[flr]

Can you explain this? It seems to me that the 'China Nation' thought experiment simply describes a human brain on a large scale...

It is more than that.
The “China Brain” (or “china Nation”) is very shortly in Wikipedia, see: https://en.wikipedia.org/wiki/China_brain
If one replaces neurons with people and connections between neurons with hand waving/shaking between adjacent peoples will the resulting state be conscious as the imitated neuronal state?
If the function is all that matter for consciousness then the pattern of hand-waving of people is similar to connections between neurons and therefore perform the same function as neurons but implemented in a different substrate.
The question is: Will there be consciousness out of the ensemble of people imitating through hand-shaking/waving neuronal pattern? If yes of whom? Of a new individual of the consciousness of the person (whose pattern is imitated) is expanded over the “china brain”?

If functionalism is correct (the function is all that matter and not the substrate) then the “china brain” is conscious.

There are variations of china brain though experiment. For example if we replace neurons with silicon neurons will we get conscious? Or if create an accurate holographic representation of the brain will that hologram be conscious?  Or if  succeed to record in a computer a copy of the neuronal connections, will that Turning Machine copy be conscious? If not, then why neurons generate consciousness? How are the real neurons different?

It appears to me that essential to functionalism is the question: What is/are the physical process(es) that underpin a conscious percept and at what scale it  occurs?
Certain physical configurations (by configuration  I mean its time-dependence as well, and not some structure frozen in time) are accompanied by aware percept (such as how-is-it-like-to-experience -sweet when eat a candy candy).
I suspect that   aware percept necessitate so tight restrictions on the underpinning  physical substrate that only neurons have it and silicon or china brain does not.  But what if we imitate in china bran (or hologram or silicon) all that is measurable about neuronal connections, why would not china brain be conscious? Then what we copy (at the level of classically describable and macroscopically measurable interneuronal connections) is not what underpin an aware percept.

[flr]

in 2014 was 1 chip that holds 1 million neurons and 256 million connections, or about 1/400,000 of a brain.
Cobble together 10,000 of these chips in a data warehouse, and you have 1/40 of a brain.

I am pretty sure one will get something totally different from a brain.

What "1/40 of a brain" is even supposed to mean? If a tight vessel blow inside my brain and do some damage not even visible without microscope my entire mind/personality could be lost and I could became in a vegetative state; that is how interconnected things are inside brain.  I truly have reasons to believe you will not get a brain (or 1/40 (sic!) of a brain) by clogging many CPUs together.

[dlorde]

...I suspect that   aware percept necessitate so tight restrictions on the underpinning  physical substrate that only neurons have it and silicon or china brain does not.

But why do you suspect this?  What properties of the substrate do you think might be relevant? (because if we know the relevant properties, we can include them in the emulation).

I grant that the 'China Brain' would function incredibly slowly compared to a human brain, which would mean it would need appropriate stimulation at a similar rate, and its output responses would be correspondingly slow, but it's just a thought experiment, and I'm assuming it is intended to exactly represent the function of a brain - neurons (& other active cells), in their networks and structures of networks, scaled up so that the functionality is similar - otherwise what's the point of the thought experiment? If the people can't communicate with as many others, or at a frequency or processing speed as would occur in a real brain (but suitably scaled up), they obviously wouldn't be able to function usefully at all.   

Suppose we approach it from another direction, and, say, imagine a thought experiment, where we develop artificial programmable neurons that can emulate the behaviour and connectivity of any neuron in the brain (chemical sensors & effectors also allow it to respond to and modify the chemical environment around it, just like a biological neuron). So we scan someone's brain (a willing volunteer!), identifying every neuron and it's connections, and measuring how each one behaves; then we start replacing them, one by one, with the artificial electronic neurons, programmed to behave identically. Assume each replacement replicates the relevant behaviour of the original exactly.

Do you think there will come a point in this cell-by-cell replacement when the volunteer will no longer be recognisably conscious, despite there being no functional change in his brain? If so, can you explain why you think so?

[evan_au]

What is "1/40 of a brain" even supposed to mean?

This was mentioned in the context of comparing how many neurons and synapses there are in a human brain vs a silicon computer which tries to implement neuron-like structures.

I know we are comparing apples and aardvarks here, because there are many things we don't know about how individual neurons work, or how different neurons differ from each other.

One thing we do know is that it takes a lot of brain power to control muscles smoothly, and a large fraction of the increased brain size in larger animals is due to the extra neurons it takes to control the extra muscle fibers.

Fortunately, developments in Diffusion MRI are allowing us to map the high-level connections within the brain, as part of the Human Connectome Project. This will assist us in identifying functional groups of brain areas, which will help us understand the human brain, and perhaps better understand the human mind. It will also guide the design of AI computers; it is very inefficient to blindly connect every computer chip to every other computer chip - breaking them up into functional blocks makes a lot of sense.

China Brain...Will there be consciousness out of the ensemble of people? If yes of whom? Of a new individual or the consciousness of the person (whose pattern is imitated)?

I suggest that if you have a number of conscious beings which are able to communicate, as soon as you put a group of them together, you create a new conscious being with a different identity than the individual pieces - an organization. This larger consciousness may be a family, a tribe, a political party, a religion, a discussion forum, a nation, a pod of dolphins, etc. This larger consciousness will have slightly different goals than the individuals from which it is formed.

These different goals create tensions. Sometimes an individual gains control of an organization, and bends the goals of the organization to match his or her own personal goals, with negative results for other individuals, and the organization as a whole.

Getting individuals to cooperate is hard, since they often compete for the same resources; getting organizations to cooperate is also hard, since they also compete for resources and individuals.

With a few small exceptions (like interactions between humans and their work animals), human organizations do not yet extend outside our own species. Perhaps AI might be the first such extension, causing a shift in the outlook of humanity?

[flr]

But why do you suspect this?  What properties of the substrate do you think might be relevant? (because if we know the relevant properties, we can include them in the emulation).

Because we can solve the tiling problem or halting problem for cases computers (i.e. Turning Machines) cannot do it.
What means to solve a problem with a computer? It means to generate a physical-state (inside machine) isomorph with the proof of the given problem. Indeed if one will follow the electronic states (in both space and time) inside computer one will find a physical structure isomorph with the given problem
What means a non-computable problem? It means that it cannot be solved within the computing model of a TM. Can we communicate to computer a non-computable problem? Yes but it is as if one write can incomplete program. 
What specific example of non-computable problem can be given? The halting and the tiling problem. (just google “the tiling problem”, see also wikipedia). The tiling problem can only be solved by computer (TM) only if periodic patterns (like in crystals) are generated. For non-periodic cases TM will never find solution and cannot figure out that it will never find solution. It is not that the programmer did not write a smart enough program, instead it can be mathematically proven that the computing model is insufficient and it is not possible to write a code for TM to solve the tiling problem in finite time for aperiodic case.
Our minds can identify and solve (for simple cases) problems that computers/TM cannot solve, and we can even explain why this is the case. When doing so it must be the case that a physical process/structure isomorph with our thinking/proof occurs  inside our brain. In other words, if I solve a problem that TM cannot solve it then my hardware (brain) must be relying on physics isomorph with the problem. Then if I solve a non-computable problem then my hardware may be relying on a non-computable physical state/process (isomorph with the problem).
What physical process is non-computable, or how would look like?  Well, physics similar to (or relying on similar events as in) tiling for aperiodic case is noncomputable. Do such non-computable physical events exist in reality? Perhaps. Rodger Penrose suggested that quantum gravity is non-computable. Classical physics is computable as far as I know.

What the “tilling problem” have to do with the ability of conscious entity to aware the reality? Perhaps nothing. However, it is a step forward in showing that our minds are more that Turning Machines and certain aspects of our minds are not representable in TM. In other words, if one try to put our minds inside a computer then mathematical proof can be given that it cannot be completely done, or, some processes from our brain will be left not-representable  inside the computer/TM.
 

(because if we know the relevant properties, we can include them in the emulation).

It is also possible that, based on the known relevant properties, one can mathematically prove that an emulation in finite time is not possible (essentially due to the limitation of the machine and its computing model). For example, one cannot emulate in a computer/TM how to solve in finite time the tiling problem for aperiodic case. To solve it (for aperiodic case and in finite time) one would have to tape into a completely different type of physics than the elementary steps of computer/TM. Our brains might do it because we can identify solutions (sometimes very quick) at least for certain simpler cases.

[flr]

Suppose we approach it from another direction, and, say, imagine a thought experiment, where we develop artificial programmable neurons that can emulate the behaviour and connectivity of any neuron in the brain (chemical sensors & effectors also allow it to respond to and modify the chemical environment around it, just like a biological neuron). So we scan someone's brain (a willing volunteer!), identifying every neuron and it's connections, and measuring how each one behaves; then we start replacing them, one by one, with the artificial electronic neurons, programmed to behave identically. Assume each replacement replicates the relevant behaviour of the original exactly.

Do you think there will come a point in this cell-by-cell replacement when the volunteer will no longer be recognisably conscious, despite there being no functional change in his brain? If so, can you explain why you think so?

With current (or near future) technology a likely outcome  is that the ‘volunteer’ will start developing dementia-like syndromes progressing as the replacement of neurons occurs.

For argument, let us assume that it is possible for artificial neurons to imitate all functions (including chemical synapses, all time transients, etc) of the natural neurons. To quantify this better, let’s assume that classical physics and macroscopically measurable properties (above thermal fluctuations sqrt(N)) are sufficient information to find all that is needed to make artificial neurons behave like natural ones, and at the end of replacement the artificial neuronal network behave measurable similar to original natural neurons. What do we get at the end of the process: an insentient machine (philosophical zombie?) or the original sentient being?

If we get a zombie then we lost some physics, likely at quantum level. Quantum mechanics (QM) is working in anything by e.g. holding atoms together, without QM atoms e- will fall into nucleus, but the question is: is QM directly involved in consciousness? As QM demonstrably play a direct direct role in biology (photosynthesis and taste), I find it hard to believe it is not involved directly in such a special and important process such the aware-ing/conscious-ing of the reality.

QM can provide speculative ground for the nature of non-computing process that might underpin (or be isomorph with) some of our thoughts [HOWEVER we don’t know if indeed our brains rely on non-computable physics because we can only solve particular cases of non-computable (for TM) problems and maybe our NN figured out a way to represent those particular cases with classical computable physics, relying on the simplicity and particularity of the solved problem; on the other hand counterarguments can easily be brought].

Speculations aside and based on what we know, direct involvement of QM can provide inside our brains two things: i) a more efficient energy transfer from point to point (due to electronic or vibrational states extended over many atoms) and ii) faster Turning Machine computations if the results of those computations somehow survive from picoseconds to tens of miliseconds in order to be interfaced with the timescales of NN classical processes – the missmatch of decoherece timescales  is a loooong shot but maybe nature found a way.

Back to your question, if the QM is directly involved in the process of consciousness (via e.g. space-extended electronic states or quantum vibrational states on certain parts of neurons) then it dramatically restrict the molecular substrate that can be used, and the artificial neuron may have to be very similar to natural one in order to reproduce the quantum states directly involved in conscious -ness (-ing).

It is conceivable to end up with a artificial NN that imitates perfectly at the classically describable level of interneuronal connections the original natural NN but it completely misses to generate some quantum states of the natural NN because the ANN does not have the right physical configuration. If those missed quantum states are essential to consciousness then ANN is a mindless machine even if at classical and interneuronal connection level imitates perfectly the natural NN

===
Abreviations:
TM = Turning Machine
QM = Quantum Mechanics
NN = Neuronal Network
ANN = artificial NN

[dlorde]

Because we can solve the tiling problem or halting problem for cases computers (i.e. Turning Machines) cannot do it.
...
Our minds can identify and solve (for simple cases) problems that computers/TM cannot solve, and we can even explain why this is the case.

I'm familiar with the halting and tiling problems. They're Decision Problems - the decidability of tiling problems mathematically depends on the decidability of a  halting problem. But they are statements about arbitrary instances, not specific instances. Both humans and computer programs can solve certain (simple) specific instances. The compiler in a decent IDE will comfortably solve halting problems for even some quite complicated code (better than a human, which is why the feature is provided), but not for all code. As humans generally use algorithmic techniques for anything more than trivially simple instances, I'm curious to see your reference for instances that humans can solve, but for which no computer algorithm can be written.

[dlorde]

With current (or near future) technology a likely outcome  is that the ‘volunteer’ will start developing dementia-like syndromes progressing as the replacement of neurons occurs.

This is a thought experiment, the assumption is that the technology will exactly mimic the original.
 

If we get a zombie then we lost some physics, likely at quantum level. Quantum mechanics (QM) is working in anything by e.g. holding atoms together, without QM atoms e- will fall into nucleus, but the question is: is QM directly involved in consciousness? As QM demonstrably play a direct direct role in biology (photosynthesis and taste), I find it hard to believe it is not involved directly in such a special and important process such the aware-ing/conscious-ing of the reality.

Why? Do all 'special and important' processes have to directly involve QM? Is it more than just a case of consciousness is mysterious and QM is mysterious, so maybe they're connected? I'm happy to accept that there may be particular neural processes where QM effects could play an optimizing role, as in other biological systems, but I don't see the rationale behind claiming it is somehow the key to consciousness.

Speculations aside and based on what we know, direct involvement of QM can provide inside our brains two things: i) a more efficient energy transfer from point to point (due to electronic or vibrational states extended over many atoms)

What 'electronic or vibrational states' in particular? How does that help cellular communication? The timescales of the neural membrane depolarisation and synaptic transmission are consistent with the overall activity that's observed, e.g. sensory input takes about 300ms of processing to reach the threshold for the wide-scale activation of the cortex and other areas, that is consistent with the start of conscious awareness - which takes a further 300-400 ms to activate areas associated with generating a response. Where does QM help?
   

ii) faster Turning Machine computations if the results of those computations somehow survive from picoseconds to tens of miliseconds in order to be interfaced with the timescales of NN classical processes – the missmatch of decoherece timescales  is a loooong shot but maybe nature found a way.

I thought the claim was that a Turing Machine couldn't emulate the non-computable functions of the brain, which was what QM was being invoked to explain? having QM speed up Turing Machine computations wouldn't help with non-computability...

..if the QM is directly involved in the process of consciousness (via e.g. space-extended electronic states or quantum vibrational states on certain parts of neurons) then it dramatically restrict the molecular substrate that can be used, and the artificial neuron may have to be very similar to natural one in order to reproduce the quantum states directly involved in conscious -ness (-ing).

That's a big, unsupported 'IF' - but the thought experiment assumes that any QM involvement can accounted for in the emulation - we know how QM behaves, so we could, in principle, duplicate its influence on our artificial system.
 

It is conceivable to end up with a artificial NN that imitates perfectly at the classically describable level of interneuronal connections the original natural NN but it completely misses to generate some quantum states of the natural NN because the ANN does not have the right physical configuration. If those missed quantum states are essential to consciousness then ANN is a mindless machine even if at classical and interneuronal connection level imitates perfectly the natural NN

But if we do generate all the relevant QM states, as the thought experiment assumes?

It seems to me that either you know of some QM effect that can't, in principle, be emulated in an artificial system, or your QM argument is not a valid objection to AI consciousness. I don't know any compelling evidence that QM effects are involved or are necessary, and all the neuroscience evidence I've seen suggests that the brain functions just as one would expect if special QM effects weren't involved.

The EU funded Human Brain Project is aiming to create a neuron emulation faithful to molecular scales, where any required QM effects could be incorporated. So far, their very limited emulations of parts of biological brains (of rats), have behaved just like their biological counterparts; they may be nowhere near complex enough to be more than proof-of-principle models, but no evidence of, or need for, QM effects has been seen.

We can't yet define precisely what we mean by consciousness, so it's not really surprising that we don't yet know how it works.

[Ethos_]

We can't yet define precisely what we mean by consciousness, so it's not really surprising that we don't yet know how it works.

I suggest that consciousness becomes possible with AI when it's calculations demonstrate opposition to the programmers aims. When AI defies instruction, a conscious rebellion becomes evident. Rebellion is the precursor to self identity and a conscious personality.

[flr]

The EU funded Human Brain Project is aiming to create a neuron emulation faithful to molecular scales... So far, their very limited emulations of parts of biological brains (of rats), have behaved just like their biological counterparts;

Their biological conterpart generated conscious states (that was its major purpose after all). Do you know if the emulation itself was conscious as well?

[flr]

I suggest that consciousness becomes possible with AI when it's calculations demonstrate opposition to the programmers aims. When AI defies instruction, a conscious rebellion becomes evident. Rebellion is the precursor to self identity and a conscious personality.

If you could only imagine how much my codes run against my wishes/aims in the debugging stages....u
You don't even know how much 'rebellion' (well, kind of) I encounter from the pieces of software I write till I get them debugged....
Jokes aside, how can I ever know is indeed opposition and not just mechanical mindless machine steps that just fitted the context ? In other words, how would I ever know the "opposition" is in machine rather than in the eye of the beholder?

I suggest that consciousness becomes possible with AI when ... demonstrate opposition to the programmers aims.

And if my aim is: "Defy my aim"?

[evan_au]

All living organisms are conscious entities

Let's probe the boundaries of this for individuals:

• I am leaning towards thinking that a single neuron is not conscious(?)
  
• I am not sure whether ants (≈250,000 neurons) or bees (≈1,000,000 neurons) are conscious - how could you prove it?
  
• Humans (≈90,000,000,000 neurons) define ourselves as conscious (without really knowing what it means)
  
• Elephants (≈250,000,000,000 neurons) have more neurons overall 
  
• The Long-Finned Pilot Whale has more neurons in the cerebral cortex (≈40,000,000,000 neurons) than humans (≈20,000,000,000 neurons)
  

I suggest that if you have a number of conscious beings which are able to communicate, as soon as you put a group of them together, you create a new conscious being with a different identity than the individual pieces - an organization.

Both bees and ants form colonies; I am not sure whether an ant nest or bee hive is conscious - how could you prove it?

• Most bees are social - a typical hive may have ≈50,000 individuals, or ≈50,000,000,000 neurons
  
• There are species of solitary bee, which seem to achieve most of the functions of hive bees. But without task specialization, they don't produce some complex structures we see in bee hives.
  
• Ants live in nests, containing ≈1,000,000 individuals, or ≈250,000,000,000 neurons
  
• I had trouble finding examples of solitary ants; most ants seem to be eusocial (most of the examples seemed to be ants which forage alone, but still live in a specialized community). One could argue that ants seem to achieve things as a group that individual ants could not accomplish alone - like moving very large pieces of food, or building bridges across gaps.
  

Perhaps ants and bees are examples of the China Brain experiment in nature? Super-organisms with a comparable number of neurons to an individual human?

But maybe humans really define our consciousness by culture - the ability to learn faster than genetic mutations, and to transmit this learned culture between generations?

[cheryl j]

It seems to me there are two separate questions in this discussion. One is - if you artificially replaced or replicated every neuron in the brain, would it be conscious? (that seems obviously true if your replication is perfect.)

The other is, does a system that duplicates the function of a conscious animal, have to be conscious as well? In biology there always seems to be more than one way to skin a cat, different ways for locomotion, different engineering for flight or sensation or reproduction. Consciousness may be our OS, but I don't see why it should be the OS of every "intelligent" system.

I must admit, I only half understood much of what dlorde and fir were debating, and I hate to sound like thick headed biologist. But if I shut down all your sensory systems, your eyes, ears, smell,touch, and propriroception, how long would your consciousness function, if you were literally a brain in a jar with all the nutrients and oxygen its cells needed?

[cheryl j]

The interesting about consciousness and qualia is this:
Some people say, well, you know, your whole experience of reality is just an artificial construct of your brain. But the converse is equally true - there is no thought, idea, memory, or imagination in the brain that is not connected to some present or past physiological sensation - a sound or image or other sensation - even our most abstract concepts. There are no symbols or meta-representations in the brain that are not basically derived from  or coded in sensory experience.

[RD]

how long would your consciousness function, if you were literally a brain in a jar with all the nutrients and oxygen your brain needed?

Without liver & kidneys attached a few days max. This dog's head only survived hours ... https://youtu.be/pQOZTpEApfA#t=4m27s

[cheryl j]

how long would your consciousness function, if you were literally a brain in a jar with all the nutrients and oxygen your brain needed?

Without liver & kidneys attached a few days max. This dog's head only survived hours ... https://youtu.be/pQOZTpEApfA#t=4m27s

Well, yes. I'm assuming waste management as well. My point is, whether consciousness would exist without the body and other biological attributes that may not be present in other systems.

[dlorde]

The EU funded Human Brain Project is aiming to create a neuron emulation faithful to molecular scales... So far, their very limited emulations of parts of biological brains (of rats), have behaved just like their biological counterparts;

Their biological conterpart generated conscious states (that was its major purpose after all). Do you know if the emulation itself was conscious as well?

Good grief, no; this was an early proof-of-concept, a relatively tiny emulation (31,000 virtual brain cells connected by roughly 37 million synapses) of a cortical column in the rat brain somatosensory cortex - see this public resource. It didn't include glial cells, and had no plasticity. These will come later. The videos of spontaneous activity in the model when given different calcium ion concentrations are particularly interesting.

A neuron-by-neuron reconstruction requires tracing the type and connectivity of every neuron & synapse in the piece of brain being studied. This is a monumental task - by contrast, it makes generating the computer model look easy.

[dlorde]

The other is, does a system that duplicates the function of a conscious animal, have to be conscious as well? In biology there always seems to be more than one way to skin a cat, different ways for locomotion, different engineering for flight or sensation or reproduction. Consciousness may be our OS, but I don't see why it should be the OS of every "intelligent" system.

I agree - it depends what functions of a conscious animal you want to duplicate. If you wanted to duplicate the consciousness of the animal, you'd have to duplicate a lot of other mental functions, because it appears that consciousness efficiently delegates as much of the 'grunt work' as it can to the fast, parallel, 'subroutine' processes that manage most of our lives. But if you wanted just those functional subroutines, you wouldn't need consciousness (unless you wanted them to intelligently coordinate their activities).

... if I shut down all your sensory systems, your eyes, ears, smell,touch, and propriroception, how long would your consciousness function, if you were literally a brain in a jar with all the nutrients and oxygen its cells needed?

I expect it would function for quite a while - people have spent hours in sensory deprivation flotation tanks without harm - although the lack of input often causes circuit noise and spontaneous activity in the CNS to be boosted into vivid hallucinations of all kinds. I couldn't even guess how long someone could endure total sensory deprivation and stay conscious and sane - I suspect it would depend a lot on the individual involved.

[flr]

A neuron-by-neuron reconstruction requires tracing the type and connectivity of every neuron & synapse in the piece of brain being studied. This is a monumental task - by contrast, it makes generating the computer model look easy.

It is possible that at the level of inter-neuronal connection the simulation became more and more accurate.

A key question is: what is the point of that simulation if the simulation itself is not a conscious state?
If the simulation itself is not a conscious state, isn't then the case that the simulation missed the main point and purpose of the real thing: to generate/supervene/trigger aware states?

If an accurate simulation of the brain function do not generate conscious states (like the natural one) isn't this a proof that conscious needs more that those patterns and changes in patterns?

[dlorde]

It is possible that at the level of inter-neuronal connection the simulation became more and more accurate.

I can't make sense of that - what do you mean?

A key question is: what is the point of that simulation if the simulation itself is not a conscious state?
If the simulation itself is not a conscious state, isn't then the case that the simulation missed the main point and purpose of the real thing: to generate/supervene/trigger aware states?

The project is not aiming to simulate consciousness, it's aiming to emulate the brain for medical reasons - drug response & interaction, effects of physical and chemical damage, genetic conditions, etc. If they eventually achieve a fully active emulation of a whole brain, they may get evidence of some form of consciousness, but that's a decade or two in the future, and they'll probably be working mainly with brain subsystems.

Don't forget that the acquisition of the brain data for building the full model (scanning, tracing, etc.) will probably take several years, and may use a number of brains, so it will probably not be a single coherent snapshot of a single brain, so is unlikely to have a single coherent set of memories, sense of self, and, perhaps, consciousness - add to that the possibility that it may not be possible to emulate some memories if that data is stored at a molecular level (which I doubt, but has been suggested as a possibility). Also, the emulation will not have been through the development and experiences of a biological brain, which may be relevant, who knows...

By the time a full emulation is ready, there will be ethical considerations which may well prohibit the level and breadth of activation that could result in consciousness (e.g. in case it could suffer or die, etc). I would expect a lot of noisy debate about this, from ethical campaigners and medical researchers concerned with consciousness.

If an accurate simulation of the brain function do not generate conscious states (like the natural one) isn't this a proof that conscious needs more that those patterns and changes in patterns?

What patterns? It would (obviously) be a proof that it isn't an accurate emulation.