The Hard Problem of Gravity

[Ichneumonwasp]

But if consciousness is produced by analogue physical actions, then it cannot be emulated digitally. Simulating the release of hormones will not emulate the action.

Then you misunderstand. There is nothing inherent to the chemicals involved that makes the system not emulatable. That's just the way the system evolved. They still result in changes in firing rates, which is digital information. There are lots of inputs with varying degrees of impact, some long-term, some short-term; that is all.

 

[yy2bggggs]

Generally an inability to solve an easy problem indicates an inability to solve the harder problems.

Generally? Is there another similar class of problems that has worked out this way?

ETA: You're also confusing the HPC with the problem of creating conscious entities, for which I was more serious about mommy and daddy being able to create such a thing than you may have initially thought.

 

[cyborg]

I already indicated that producing jokes was an example of selecting particular patterns from a finite set. It is an easy problem, which has not been solved.

Humour is relative - you cannot 'solve' it, you can only 'solve' it with respect to some given cultural benchmark of what humour is.

 

[AkuManiMani]

But don't you see that's another turtle ? "The same way that we know anything else exists" means that EVERYTHING is experienced, INCLUDING qualia.

Quale is just a name for experiences. The world exists whether or not we experience qualia. Of course, as with anything else, one can have an endless regression explanation but this does not negate the reality of things IAOI.

So now the question becomes: do we experience experience ? Where do we stop ? No. Qualia is just a label, but it doesn't represent anything real.

Define what you mean by 'real'. When I use the term I mean any actual phenomena. We experience sense impressions consciously and qualia is a label for that experience. If you don't like the term, feel free to create another synonymous word to label what is being discussed but there is not question that conscious experience [whatever one chooses to call it] is an actual phenomenon. The point is to gain a scientific understanding of it -- which is one of the most central concerns of neuroscience. In the field of neuroscience, there are numerous hypothetical models of what consciousness may be but there is nothing definite yet. In the mean time, its more than a little silly to proclaim that its already been explained while claiming it doesn't exist.

The "experience" of seeing a cat is simply a physical extention of the cat itself. There is no "image".

Hmm...thats an interesting interpretation. If you think about it every entity is an extension of all others -- the universe, and everything in it, is singular.

The problem is, in order to have any meaningful language, one must have some categorical system of distinguishing between things. Its important to distinguish between the perception of a thing and the actual thing being perceived. Even assuming the position of idealism it is still necessary to distinguish between subject, object, and the relations between them.

No, no. You are assuming that qualia exist, and then trying to understand them.

I'm not 'assuming' qualia exist; I'm directly observing a phenomenon and choosing to label it qualia. Its not much different than 19th century biologists labeling hereditary elements 'genes' before they scientifically understood them or current thinkers labeling cultural elements 'memes'.

 

[westprog]

Yep. I don't know how to represent that in a computer system, but you could put what I know about computers in a paper bag. A small paper bag.

Well, here's my viewpoint.

Digital systems are a concept which we apply to describe existing analogue systems. To consider a system as digital, we ignore most of the information that passes back and forth, and consider certain conglomerations of information as if they were single binary values.

The advantage of digital systems is that they are predictable, and can be easily treated mathematically. It's also the case that you can create totally different digital systems which will give exactly the same digital output. For digital systems, the implementation makes no difference to the output. If the digital arrangement is the same, the digital output will be the same.

But we must always remember that this is an abstraction. The actual operation of different digital systems is different, as is the actual output. It is only the digital interpretation of the output that remains constant.

It is possible to exactly duplicate the digital aspects of a digital system. However, it is never possible to duplicate an analogue system exactly - and all real-world systems are analogue systems.

Thus it is of utmost importance whether we are considering the digital or analogue operation of a system. A digital emulation of an analogue system won't do the same thing at all, in any sense. All it might do is predict what such a system might do, under specific circumstances.

So it's hugely relevant whether the operation of the human mind is an analogue or digital process. If it's an analogue process, then it's highly unlikely that its effects can be duplicated by an entirely different system that has a similar structure.

Question for Westprog: Do you think that emotion/motivation states/feelings consist in an ontologically different category from other mental actions, such as calculation ability?

To me they are simply different types of problem solving. Emotions and feelings deal with different types of problems than math does, but they still are there to solve problems.

That may be so, but it's the experience of having emotions that interests me. Emotions feel like something. I don't know why.

 

[PixyMisa]

But if consciousness is produced by analogue physical actions, then it cannot be emulated digitally.

Wrong. Not only can it be simulated digitally, but given the nature of our Universe, it can be simulated digitally to a degree of precision where it is indistinguishable in principle from the real thing.

Simulating the release of hormones will not emulate the action.

Wrong, wrong, wrong, hopelessly wrong. Scientists and engineers - and these days, preschoolers - use that sort of simulation every day.

I already indicated that producing jokes was an example of selecting particular patterns from a finite set. It is an easy problem, which has not been solved.

Wrong. Doubly wrong, in fact. It's a hard problem, with a great deal of complexity and subtlety to it - natural language, psychology, social cues. The complete opposite of consciousness, which is a simple mechanical process.

And it's been done. No quack.

Generally an inability to solve an easy problem indicates an inability to solve the harder problems.

What does consistently mistaking easy problems for hard ones and vice versa indicate?

 

[PixyMisa]

Well, here's my viewpoint.

Digital systems are a concept which we apply to describe existing analogue systems. To consider a system as digital, we ignore most of the information that passes back and forth, and consider certain conglomerations of information as if they were single binary values.

Doesn't need to be binary, and you're switching back and forth between two different (but related) definitions of information, but yeah.

The advantage of digital systems is that they are predictable, and can be easily treated mathematically.

So can analog systems. I mean, personally I prefer algebra to calculus, but that doesn't mean calculus doesn't exist.

It's also the case that you can create totally different digital systems which will give exactly the same digital output. For digital systems, the implementation makes no difference to the output. If the digital arrangement is the same, the digital output will be the same.

Information is substrate-neutral, yes.

But we must always remember that this is an abstraction. The actual operation of different digital systems is different, as is the actual output. It is only the digital interpretation of the output that remains constant.

Actually, no. The physical output can be identical too. I mean identical. As in, not distinguishable, even in principle.

It is possible to exactly duplicate the digital aspects of a digital system.

Sure.

However, it is never possible to duplicate an analogue system exactly - and all real-world systems are analogue systems.

No. All real-world systems are really complicated digital systems. Or, in one case, a reasonably simple one: The hydrogen atom.

Thus it is of utmost importance whether we are considering the digital or analogue operation of a system.

No it isn't.

A digital emulation of an analogue system won't do the same thing at all, in any sense.

A simulation will act like the simulated system to whatever degree of precision and accuracy the simulation simulates the simulated system. Hence the demand for great big computers.

All it might do is predict what such a system might do, under specific circumstances.

I see. And how does it make this prediction, without acting like the system?

So it's hugely relevant whether the operation of the human mind is an analogue or digital process.

This is triply wrong, which may be a first. First, it doesn't ultimately matter for anything. Second, everything is ultimately digital. And third, when it comes to information-processing systems, which is what the brain is, the substrate is immediately irrelevant and not just ultimately so.

If it's an analogue process, then it's highly unlikely that its effects can be duplicated by an entirely different system that has a similar structure.

Right. Tell that to all the CFD people.

That may be so, but it's the experience of having emotions that interests me. Emotions feel like something. I don't know why.

I do! I do!

 

[rocketdodger]

So it's hugely relevant whether the operation of the human mind is an analogue or digital process. If it's an analogue process, then it's highly unlikely that its effects can be duplicated by an entirely different system that has a similar structure.

This is merely an assumption. You have no evidence to support this assumption.

Evidence to support this assumption would be:

Data that suggests the function of brain components relies upon an infinite level of precision.

Evidence that we have actually gathered:

Data that suggests the function of brain components relies upon a finite level of precision.

If you shift the molecules of a neuron ever so slightly, the behavior of the neuron will stay the same. We know this, because we know molecules are in constant motion anyway. If you shift a neuron around ever so slightly, the behavior of the neural network will stay the same. We know this, because we know neurons are in constant motion anyway. And so on and so forth.

There isn't anything to suggest that consciousness requires infinite precision. That means there isn't anything to suggest that consciousness requires an analog substrate.

 

[Ichneumonwasp]

That may be so, but it's the experience of having emotions that interests me. Emotions feel like something. I don't know why.

I'll get to the rest as I have time since I don't have time to read through it all right now, but this caught my eye.

I would say that they have to feel like something or they would be nothing. To think about this issue, we need two bits of info: (1) nature only works with what its got; and (2) we must ask ourselves what an emotion or a feeling *is* or what it does.

From an evolutionary perspective, emotions and feelings came before language and complex calculation ability, so we are stuck with old ways of solving problems. Emotions and feelings are ways of solving social and environmental problems (mostly emotions) and ways of providing value to what goes on out there. Emotions motivate us toward certain actions. For them to do so they must take some form -- that is why we feel them. That is simply the form they take. Same with feelings, but feelings are much more complex cognitive processes than simple emotions.

They feel like something because that *is* the motivation and orientation toward action that they do.

I would imagine that we could recreate the same thing in a computer system, but since emotion and feeling are only "movements toward some action" or action tendencies, the resulting computer would not work the way that we want computers to work. Instead of doing precisely what we tell it to do, it would act in a less predictable fashion. That is how we are set up -- with behavioral tendencies rather than set programs. Consciousness is the process of sifting through those behavioral tendencies.

 

[westprog]

Then you misunderstand. There is nothing inherent to the chemicals involved that makes the system not emulatable. That's just the way the system evolved. They still result in changes in firing rates, which is digital information. There are lots of inputs with varying degrees of impact, some long-term, some short-term; that is all.

How can a change in a rate of release be digital information? It might be something that can be approximated digitally, but it is unquestionably not digital.

 

[westprog]

This is merely an assumption. You have no evidence to support this assumption.

Evidence to support this assumption would be:

Data that suggests the function of brain components relies upon an infinite level of precision.

Evidence that we have actually gathered:

Data that suggests the function of brain components relies upon a finite level of precision.

If you shift the molecules of a neuron ever so slightly, the behavior of the neuron will stay the same. We know this, because we know molecules are in constant motion anyway. If you shift a neuron around ever so slightly, the behavior of the neural network will stay the same. We know this, because we know neurons are in constant motion anyway. And so on and so forth.

There isn't anything to suggest that consciousness requires infinite precision. That means there isn't anything to suggest that consciousness requires an analog substrate.

This misses the point. An analogue system is a fundamentally different real-world thing to a digital system. It doesn't matter what precision the neurons need, if it is the actual physical operation of the neurons which creates consciousness.

I find is strange that it is supposed proponents of materialism that attempt to refute the possibility that consciousness has a physical component.

 

[Ichneumonwasp]

Well, here's my viewpoint.

Digital systems are a concept which we apply to describe existing analogue systems. To consider a system as digital, we ignore most of the information that passes back and forth, and consider certain conglomerations of information as if they were single binary values.

The advantage of digital systems is that they are predictable, and can be easily treated mathematically. It's also the case that you can create totally different digital systems which will give exactly the same digital output. For digital systems, the implementation makes no difference to the output. If the digital arrangement is the same, the digital output will be the same.

But we must always remember that this is an abstraction. The actual operation of different digital systems is different, as is the actual output. It is only the digital interpretation of the output that remains constant.

It is possible to exactly duplicate the digital aspects of a digital system. However, it is never possible to duplicate an analogue system exactly - and all real-world systems are analogue systems.

Thus it is of utmost importance whether we are considering the digital or analogue operation of a system. A digital emulation of an analogue system won't do the same thing at all, in any sense. All it might do is predict what such a system might do, under specific circumstances.

So it's hugely relevant whether the operation of the human mind is an analogue or digital process. If it's an analogue process, then it's highly unlikely that its effects can be duplicated by an entirely different system that has a similar structure.

I'm afraid that I don't quite understand. First, computer systems are not being proposed as human beings, so what difference does it make that analogue and digital are simply two ways of describing a system.

None of us ultimately know what is going on at a deep level, so the analogue description is just as much an approximation as a digital one is. If we construct a digital representation of the process of consciousness, how have we not understood what consciousness is?

It seems as though you are veering close to an elan vitale argument here -- that there is just some 'something' in living systems that we can't put our finger on.

That argument died in the 19th century. Granted, it might be the case, but I tend to doubt it.

Perhaps one approach to this issue would be to get a better handle on what we mean when we use words like "understand" or "meaning". I suspect that when we examine them we will find that their "meaning" amounts to mapping some idea of intentionality onto computable processes. Emotions/feelings have that weirdness to them because they provide intentionality -- they are oriented toward the world in a way that multiplication is not.

 

[Ichneumonwasp]

How can a change in a rate of release be digital information? It might be something that can be approximated digitally, but it is unquestionably not digital.

Is not digital information simply "yes"/"no" and time coding? That's essentially what neurons do.

ETA:
Oh, forgot to add -- I thought neural networks largley consisted of nodes with varying connection strengths. Learning in the nervous system seems to consist of variations in connection 'strength', variation in the number of connections between cells, and variations in cell survival depending in large part on whether or not anything communicates with a cell.

Can't all of that be done digitally?

 

[westprog]

Doesn't need to be binary, and you're switching back and forth between two different (but related) definitions of information, but yeah.

So can analog systems. I mean, personally I prefer algebra to calculus, but that doesn't mean calculus doesn't exist.

Information is substrate-neutral, yes.

Actually, no. The physical output can be identical too. I mean identical. As in, not distinguishable, even in principle.

Sure.

No. All real-world systems are really complicated digital systems. Or, in one case, a reasonably simple one: The hydrogen atom.

No it isn't.

A simulation will act like the simulated system to whatever degree of precision and accuracy the simulation simulates the simulated system. Hence the demand for great big computers.

I see. And how does it make this prediction, without acting like the system?

We've been through this over and over again. No digital system is able to emulate an analogue system in any way. All it can do is predict certain behaviour of the analogue system.

Even given unlimited resources, all a simulation of an earthquake, or a car, or any other real world object can do is predict what the object will do. It cannot replace the object. You can't drive a digital simulation of a car to work no matter how good it is.

So, if the operation of the mind is essentially analogue, it can't be replaced by a digital system. Now, it's possible that the essential element of the mind is digital- but that is far from certain. Certainly its operation is not, on inspection, obviously digital.

This is triply wrong, which may be a first. First, it doesn't ultimately matter for anything. Second, everything is ultimately digital.

There might be some theories according to which nature is ultimately digital, but they are very far from certain.

At the level of a hydrogen atom, there might be discreet states to which the electron belongs. But the actual position of the electron is inherently uncertain if we attempt to measure its momentum at the same time. It does not look like a digital universe close up.

And third, when it comes to information-processing systems, which is what the brain is, the substrate is immediately irrelevant and not just ultimately so.

There are plenty of analogue information processing systems which rely on the precise substrate in order to operate. That's why the substrate is there. A car takes information from the driver, adjusts its behaviour, and displays information for the driver to read. It can't be replaced by a digital system - even in principle.

Right. Tell that to all the CFD people.

I do! I do!

 

[westprog]

So can analog systems. I mean, personally I prefer algebra to calculus, but that doesn't mean calculus doesn't exist.

Digital systems can be predicted exactly. Analogue systems can only be approximated. Indeed, some simple analogue systems can't be predicted at all. I'm not sure if all digital systems are predictable in principle, but I suspect they are.

 

[westprog]

Then you misunderstand. There is nothing inherent to the chemicals involved that makes the system not emulatable.

I didn't say that the system couldn't be emulated, I just said that if it's analogue, then it can't be digitally emulated. It can be digitally simulated.

 

[westprog]

Is not digital information simply "yes"/"no" and time coding? That's essentially what neurons do.

No. Digital information is sequential. It isn't time coded. If something is timing dependent, it isn't digital.

ETA:
Oh, forgot to add -- I thought neural networks largley consisted of nodes with varying connection strengths. Learning in the nervous system seems to consist of variations in connection 'strength', variation in the number of connections between cells, and variations in cell survival depending in large part on whether or not anything communicates with a cell.

Can't all of that be done digitally?

It all depends what the significant factors are. Pixy has made up his mind that only the digital factors are significant. IMO we know far too little about the operation of the mind to be sure of any such thing.

 

[Ichneumonwasp]

No. Digital information is sequential. It isn't time coded. If something is timing dependent, it isn't digital.



It all depends what the significant factors are. Pixy has made up his mind that only the digital factors are significant. IMO we know far too little about the operation of the mind to be sure of any such thing.

OK, my mistake. I don't think we know enough either, but I still don't see what the big deal is. Whether or not we are discussing a mere simulation, that still serves the same purpose -- which is to produce a scientific account of consciousness. A scientific description of running isn't running. If we can simulate consciuosness, then we have our scientific understanding of it that can be represented mathematically.

Isn't that what we are after?


ETA:

Or to head things off at the pass, we can define consciousness by "what it does" and reach some sort of answer here.

If we define it by "what it is" then, by definition, it's in a different ontological category and we've entered the neverland of dualism.

If we define it any way but functionally there is no sense in playing the game.

 

[rocketdodger]

This misses the point. An analogue system is a fundamentally different real-world thing to a digital system. It doesn't matter what precision the neurons need, if it is the actual physical operation of the neurons which creates consciousness.

I find is strange that it is supposed proponents of materialism that attempt to refute the possibility that consciousness has a physical component.

The fundamental tenant of modern materialism is that the only thing which exists are properties.

It is mathematically impossible to show otherwise, since anything you observe about an entity amounts to some kind of property.

So a neuron is a set of properties. You can't really dispute this.

Furthermore, all properties can be emulated / simulated.

You don't quite understand how this could be done, because you still haven't wrapped your brain around the fact that what you call "physical" properties are relative.

The interactions between particles are relative. The interactions between you and your car are relative. The interactions between the neurons in a brain are relative. Location, velocity, momentum, time, even energy -- all relative properties.

When a relative property is simulated, all relevant sides are included. If not, then the property is not being simulated. Plain and simple.

Obviously, when a car is simulated and you are not, you are not included in the simulation. Obviously, the simulated car and the real you can't interact -- the physical properties of the simulated car are not relative to you anymore.
Obviously, if you were simulated as well, then you could interact with the simulated car.

If you disagree with any of this, then feel free to mathematically prove that we do not live in a simulation right now. Or, simply tell me how it might even be possible to figure out whether we are living in a simulation or not.

 

[rocketdodger]

Digital systems can be predicted exactly. Analogue systems can only be approximated. Indeed, some simple analogue systems can't be predicted at all. I'm not sure if all digital systems are predictable in principle, but I suspect they are.

Wrong.

Analog systems can be predicted exactly. That is the whole point of using rational and real numbers instead of just decimal approximations.

I pour out 1/3 of a cup. How much is left? 2/3 of a cup. Is that not a perfect prediction?