons are good enough for me."

As far as I know, neural networks, deploying simple integrate-and-fire neuron has been regularly engineered over the past years and decades, but there is no evidence, no reason to believe that these computer programs generate qualia..  My apology though if I understand you wrongly...

" Let's say consciousness is integrated over multiple neurons with some quantum effects. How does any quantum effect select which neurons to integrate and which not? ... In contrast neuronal spike communication based on hard-wired axons and synapses  has the selectivity and flexibility to get the job done."

The brain works as a system, and even if the neurons are quantum devices, it still needs networks and spike-based communication to complete a quantum network. I believe that the system must be complete. If you replace the neurons with something simpler, it (the qualia) will not work. And if you replace the spike-based communication with something simpler, it will also not work as the interaction required between the neurons will not be there. 

In fact, instead of thinking like the blind men who held different parts of the elephant trying to figure out what it is, it might be beneficial to adopt a system view and try to see things as a whole.

Human and other animals have an experience reflecting what they are doing and what they are doing is not represented anywhere but result of the interaction of their body with certain aspect of their environment. My hypothesis is that there is a dynamic nexus of interaction that is a informational singularity and where the growth of the interaction is emanating.  By informational singularity I mean that at the nexus of interaction there is no decision algorithm that is applied, there some form of information interaction totally unspecifiable is taking place for the decision of the growth.  It is why I call it a singularity.  It is where learning and growth emanate from.  If you find the idea vague, I agree.

Anyone here close to David Chalmers, Stuart Hameroff and Ramachandran?

Since theirs were the main papers I cited, I would very much like to communicate with them, to get their feedbacks, however damning.. But no response yet from anyone of them, even after 2 emails about 2 weeks apart.. 

Hi Jonathan,

And coming back to the main problem that we were discussing, whether or not you also need to introduce a "new law of nature",

"There is nothing 'physical' about physical events that we can recognise independent of our experiences so 'physical' is ultimately defined by a disposition to engender certain experiences. And that being so experience is required by physics.",

the above argument seems to be mixing up, 

"experience is required by physics" with "physics is experience (a law of nature: objective X is subjective consciousness)". Taking for granted the former, does not automatically lead to the latter. Therefore, my position is still that you are in fact introducing a new law of nature,

"quantum effects in a single dendritic tree is subjective consciousness".

BTW, I don't have any problems with this at all. I just want to claim that your stuff is equivalent to all other stuff (e.g. Chalmers' Dual Aspect Theory, Baars' Global Workspace) including mine, in terms of the need to introduce a new law of nature to "explain" subjective experience.

I thought your only possible escape was relating it to (and expanding) the extreme interpretations of quantum physics, but I am relieved to hear you saying,

"I do not mean any WIgner type nonsense about wave functions."

 Another way to look at it is, doesn't your arguments also exactly apply to my hypothesis that "algorithms is experience", because after all, algorithms are realized by neural processing and that is physics.

Hi Nordin,

" As far as I know, neural networks, deploying simple integrate-and-fire neuron has been regularly engineered over the past years and decades, but there is no evidence, no reason to believe that these computer programs generate qualia..  My apology though if I understand you wrongly..."

Yes, I basically agree, although I take the position that there is no way to prove or disprove that even Chalmers' thermostats are conscious as of now, and that perfectly applies to modern artificial neural networks! But yes, I am thinking that we need to push things further to get close to the real thing, and then we can start testing as in my other Q&A thread: 

https://www.researchgate.net/post/How_can_we_test_machine_consciousness_How_can_we_test_various_theories_of_consciousness

But importantly, I take the position that its the complexity of the circuitry, not that of the single elements, that matters in the end. (But then again, we might need to go a little further than integrate and fire neurons,,,)

"In fact, instead of thinking like the blind men who held different parts of the elephant trying to figure out what it is, it might be beneficial to adopt a system view and try to see things as a whole."

I completely agree that we need a system view and the algorithm hypothesis is my attempt to do so. The difference is, whether or not we are introducing a "ghost (quantum effects)" to the system.

My prior point would be that this "ghost" is not selective in terms of which neurons to integrate, and if not, it cannot be the source of our subjective experience, even if its causality affected by the more usual neuronal firing mechanisms. From the quantum effects side, subpopulations of activated neurons from dissociable neural circuitry only appear as a single population of activated neurons, because it cannot make use of the sophisticated and selective synaptic connectivity for integration.

Is there a way at least to "tag" quantum effects by firing properties such as synchrony or oscillation and only integrate the "tagged" ones? Then I see some potentials, given that the quantum effects reach beyond the size of single cells in the first place (Isn't that heavily debated? Which side is winning now?)

It is interesting that we have a three way relationship between you, Jonathan and I. Because in terms of introducing quantum effects, Jonathan's attempts appeal to me as a cleaner approach, not having to deal with the problem of combining quantum effects beyond the distance of a single cell. 

Matasaka,  

I think you are actually right; the past few hours, I'm quite stuck on this one...

On the one hand, I would like to think of qualia as being distributed over neurons. On yet the other hand, I have difficulty in figuring a solution to the ghost effect...

Maybe, the solution is what Jonathan has been saying all along about my paper - to put the "combining thing" to rest. So, the effect of each neuron will be like a pixel on the screen. It lights up, but there is no need for explicit combination of qualia packets, of "individual pixels"... The overall effect, the overall qualia, will be like a full screen, a computer monitor..

Then again, there might be "tagging" as u've suggested..

The integration is only over firing neurons?   ... my thoughts a bit clouded now, will come back to this.

Hi Louis,

Welcome to the thread!

Your singularity arguments sound interesting, but its a little difficult to grasp it from what you have described. Do you have anything published, or can you describe it in little more detail? 

I am sticking to my position, Masataka. I am not saying physics is experience, just that physics assumes that all these dynamic relations are known to us to be the instances we think they are because they give rise to experience somewhere in our head. I am simply saying that the final event in the chain is located in a relation in a dendritic tree. That is no more really than saying that it is the liver that makes bile or the hypothalamus that controls hunger. I admit that there are some details of correspondence, which I have yet to make any real headway in working out, between specific patterns of potentials and specific experiential features but I have not got into those yet.

The bottom line is that you are proposing a relation between physical dynamics and experience that is NON-LOCAL in contrast to the rules of locality of physics. That is the sense in which you are proposing a new law for me. I think we need a very very strong argument for introducing non-local laws because locality underpins all our abilities to make testable predictions. If you have non-locality then all predictions become contigent on what is happening somewhere else and if you start asking when that is allowed to be you end up in infinite boundary regresses.

Dear Nordin,

If one wants to integrate over signals from a set of neurons then I think you have to include non-firing or null signals. Complex patterns arise from some elements being 'on' and others 'off' so we have to include the non-firing ones. That poses a huge problem for all the neuroscientist who want to 'integrate' over 'cell assembies' defined by the fact that they are firing together. They really need to include cells that are synchronously failing to fire - which of course is a contradiction in terms!

If you want to use a computer monitor or screen analogy that has a venerable precedent. Leibniz proposed it in 1704 and I think this is where the screen analogy comes from. But note that for Leibniz nothing is viewing this screen in the way that the caricature homunculus (that nobody ever proposed) does. It is the screen itself that is getting the combined messages. And the screen is inhabited by a dynamic unit or monad which is what has the 'phenomenal experience'. In my model the screen is the dendritic tree receiving positive and null signals from thousands of cells in a defined pattern. The receiving provides the binding, as William James said it had to. The receiving unit is functionally indivisible in the way it receives. That is something condensed matter physcis allows. It does not allow non-local 'adding' of events.

Hi Jonathan,

There seems to be critical differences in how we think about the problem of consciousness and what needs to be done to say that we have solved the problem.

What is your take on the "Hard Problem" and the "Explanatory Gap"?

Hi Matasaka,

Rather than sidetracking into yet another fallacious thread that goes round and round in circles, why not work out (together?) the details of something like as follows:

Experiment 1

Keep network and firing constant.

Vary the neurons complexity -  from simple integrate-and-fire to full-blown neuronal architecture complete with microtubules and all..

Experiment 2

Keep the neurons constant.

Vary the network pattern complexity.

.. just my 2 cents..

Ok, maybe that has already been your intent...

What I was trying to say...  what sort of experiment that can answer reliably the question of: where is the qualia? Is it in the network or is it in the neurons.

That was the real question, right? Then, the task would be the design of the "perfect" set of experiments to address this..

Hi Nordin and Jonathan,

Yes, there is a danger in spiraling down into limbo,,,

So let me just clarify my position and maybe we move on.

I am thinking that our take on the "Explanatory Gap/Hard Problem" reflects on what kind of assumptions we make and how we tackle the problem of consciousness.

If we treat it as a problem like all other problems that science have dealt with in the past, Jonathan's point of "non-locality" makes all sense. Yes, it is hard to believe in today's  scientific standards that parallel distributed processing can ever holistically seat qualia (e.g. sense of bilateral vision). 

If we treat it as a "Hard Problem", we need to at least admit the need of a "new law of nature" to make any possible links between objective phenomena and subjective consciousness. After all, previous science is all about explaining objective phenomena and has nothing to say on subjectivity. That is the whole point of the "Explanatory Gap". Under this position, "non-locality" is not something that we can simply deny. It is a serious candidate mechanism that many hypotheses rely on (e.g. IIT, Global Work Space Theory)

But don't get me wrong. I do respect Jonathan's attempts to bring down the problem to where we may tackle it better with existing science. If you admit that the final link from "quantum effects in a single dendritic tree" to "subjective consciousness" is a new law of nature, I don't see any problems at all as a hypothesis. If you don't, I am really curious to know how you have solved the "Hard Problem" with existing science.

I would like that someone explain not how to solve the hard problem but how to describe it so that it is framed as a scientific problem.  For a mathematical problem to be hard, someone has to specify it.  Could someone specify in clear scientific term the hard problem?  There are many problems that science  by its method cannot  even formulated.  If someone think that consciousness is a problem that can be formulated scientifically, I am listening.  A lot of scientist want to solve the so-called problem but its seems that none of them bother formulating what they are supposed to solve. By formulating the problem I do not mean formulating it in all its breath.  If someone could formulate it for pain, colour, or whatever conscious aspect I would be satisfied.

Hi Everyone

I like Nordin's attempts to come up with a condition that dissociates our positions. I was also thinking of something like below, and curious to know what Nordin means by "constant".

So how about this?

5) At time T, all neuronal firings are blocked and all dendritic voltage distributions are clamped and held constant to the instantaneous values of time T.

I am hoping that the quantum effects remain intact. If so, does this lead to consciousness under your hypotheses?

Louis,

What do you mean by "framed as a scientific problem"? Can you give an example of what sort of description should be sought? 

Hi Nordin,

"That was the real question, right? Then, the task would be the design of the "perfect" set of experiments to address this.."

So are you talking about a real doable experiment or a thought experiment? I am interested in both of course!

Dear Louis,

I will have a go at formulating the hard problem. However, I think it may benefit from a little delayering first. David Papineau has suggested that we all intuitively find it hard to see how phenomenal experience could come from physical matter. He may be right but I worry that the 'hard to see how' is actually an artefact of the naive way we have been taught science. I have never myself had any real 'hard to see' about this although I do have some idea what is meant.

So I would suggest winding time back 374 years and considering the 'hard problem' as it appeared to Monsieur Descartes. Descartes elaborated Galileo's idea that the world consisted of one type of matter. He decided the essence of matter was extension - by which he meant the 'ownership' of a domain of space such that it excluded other matter from that space. He then decided that all the laws of dynamics would be determined by the way matter collided with matter.

But Descartes was also convinced that there were 'thinking things' such as himself in the world. And here the hard problem arose. The mechanical interactions he proposed for matter entailed that all matter had parts and that each part could only collide with one other part of matter - at a particular point. Moreover, it was quite incomprehensible that matter should instantiate logical thought or understand language. At a deeper level there was the problem that matter had to be infinitely divisible (to avoid the Democritan paradox) and so there could be no such thing as a 'material substance'. A 'substance' for Descartes was an entity that did not depend on other entities, such as parts, for its existence. The usage of the word may have not been entirely consistent but essentially there could be no material substances - so Descartes cannot have been a 'substance dualist' (how sloppy these philosophers are!!).

Descartes proposes that the only option for there to be such a thing as a thinking substance is for thinking substances to be other than matter. He has clear grounds for making the sort of statement Chalmers makes; that 'it seems quite unreasonable that matter should be capable of experience' or whatever. That is a respectable scientific hard problem. Matter is no good because it doesn't even give us individuals. In as much as there are bits of it they can only interact with one other thing in each place, in a divisible way. Such 'bumpings' provide no basis for a syllogism in which two or more premises lead to a conclusion. Language makes no sense. Experiencing a complex scene makes no sense.

However, we only have to wind forward again about ten years and Huygens and Hooke and others establish that extension is not the basis of matter. It is internal force. And within fifty years Leibniz has worked out that while matter may still strictly speaking be infinitely divisible (maybe as in quantum foam) it is nevertheless constituted at various levels by dynamic individuals and that these individuals will relate indivisibly to everything else (as in the time dependent Schrodinger equation). So LEibniz sees that Descartes's hard problem was unnecessary. Descartes had thought that there were two mutually exclusive dynamic species. Leibniz realised that the difference was merely between individuals and aggregates - as in the difference between quantum dynamics and classical dynamics.

So since Leibniz there has been no need for a hard problem as long as we take dynamic individuals seriously. Leibniz himself is not helpful when it comes to pinning down the relevant dynamic individuals for human experiences although he gets tantalisingly close in New Essays in response to Locke.

Dear Masataka,

I do not make any specific claim that 'quantum effects in a dendritic tree link to subjective experience'. Firstly, we can forget about quantum level effects such as spin phase, since for the relations I am interested in a classical description is quite adequate. What matters is that it is an immediate or fundamental (or local) relation to an individual mode. Statistical thermodynamics recognised modes like this long before QM did. 

So all I am really saying is that subjectivity is associated with the dynamic relations of physics at the immediate or local level. That is why we only ever experience what is here now - or at least what is being signalled internally here now. Physics had always had an unread law that an observer will always observe only where he is when he is - here now. So we just respect that law right up to the limit.

Note that I am not suggesting that subjective experience is restricted to dendritic tress. In this sense I take a panexperientiaist view. I assume that some sort of subjectivity will be associated with any immediate dynamic relation in the universe. But since none of them other than those in neurons are ever here now for 'me' I will never be able to confirm that. The reason for choosing dendritic trees for huamn experiences is simply that that is where information about scenes is available to something. The reason for proposing acoustic modes is that they would be in a position to experience all potentials in a dendritic tree. The reason why we only talk about this type of subjectivity is that it is the only type that is correlated to a readout from neural circuits we call 'reporting our experience'. We do not even talk about subjectivity in cerebellar cells because they are not hitched up to a relevant readout. That does not entail that there is no subjectivity there.

Hi Matasaka,

"5) At time T, all neuronal firings are blocked and all dendritic voltage distributions are clamped and held constant to the instantaneous values of time T."

Practically, I suppose, rather than an instantaneous time T, maybe a period of time,t   T0 < t < T1 would be more appropriate.  But what do we do during that period, to test the presence of qualia.. I have read your thread on testing for consciousness, but I'm not sure what to be drawn from that. 

Will something like the following work?

1) Identify first (through a set of experiments) the set of neurons that correlates with a certain, very specific qualia - color, taste, 'pure" emotion, etc.

2) Now, during the time period t, during which you block all neuronal firings to that specific neurons, somehow present the participant with a qualia stimuli and ascertain their qualia perception - is it still there, for example the redness of red, the taste of coffee.

I'm talking about experiments that can be really done sometime in the future... of course, I might be saying things easier said than done...

Here's another thought:

I remember Michael Persinger's experiment. His team reportedly stimulated a certain feeling, a metaphysical sensation, by introducing a certain magnetic field into somewhere in the temporal lobe.

Suppose it can really be done. Can we perform the test for qualia as follows:

1. Select a very specific qualia, Q - say color, or the taste of something.

2. Identify (through a series of experiments) exactly the set N of correlated neurons and the firing patterns F.

3. Now, to check the hypothesis that qualia is isolated in the neurons, directly "arouse" the neurons that have been identified, through some artificial means and ask (or observe) the participants to see evidence of Q.

4. To check whether qualia is in the firing "algorithm", in the same or another set of participants, replace N with artificial neurons, and present to the participants a stimuli that is associated with Q. Observe whether or not they can perceive it. Presumably, animals have to be used in this step.. but pls - no cats,dogs - those are pets.

Dear Jonathan,

Nice effort.  I will comments on some aspects of what you said.  

Descartes in the first meditation did a process of epistemological elimination by doubt.  He was totally focus on thinking and so at the end he logically had to eliminate everything except the thinker doing the elimination.  He was right about the phenomenal difference between Knowledge and the act of thinking.  The former are theories, information and cannot be taken as true without doubt, this is the falsifiability of scientific knowledge, and the ''I think'' is not a knowledge but a thinking action. It is a what is like to think. It is phenomenal, like the ''I see'', ''I feel pain'', etc.  Husserl credited Descartes for differentiating the phenomenal from knowledge.  He was also right in thinking that science is limited to extension.  Extension is what geometry and mathematic allow to express.  Physics can expressed some invariant of dynamic systation but not the acting. And he was not a substance dualism, but he saw difference between knowledge and phenomenality and saw also that knowledge is not phenomenality/actopm.  Those are two type of  beast.  It is not a separation of the world in two domain of substances with a problem for one to act on the other.  Descartes was not an idiot.  Nobody yet have shown that science what can be expressed mathematically can specified an aspect of phenomenality. The Doing and the Saying it. It is why we cannot even specify the hard problem.  It is not a question of progress of science.  It is a question that action is a different domain of what can be said about the world (information).  There is a irreductible core to action. Leibniz focus on the actors amd call them monads. And he saw that science so far had no way to express the reality of an actor. Leibniz elarge the domain of phenomenatity/doing.  Descartes focus on the thinking only.  Leibniz consider perception as a doing.   The second contribution of Leibniz is that science (what can be say) is not about what is but about invariant relations among things.  Leibniz want to re-build the philosophy of Aristotle with the new science.  Aristotle had distinguish the living, the body that have a growing forms, that have a nature and the thing that do not have a nature, the agregate, the inanimate objects.  The unity of the body is its nature. I consider that it is the growing of the interaction.  The growing of the interaction is both the building of the body which is the visible interface of the interaction and the growing of the interaction/action.  

There is no hard problem can be stated scientifically because what has to be explain is the acting, and there is no explaining of the acting. Only invariant aspect of the action can be studied, not the irreductible/intrinsically impredictible aspect of the growth of the acting.  Science is only about invariant relation among existing things that can be specified in terms models. It is kind of obvious that growth creation of the acting is not stable and if it is not it cannot be expressed as a scientific model.  

Dear Nordin,

In a sense all your experiments have been done. The neurons whose firing correlates most closely with 'basic qualia' are those in primary sense receptors. If you poison these with simple things like a very bright light (to bleach retinal receptors temporarily) or chilli pepper (to block taste buds) or background noise (to block high frequency hair cell responses in the cochlea) you lose the qualia. You can replace the qualia with bionic implants at least for cochlea and now a bit for retina. Things are a bit more complex because colours do depend on integration in visual cortex, but we have a pretty good idea which neurons correlate best to which qualia.

But that has got us nowhere. Because it seems very clear that you do not actually experience anything if only these cells fire and not some cells further forward in the cortex. You get 'cortical blindness' and such things. Yet you need the early cortical cells even for imagining the qualia. The problem is that in a functioning brain the firing of specific neurons is irrevocably correlated with certain pathways of message sending and it seems likely that qualia arise once those messages have been sent and arrived somewhere where they can be experienced. Where are the 'qualia' in an expresso coffee machine? For sure they are in the sachet you put in the top - either dark arabica or columbia mild aromatic. But you get no taste until you put a cup under the bottom and wait for the machine to work. 

Qualia are not where a certain cell is firing. That I think we can be sure of, because the firing of a cell soma means nothing to anything until some neurotransmitter has arrived further along. Only God could know the cells are firing otherwise and it is 'me' seeing the red, not God. The tricky part is knowing what 'me' is. What is 'me' in your model?

Hi Jonathan,

"Where are the 'qualia' in an expresso coffee machine? For sure they are in the sachet you put in the top - either dark arabica or columbia mild aromatic. But you get no taste until you put a cup under the bottom and wait for the machine to work. "

I think of it this way.. the qualia is in the sachet you are right. Its higher order consciousness that performs the integration, producing the final "taste".

"What is 'me' in your model?"

Qualia + Higher-Order Consciousness = Me (the coffee machine)

Hi Jonathan,

"Because it seems very clear that you do not actually experience anything if only these cells fire and not some cells further forward in the cortex. You get 'cortical blindness' and such things."

Sorry if my understanding is a bit (or maybe "very") simplistic, but if it all work like in a pipeline, where physical event comes in at one end, generates a qualia, travels down the pipe to the other end where higher-level consciousness attend to it, then surely, any damage to any part of the pipe will result in some sort of blindness.

And if its really like a pipeline, then can we not work progressively down the pipe? Activate one end of the pipe, block the rest. Test for qualia. Activate a quarter of the pipe, and test again. Repeat again for half, 3/4 and so on... See at which point exactly is the qualia perceived.. At the very least, a more precise statement on the "where" issue can be made.

But Nordin,

1. Higher order consciousness is not an entity. It is a property of some entity, maybe, but what makes it higher order if it is experiencing basic qualia? What is lower order consciousness? It all seems a bit vague.

2. Whatever tastes is not tasting the sachet. That is left at the top. In fact of course my analogy is overgenerous because in the brain what passes through the 'coffee machine' is just some code for arabica or columbia - yet that code tastes good.

3. Me isn't the coffee machine because the coffee machine is not what tastes the coffee. That is why I use the analogy. A lot of neuroscientists assume that 'me' can be the coffee machine but this is simply absurd in common sense terms. Me is what gets the signals. The coffee machine is what sends them.

4. I don't see how qualia plus HOC = either me or the coffee machine. 

We need something to experience the qualia - exactly as Descartes said we did. There was nothing wrong with his logic there. The only mistake I think he may have made is to assume there was only one tasting me in a brain.

We seem to have crossed posts maybe. The problem with the pipeline is that his one is like ten billion balls of string all unwound and tangled together. Not easy to know which pipe you are blocking where!

The other point is that qualia do not pass down the pipeline. Only codes. I am not quite clear what you are wanting qualia to be. They are facets of experience - of some sort of input to something that expreiences. They are not something moved from here to there - we would use a biophysical description to cover that, not a phenomenological one.

Hi Jonathan,

I've been questioning myself, how a distributed neural network can holistically "sense" its spatially distributed inputs. I have to admit that, In this aspect, your hypothesis provides a clear-cut potential answer.

So my question is, what are your criteria for "something to sense its inputs",  that forces you to conclude that the dendritic tree is the one?

Is it that the inputs are concentrated enough in small space, so that, with help from quantum effects (e.g. spread of quantum wave packets), it may ultimately converge onto a single point? Or is it that some casual interaction becomes possible in such confined space?

Hi Nordin,

"1) Identify first (through a set of experiments) the set of neurons that correlates with a certain, very specific qualia - color, taste, 'pure" emotion, etc.

2) Now, during the time period t, during which you block all neuronal firings to that specific neurons, somehow present the participant with a qualia stimuli and ascertain their qualia perception - is it still there, for example the redness of red, the taste of coffee."

Maybe I don't understand. What do you mean by "present the participant with a qualia stimuli"? Are you talking about peripheral stimulation? If so, wouldn't this not work for any of our hypotheses?

My intention of the "freezing of PSP distribution", was to keep the quantum effects intact and discuss how much we can detach it from the more usual neural processing.

Dear Masataka,

There are two reasons for choosing the dendritic tree of an individual neuron. 

The first is purely informational. The integration of post synaptic potentials in a dendritic tree leading to an action potential (or not) is in a sense a single closed causal juncture. There is one output that is dependent on those inputs and no others, with no effective 'inbranching' or 'outbrancing' of the causal chain other than what happens in the dendritic tree. So if the system is going to develop a way to track and report individual events this is a good start.

The second is the more critical reason. The idea is that a dendritic tree is the sort of ordered bounded structure that might support an indivisible dynamic mode. There is no real need to worry about 'quantum effects' here. We are talking about energy bearing modes, which existed in statistical thermodynamics long before QM. Also, it is important to note that quantised modes are not 'small' and do not deal with points. They are dynamic units with a field-like format that can occupy any space you like, as long as it is defined by certain types of asymmetry. So a seismic wave following an earthquake is a quantised mode, occupying cubic miles of molten lava. Fermionic modes tend to have rather small domains but not always. In metals electron modes can be enormous. Massless Bose modes, like acoustic modes, have no real size limits. However, they only propagate within bounded ordered domains.

The problem with brain tissue at a larger scale than individual neurons is that it seems very unlikely to constitute a single bounded ordered domain of the sort that would make sense in terms of signal sending. A neural net is clearly an aggregate of bounded domains, just as a piano is an aggregate of vibrating strings. Giuseppe Vitiello has suggested that long range modes might span many neurons but the sorts of asymmetry he invokes are well outside standard interpretation of what a quantised mode might be. Moreover, I have a suspicion the time dependent Schrodinger equation equivalents are unwritable. Stuart Hameroff would like modes in microtubules to extend across many cells, but I am doubtful that the couplin gcould work. Moreover, this would go against reason one. You would not longer have a juncture that was closed and local in causal informational terms - generating an unstoppable regress I think.

Hi Jonathan,

Just a simple question: If qualia exists indivisibly in single neurons, why would neuronal activation involve multiple neurons? 

Dear Nordin,

Neural activation is the means by which each individual neuron sends messages to certain other neurons. The activation of a neuron X is only of significance in that it will cause a message to reach neurons A,B,C,D. There is no group function of the activation of many neurons.

A connectionist neural network model shows us that having large numbers of signal sending units connected in a divergent and convergent way can achieve complicated computations like recognition of faces or inductive discovery of rules using learning. Nobody ever suggests that the activation of many nodes in a connectionist net means that this activation has some concerted function - it is just the totality of all the nodes sending local messages. So there is no reason to think that there is any concerted function of activation of neurons in brains. There are a number of neuroscientists who seem to suggest this but I do not think with any justification.

Activation patterns are not experiences. The experience is not the combined bowing of all the violinists, it is the sound reaching a receptive ear. The combined bowing may correlate with what the ear hears but if there is nobody to hear it there is no experienced sound.

Hi Jonathan,

Found your page in UCL, and the writings you put up there....for the first time, it becomes clear to me where you are coming from.

It is an interesting idea that you have there.. though intuitively, it is somewhat hard to digest from an information theoretic perspective... I'll think more on this.

Hi Jonathan,

Thanks for the detailed answer!

"one output that is dependent on those inputs and no others, with no effective 'inbranching' or 'outbrancing' of the causal chain."

"support an indivisible dynamic mode. "

Under certain conditions, wouldn't it possible for a neural network consisting of  subpopulation of neurons in the brain to fulfill the two criteria? Of course, the "one output" needs to be extended to a vector.

Dear Nordin,

It took me about two years to digest the idea myself so give it time!

Dear Masataka,

I cannot see how a network will do. There is no principled reason to 'add together' inputs to several cells because the input to one cell has no causal relation to the output from another cell. It is causally or informationally non-local and so runs into baulking paradoxes. There are also serious issues about how you can expect all the incoming signals to have co-dependent semantic values - simply because they are not causally co-dependent and we need a causal account of semantics.

And although Giuseppe Vitiello has suggested modes in networks he cannot tell us what asymmetries these are based on. The asymmetries that underlie acoustic modes are well understood and you can write equations for energy content and transfer in a way that I do not think you can for his suggestions.

And the question remains - why do we need to include lots of cells? 

Jonathan,

Behavior is caused by activation patterns and the interaction that is produced is affecting back the activation patterns. It is non local but this non-locality does not prevent the activation patterns to be felt as unitary since its may find its unity into the intended action.  So if we locate consciousness into the action dynamic that are steering the action then it does not matter if this action dynamic need one thousand billion neurons to be effective.

Louis,

I just do not buy this. The activation pattern is felt by what? Physics does not have any place for non-local events in the chain from observed to observer in all cases where we can ascertain locality. Within the brain ascertaining locality is more difficult but to say that an experience can be based on an aggregate of non-local separate dynamic events is completely contrary to physics. If the suggestion is a new dualist non-local rule of causation - more anti-physicalist than Descartes, then what is the justification for such a radical move. There is no need for it so why make it?

I agree that the activations lead to behaviour but that is no reason to bundle them all together as an experience had by some entity. 

We could locate consciousness in the multiple sites of signal reception if we want to talk of consciousness in more global operational terms. But Nordin is talking of qualia, not global input-output relations.  We are considering consciousness in Chalmers's what it is like terms. .

Jonathan,

You experience your doing and not all the billion of events allowing it.  You move your leg. If you pay attention to it, you do not really know how you move it but only experience the leg moving and you eyes seeing it. What do we really experience is not even how exactly the leg is moving but just some aspect of that.  Usually we do not pay attention to our leg moving and so we expererience ourself moving on the sidewalk vaguely knowing the cyclical alternation of our leg but we normally put our attention on what is new in the happening, what is threatening, or what is interesting and un-usual.  We do not even move our attention.  OUr attention move to these and we notice what it select.  So there is a nexus of action where the attention move.  This nexus of action is controling this thousand billion cell being.  It does not need to be fixed in the organism and it depend as much on internal events than on external events. The nexus of action might by be free floating location in the nervous system. 

  Quantum physics is not a local physics. THere are quantum systems, even macroscopic ones where all locations immediatly interact with each other.  In the EPR experiments , far locations are in immediat interaction.  Plasma have some of these behaviors.

Hi Jonathan,

"There is no principled reason to 'add together' inputs to several cells because the input to one cell has no causal relation to the output from another cell."

If you take the full-passive position that "something needs to receive and interpret the input instantaneously at that time of input", then you run into the above problem.

But on the other hand, you also do make use of dynamics. If you allow the network dynamics, instead of your dendritic dynamics, to fold-out in time, and let it causally interact with other neurons to "interpret" past input, you can overcome the above "no causal relation" problem. Libet's neural adequacy which goes up to 500ms is in line with this way of thinking.

Hi Jonathan,

"And although Giuseppe Vitiello has suggested modes in networks he cannot tell us what asymmetries these are based on. The asymmetries that underlie acoustic modes are well understood and you can write equations for energy content and transfer in a way that I do not think you can for his suggestions"

Better understanding of something has nothing to do with whether what is true, especially at this stage of science.

If you take that for granted, any of the modern physics would not have developed.

So would you say, "modes in networks" is a possibility, just that its not well understood yet?

Louis,

I do not really understand your descriptions but they are clearly non-local in a classical framework and we know that never happens in physics. You have to invent special non-physical rules to explain that way and because of the locality the explanations can never be testable. It doe not work.

Quantum physics is absolutely local but according to a new set of rules that differ from the classical rules such that in classical terms there is non-locality within a coherent mode, but never beyond that. So we have absolutely precise rules for locality still. They allow richness of relation within an individual mode but not otherwise. So my argument is that experience needs to be considered within a single mode. So I agree that quantum physics is (classically) non-local and in that sense that it rescues us from the impossibility of rich experience in classical physics. But it is non-sequitur to then suggest that we can have non-local explanations in a classical framework of the actions of limbs and neuronal nets.

Masataka,

You ask: So would you say, "modes in networks" is a possibility, just that its not well understood yet?

I was trying to be polite because I have huge respect for Giuseppe. I owe him my understanding of Goldstone theorem. But I said that I do not think you can write equations for energy content for the notional modes he wants to associate with his notional asymmetries. No such equations have been attempted as far as I know. I personally think they cannot be written for fundamental theoretical reasons.

The issue is the coherence of the theoretical framework, not the issue of being well understood which was just a lead in remark.

Masataka,

Spreading to network dynamics is non-local, as I have said. Physics does not allow it. There is no point in having interpretation in a series of events because at each new event you have to interpret again. Interpretation is not carried over in the world. You have the infinite regress of the homunculus. All networks can do is turn one set of signals into another set of similar signals - nothing achieved.

The only solution is Descartes's solution - the 'interpretation' comes free with the immediate local relation of input. There is no reason to suggest otherwise (since as indicated above it achieves nothing) and there is no need to violate what we already have in physics since physics leaves 'interpretation' to the observing event.

Hi Jonathan,

So what would be your requirements for "modes" to be the one? I don't think having nice equations is a factor.

Jonathan,

You hold  that for a state of visual perception of a complex animal body to be used by an animal, it has to be localized at one single location or many single locations. 

It take approximatly 1/10 sec for for a state of visual perception to be generated: the microgenesis process (Jason Brown). 

 The external surface of the animal in the viewing direction has a certain topology.  Now you put your gaze on the animal and a luminance surface of this animal surface profile is optically formed on the surface of the retina.  I have observed thousand of luminance surfaces with graphical software and most of the luminance surface structure correspond to the animal surface structure and the imaging process only destroy a small fraction of the original topological structure of the surface.

Suppose that the visual process proceeded into a number of steps, each steps being a step of the differentiation of the surface of the animal that is very similar to the actual embryological morphogenesis formation of the surface of the animal.  Each step of the surface formation being created by a symmetry breaking event in the differential geometry of the surface.  Refer to my Ph.D. thesis for details.  

Assuming that the visual system has built-in philogenic schemata tree of all possible surface structure and that the visual process in 1/10 of a sec activate a single path from the root of this tree to a leaf corresponding to the structure of a complex surface corresponding to a particular animal profile then the state of the neuronal mechanism instantiating this state will reflect the perception of the animal. Notice that the visual process history that took place in 1/10 of second (microgenesis) is mathematically similar to the ontogeny of the animal surface. 

I did not work out the neuronal mechanisms instantiating this but I do not see why the state of activation should be spatially localized.  A specific neuronal state might be distributed and when a specific activation state realized trigger other specific activation states.  A given activation state corresponding to a specific animal will determined if this animal is dangerous , small or big, and determine a range of possible action patterns themself being specific pre-existing activation state into a tree of possible behavior.  

Regards

Masataka,

We need a causal chain to pass through the mode on its way to output. The input seems pretty sure to be electrical potentials, because we know these are in some way integrated to co-influence neuronal firing. Neurotransmitter molecules are not integrated in this sort of sense so I think the rich input must be electrical. At the moment the assumption is that electrical potentials simply summate by cable effects and then trigger ion channel opening somewhere at the base of the dendritic tree, although there is increasing evidence for a more complicated mathematics of integration. In this case the modes influenced by the potential would be  at a molecular level in the ion channels and rather than receiving rich patterns they would just get a single suprathreshold input (for practical purposes). 

So we want a mode that can take up energy from electrical potentials in a complex pattern based way - which means a mode that extends throughout the dendritic tree along which individual PSPs are distributed. At this scale the modes of condensed matter physics are no longer at a molecular level. There are two options. One is long range electron modes of the sort that you get with valency electrons in metals. The other is phonon based modes of the sort that mediate piezoelectric vibration in quartz crystals in old fashioned radio receivers. Electron modes suffer from having very short decoherence times - maybe picoseconds. Phononic modes have no real limits on decoherence times since they are massless. There are a whole range of other arguments but I still think a phononic mode is most likely what we are looking for. I now think it is probably based on acoustic unit whose parameters are primary determined by microtubules rather than membrane, after hearing the work of Anirban Bandyopadhyay.

I think writing equations is important. I am talking in very broad terms here. For instance if you find when you sit down to write an equation that two terms which need to represent different causal elements are in fact the same thing then you can be pretty sure that whatever equation you write will have no determinable solution. You only need to write the equation at the level:

X = f(a,b)

to do that. 

Dear Louis,

The 'state of activation' of a bank of cells sending a pattern of signals does not need to be localised - for sure. But for those signals to meaningfully count as signals they have to arrive somewhere. For them to generate an effect at a single pass that is dependent on every detail of the pattern then all of those signals must converge on some computational event. The only way to do that in a single step is to have all the signals arrive at the dendritic tree of at least one neuron. If you allow several passes or steps you can use a pushdown stack or Turing tape system that simulates massive convergence by breaking it down into a series of comparisons of parts of the pattern in turn. But the brain has no time for that. Moreover, if it is done like a Turing tape then there is no event that could be an experience of the animal. It is all done bit by bit. 

So yes, the activation of the sending cells will of course be distributed but that is not experience. Experience means getting an input - unless you want to change the meaning of the word.

There is, I admit, a much more complex analysis needed if we are to understand how we can report these experiences, and in fact how we can think the way we do at all. We need some cells that do a subtracting or truth assessing computation - to provide a continuous internal feedback that tells the system its reports actually reflect inputs. In simple terms we need a juncture that takes in two complex patterns and checks that they are identical. That certainly requires inputs of complex patterns to individual cells, unless again you use a slow serial process. 

Masataka,

"My intention of the 'freezing of PSP distribution', was to keep the quantum effects intact and discuss how much we can detach it from the more usual neural processing."

What if the input pattern is important for a neuron, i.e. it affects the integration and firing of the neuron and its quantum effect?

Jonathan,

"The problem with the pipeline is that his one is like ten billion balls of string all unwound and tangled together. Not easy to know which pipe you are blocking where!"

I'm referring to the vision or color pathway. the neurons in the V1, V2 ,V4 regions and beyond, as per, for example, the 3-stage processing proposed by Zeki et al in  http://www.neuroesthetics.org/pdf/threecort.pdf

Surely, the pathway (or pipeline as I called it) has already been established by now?

Masataka, Jonathan... I see the need for us to go beyond philosophical debate, and perform experiments, any experiment, that would at least provide a clue on "where is the qualia"?   I am no neuroscientist, so I cannot provide detail, but can we not do something like as follows to resolve contending issues:

Hypothesis 1: Qualia distributed in a bunch of neurons (Nordin)

Hypothesis 2: Qualia is in individual neurons. (Jonathan)

Hypothesis 3: Qualia is in the network.  (Masataka)

Steps

1. Ascertain the neurons in the color vision pathway.

2. Determine the neurons most suspected to generate qualia.

3. Replace those suspected neurons with simpler artificial device, one at a time, at each time, performing a test for qualia perception.

A number of possibilities:

1. With more and more neurons replaced, the qualia disappear or becomes "degraded"  - Hypothesis 1 is to be accepted.

2. Qualia remains down to the last neuron - Hypothesis 2 wins.

3. Qualia remains even after all the neurons are replaced - Hypothesis 3 is correct.

I fear that our knowledge of the pipeline is much more primitive than you think, Nordin. We know that there are some cells in certain areas with functions at particular levels of differentiation/inference. But for none of them do we know where all their signal branches go, nor what sort of function the receiving cells perform. There are even more serious ascertainment problems down the line but that takes a lot of explaining.

We do not have any means of replacing a single neuron functionally so this sort of experiment has to remain at the armchair level, and almost certainly always will. I think other types of experiment may be more useful in building up an idea of where the qualia are - using a lot more indirect inference.

What seems fairly clear is that the signal receiving in the occipital cortex in V1,2,4 is not sufficient for the sort of qualia we report. If you damage the output from the occipital cortex to the rest of the brain visual qualia are lost even if the occipital cortex still receives signals.

I don't see neurons as 'generating' qualia. There is nothing for them to generate qualia for.  I think we have to see qualia as something sensed by neurons, either individually or in groups if that were possible. If they are 'generated' by anything it is the incoming PSPs.

Hi Nordin and Jonathan,

I don't think we are anywhere close to replacing biological neurons with artificial neurons, if you are talking about "doable" experiments.

On the other hand, replacing neural projections might be more realistic.

I have actual plans to split the two hemispheres of rodents, and then rewire them, so that we have full access to what is communicated between the two hemispheres. If you are interested, there are some details of this experiment in the last part of my Berkeley talk, 

https://archive.org/details/Redwood_Center_2014_04_30_Masataka_Watanabe

My algorithm hypothesis predicts that, as long as the higher visual level is shared through minimal connections, the bilateral visual system would function as a single neural algorithm and hence the animal would be able to perform bilateral tasks, without necessarily exchanging all lateral visual info required to solve the task.

What would your hypotheses predict?

Hi Masataka,

"I don't think we are anywhere close to replacing biological neurons with artificial neurons.."

What we need is a roadmap. In this roadmap, we shall have milestones like "create new mapping technology", "create artificial cell", "create artificial neuron", etc. and it can be a 20-, 50-, or maybe a 100-year roadmap. Someone, with the political and scientific clout, then needs to put forth the roadmap to EU or Japanese or American funding agency, and if it goes through, hopefully anyone among us here will live to see such experiments...

Anyway, that is an interesting experiment that you are proposing to do, something to think about over the next few days..

Hi Nordin,

I mean, replacing significant amount of neurons with artificial ones and completely rewiring them back to normal sounds basically impossible to me, at least in the next one hundred years. I am totally unaware of the potential of those future nano-bots, but still, it does not seem to be something that we "decent" scientists should be creating roadmaps of.

If we are allowed to have a 100 year plan, I would go for machine consciousness and test it by connecting it to our brain. Then we have full control of the dynamics and can try out whatever hypothetical mechanism for consciousness. Well not all though! (e.g. Jonathan's idea)

Hi Jonathan,

Its a lovely Sunday morning here and I have a thought:

At this moment, I perceive qualia of my fingers feeling and tapping the keyboard, my gaze upon the monitor in front, my cat lazying next to the keyboard, dozens of different colors, thousands of pixels in my view, the sound of the birds tweeting outside my window, dogs barking in the distant, the ambient sound of the fridge in the kitchen, the warm morning sun coming in through the window, my hair still feeling  damp after my morning shower, glimpses of memory running through my head, etc. 

Assuming your locality argument, does it not naturally follow that my combined experience - the different qualia of many different things - all sit in a single cell? Shouldn't there not be a single "mother of all cell" sitting somewhere in the brain, holding connections to every senses, every instantaneous thoughts, memory, etc. ?   A single cell that represents "me"? 

In short, my question is: what is the architectural implication for the locality argument?

Dear Nordin,

You are expressing the argument put forward by Descartes, Leibniz, Lotze, Herbart, James and a hundred others in the history of the study of mind. William James in chapter 6 of Principles of Psychology (on the net) goes through the argument in detail and concludes that the only solution that is not contradictory is that there shod be a single 'me' cell. But he dismisses it as fantasy.

Ironically, James himself gives the solution to the problem in the same text. Suppose that there are a million 'me' cells, of exactly the sort you are suggesting, but lots of them instead of one. Each will get all the complex patterns of qualia in some multimodal code.  Now consider what 'story' will be coming in as signals to each of those cells. It will be about a 'me' called Nordin with fingers on the keyboard, cat by his side, birds in trees and memories of more. It will contain no reference to the nature of the receiving cell. And crucially, it will contain not a hint of a reference to the fact that 999,999 other cells are getting the same story. It will contain no information about cells at all. So it will be just like the experience you describe.

In other words your introspective evidence cannot distinguish between the presence of one me cell and a million. This important point was made by a friend of Descartes who used the pseudonym Hyperaspistes in 1641. It was also made by ELizabeth Anscombe (philosopher of mind) in the mid twentieth century. Steven Sevush and I have both tried to emphasise it in our account of the theory of individual neuronal consciousness. 

There is an intuition that there can only be one me cell but it is based purely on an illusion generated within the brain - that there is a single me there. It is hard to say that this illusion is genetically programmed but maybe one could say that once such an illusion was created, perhaps through cultural exchange, brains found it very acceptable. There is a confusion in the literature about 'unity of consciousness'. Consciousness is unified in the sense that your fingers, the cat and the birds all seem to be known together as a 'scene'. But that does not mean that it is unitary or unique. There is no reason why there should not be a million copies of the scene all being subjected to useful integrative analysis simultaneously.

Hi Jonathan,

If there is a million "me" cells, then surely each and every cell must be perfectly in sync, doing the same thing, seeing the same thing at precisely the same time. For otherwise, there would not be a coherent "me". But then, there is surely a certain probability that a cell might go out of step, might decide to be a bit different, and this probability mathematically adds up to a non-trivial value for a million "me". So, which among the "me" would be "me"? Would it be not the case that having just the slightest, minutest difference among the "me"s would result in a malady, some sort of observable brain illness?

Shouldn't it be that either I have one and just one "me"  cell, or I would need some sort of mechanism to integrate the many "me" or to select from the many just a single "me".. the latter would of course lead us further down the rabbit hole..

If you think carefully, Nordin, I think you will discover that your problems are based on circular or non-sequitur arguments.

Forget about the subjective aspect for a moment. Would it not be the case that having just the slightest minutest difference in events in cells would result in a malady - just in terms of the operational mode of the brain? I think you can see that subjectivity is actually irrelevant to that worry. Presumably the brain is synchronised well enough to handle information the way it needs to. Whether or not all the experiences are absolutely synchronised does not actually matter since none of them knows about any for the others or needs to. So your out of synch or slightly different content problem is not a problem for my theory - it is just a problem for any model of how the brain operates in behaviouristic terms.

In fact there is reason to think that millions of me cells are probably very finely synchronised in terms of input. Synchrony of output is not wanted because the brain wants each cell to output for a different reason, and if timing is important to signal meaning it would want output timing not to be synchronised.

And I come back to the fact that synchrony is completely irrelevant to 'a coherent me' because the coherence of the sense of me only needs to occur in each cell separately. There is no coherence between cells. There is no need. That is the whole point of the model. And the question 'which me cell would be me' is one that so many people ask - one colleague asked it about thirty times in the same discussion session - does not have or need to have an answer. if you ask the question you have not followed the reasoning behind the model. You need to put aside illogical preconceptions about 'only me' and stick to the logical structure of the problem. Many people seem to find that extraordinarily hard, but once you do it then the nature of the problem bgins to become clear.