Singularity

[Soapy Sam]

Yes it does. If your model is accurate, and it doesn't produce the same results, then the system works by magic.-PixyMisa

You've lost me. Why should a model, however accurate, produce "the same results" as the thing modelled? That's not what models do, it's what copies do.
There are any number of differences between a model and the thing modelled. To pluck one out of the sky, there's scaling. Would it be "magic" if it turned out that all biological intelligences are between 2 and 2500mm in diameter? Hardly- it might simply imply a signal limitation of organic brains.
A silicon brain the same size might not produce similar output. One 5 miles across might not either.

Most of the output of a brain has nothing to do with conscious intelligence anyway. Why bother modelling it?

I agree it's a whiz project and all, but- I used to have an HP41C simulator program for a pc. If I put the same data into the program and the calculator, I got the same data from both. But it wasn't the same experience. It was easier to use an Excel sheet. What made the CV a great calculator was its portability. Simulating a brain in silicon strikes me as just a futile gimmick. If it worked, what have you got?
You just lost $300 billion worth of hardware and got yourself a moron.

(By extension, Melendwyr, while I agree you probably could model the ant's neural hardware, an ant brain deprived of the chemical communications and social context that makes an ant an ant rather illustrates my point about the grander human simulator. Even if it works, what use is an ant that other ants don't recognise as an ant?)

 

[kuroyume0161]

This is where Rodney Brooks has the basic concept correct. You can't just model 'a floating brain in a vat' and get a human being (or an ant). The brain isn't a disjoint object, but part of a vast complex of interfaces, chemicals, and biological systems - which are interdependent.

That said, I think the draw to such modelings is that there is a mutual drive between the actual study (biological sciences) and the modeling (AI sciences) which support each other in making discoveries and filling in knowledge. Though, I would contend that most of the work is being done by the former while the latter reaps the benefits (or sows the sorrow).

Soapy Sam makes a good point without bringing up the "magic" word: Chaos. As he states, modeling a complex system will not (as in NEVER) produce identical results. Please read up on Meteorology, turbulence, and Chaos Theory to see why. This is why weather prediction, for all purposes, sucks. They can say, "It's gonna rain in about five minutes", but much beyond that the probabilities drop significantly towards unpredictable. Why? Because the only model that could possibly match the real thing would need to be as accurate as the real thing (analog and taking into consideration the entire influence of the universe - as in all of the energy and matter impacting the Earth as well as the same on Earth). That...is...impossible.

 

[PixyMisa]

You've lost me. Why should a model, however accurate, produce "the same results" as the thing modelled? That's not what models do, it's what copies do.

No, it's what models do.

An accurate model of a steam engine is a steam engine.

If you accurately model the brain, you have a brain - expressed in this case in chips and wires and data, but still a brain.

If it's not a brain, it's not an accurate model.

There are any number of differences between a model and the thing modelled. To pluck one out of the sky, there's scaling. Would it be "magic" if it turned out that all biological intelligences are between 2 and 2500mm in diameter? Hardly- it might simply imply a signal limitation of organic brains.
A silicon brain the same size might not produce similar output. One 5 miles across might not either.

That's completely beside the point. If your model doesn't properly represent the signal limitations, it's not an accurate model. And if it is an accurate model, then it works just like a biological brain.

(By extension, Melendwyr, while I agree you probably could model the ant's neural hardware, an ant brain deprived of the chemical communications and social context that makes an ant an ant rather illustrates my point about the grander human simulator. Even if it works, what use is an ant that other ants don't recognise as an ant?)

That's nonsense.

We can recognise people as people just by talking to them on a web forum. The modelled brain can certainly do that, which illustrates that your point is wrong.

 

[PixyMisa]

Soapy Sam makes a good point without bringing up the "magic" word: Chaos. As he states, modeling a complex system will not (as in NEVER) produce identical results. Please read up on Meteorology, turbulence, and Chaos Theory to see why. This is why weather prediction, for all purposes, sucks. They can say, "It's gonna rain in about five minutes", but much beyond that the probabilities drop significantly towards unpredictable. Why? Because the only model that could possibly match the real thing would need to be as accurate as the real thing (analog and taking into consideration the entire influence of the universe - as in all of the energy and matter impacting the Earth as well as the same on Earth). That...is...impossible.

Completely irrelevant.

We're not trying to build a predictive model, we're trying to build a working model. If both the original and the model are chaotic, that's fine - it's exactly what we'd expect.

 

[Melendwyr]

That doesn't involve any magical essence just the fact that different things arranged in different ways have different properties.

Different things arranged in different ways can have identical properties, if the properties in question are specific and limited.

Emulation, Darat. A computational system that accurately represents the meaningful patterns of information in a brain is a mind, just as much as the behavior of that brain is a mind. That's what emulation is all about!

 

[Soapy Sam]

An accurate model of a steam engine is a steam engine.

We are evidently using "model" to mean different things. A model , to me, is a simplification, an abstraction of the essentials of an object or process. A model of the Titanic would not be a copy of the Titanic; a model of a steam engine might be a steam engine. It might be a bit of wood, with "Steam Engine" painted on it.

If you accurately model the brain, you have a brain - expressed in this case in chips and wires and data, but still a brain.

If it's not a brain, it's not an accurate model.

Well- how accurate must it be? How do we know if it's a brain? Ask it?
From context you seem to mean by "model" something which emulates the activity of the brain. As brains do not have detailed self analytical ability at all levels, this would imply that an electronic brain , like a human brain (if a good and accurate model), would also have limited internal understanding of it's own operations, so it would be difficult to tell if it was a brain by asking it, just as it is hard for a human to be sure he is a brain. (This might get positively theological).

On the other hand, if the model brain was equipped with internal sensors so it could understand it's own states, it would not be an accurate model.

If your model doesn't properly represent the signal limitations, it's not an accurate model. And if it is an accurate model, then it works just like a biological brain.

What do you mean by "it works just like a biological brain"? Clearly it works very differently from a biological brain. A bullet can kill. A knife can kill. At some levels (loss of blood, shock, organ failure) they work rather like each other, but they don't work "just like each other". Two processes can have similar results yet work by quite different methods. That after all is what you are assuming.
If I add 2 and 2, I get the same answer a calculator does. So does it model my brain when it adds? Or do I model it? Clearly, neither. Two unrelated processes have similar inputs and outputs. That's all.


We can recognise people as people just by talking to them on a web forum. The modelled brain can certainly do that, which illustrates that your point is wrong.

(My emphasis).
Sorry- are you referring to an actual modelled brain currently in existence? Or is this speculation?
(I have no difficulty imagining a machine that might pass an online Turing Test. I do have difficulty imagining a machine imagining a machine that could). (Could it fool another machine, I wonder?)

Nah, sorry Pixy, I've seen nothing in computers to date which leads me to think we can reproduce the output of a human brain by using a large number of them, no matter how fast. I don't say computers can't develop intelligence or awareness, but I think if and when it happens it will be so utterly unlike the human variety as to be probably unrecognisable until it's too late. The clear anti-human malevolence of the global telephone system seems like the clearest evidence for machine consciousness to date.

 

[PixyMisa]

Well- how accurate must it be? How do we know if it's a brain? Ask it?

From all available evidence, brain function arises at the cellular level. Read the thread. We are talking about a full cell-level simulation.

From context you seem to mean by "model" something which emulates the activity of the brain.

No, not emulates, simulates. To emulate is to model the behaviour of a system; to simulate is to model the operation of a system. Important difference.

As brains do not have detailed self analytical ability at all levels, this would imply that an electronic brain , like a human brain (if a good and accurate model), would also have limited internal understanding of it's own operations, so it would be difficult to tell if it was a brain by asking it, just as it is hard for a human to be sure he is a brain.

So?

The simulated brain functions as a brain. You just pointed out that it would have the same limitations as a biological brain, which is exactly what we would expect.

On the other hand, if the model brain was equipped with internal sensors so it could understand it's own states, it would not be an accurate model.

True.

However, if the model was equipped with internal sensors (software, really) so that others could understand its states, it would still be an accurate model, and a very useful one.

What do you mean by "it works just like a biological brain"?

Operationally. What it does is the same as what a biological brain does.

A bullet can kill. A knife can kill.

And simulating the cellular damage caused by either could kill the simulation. No difference.

Two processes can have similar results yet work by quite different methods. That after all is what you are assuming.

No, that's not what I'm assuming at all.

What I'm assuming is that there's no magic. So if you simulate the function of all the individual cells of the brain, what you get is a working simulation of the brain, that processes information just like a biological brain and exhibits the same emergent properties.

And there is no reason, no scientific evidence at all, to doubt this.

Sorry- are you referring to an actual modelled brain currently in existence? Or is this speculation?

It's speculation!! Read the thread!

Nah, sorry Pixy, I've seen nothing in computers to date which leads me to think we can reproduce the output of a human brain by using a large number of them, no matter how fast.

And I have. Alan Turing also had thoughts on the matter. So does Douglas Hofstadter.

Not only that, but I calculated the requirements and the cost right here in this thread. To suggest that it's not possible is simply to claim that the brain works by magic.

I don't say computers can't develop intelligence or awareness, but I think if and when it happens it will be so utterly unlike the human variety as to be probably unrecognisable until it's too late.

That's irrelevant. What we are discussing here is a cell-level simulation of the human brain. That will have a human awareness.

 

[PixyMisa]

A model , to me, is a simplification, an abstraction of the essentials of an object or process.

Yes.

Which is the whole point.

We are not engaging in a quantum-mechanical simulation, we are engaging in a cellular simulation. That's fine, because all evidence points to brain function arising at the cellular level. (Penrose, as I mentioned before, is a nut.)

A model of the Titanic would not be a copy of the Titanic; a model of a steam engine might be a steam engine. It might be a bit of wood, with "Steam Engine" painted on it.

Read. The. Thread.

I have stated, specifically, the level of detail of the model. More than once.

A working model of a steam engine is a steam engine. If it's a bit of wood with "Steam Engine" painted on it it doesn't model anything at all.

Yes, I inserted the word "working" there, but that was explicit in the original definition of the model of the brain. It's not a bit of wood with the word "Brain" painted on it, it's a functional cell-level simulation of a brain.

 

[PixyMisa]

By extension, Melendwyr, while I agree you probably could model the ant's neural hardware

Okay, you agree that we could build a working model of an ant's brain?

Fine.

Remember Pixy's First Principle: There's no magic.

The only differences between an ant's brain and a human brain are the physical differences: the nature and number of the cells and the complexity of their interconnections.

So if a human brain is, say, ten million times as complex as an ant brain, all we need to build a working model of a human brain is a computer that is ten million times more powerful.

And that's all.

Assuming we know how the cells function and how they are interconnected - physically - we don't need to understand how the brain itself operates. We are not modelling the operation of the brain as a whole, we are modelling the function of all its components. Once we have built the brain up out of correctly modelled cells, the operation of the whole thing follows naturally.

 

[kuroyume0161]

No, it's what models do.

An accurate model of a steam engine is a steam engine.

If you accurately model the brain, you have a brain - expressed in this case in chips and wires and data, but still a brain.

If it's not a brain, it's not an accurate model.

1. If you are referring to a dynamical simulation of a steam engine in, say, a computer, then it is not a steam engine. It is a simulation of a steam engine using the most basic principles required. Steam engines, compared to brains, are simplistically and idiotically simple constructs by comparison. Bad analogy.

2. To build an 'accurate model' of a brain would be to be a brain (a biological organ consisting of neurons and so on). A mechanical simulation would not be a brain, it would be a simulation. And since brains are vastly more complex than steam engines, I'd say that simulating one accurately will take another several centuries (if not longer).

3. To your point on my statement: It is completely relevant. You want a model that simulates a human brain. Without predictability at some stage, there is no comparative analysis between them. If you wire a 100 billion pseudo-neurons together, you don't have a human brain simulation. You have a 100 billion neuron simulation. You must be able to make predictions upon the structures of the pseudo-brain that correlate to a real one in order to show simulation accuracy (or are you going to judge it on superficial response alone?). The real brain is not a random or chaotic system, but it is complex enough that in modeling it, chaos and randomness may crop up (almost certainly, as a matter of fact). Unless you are using specially tailored (and awfully slow) arbitrary precision floating point mathematical calculations, you are going to run into the time-tried and well-known wall of limited precision floating point numbers on computers as well as precision loss (roundoff and truncation) if not countered by an almost AI-level monitoring and restoration system to go along with it. You cannot circumvent such limitations easily.

ETA: And there is still the 'halting problem' with which to contend. Your multi-billion dollar brain may end up in an infinite loop (i.e.: bloody mad!). These are NP-hard problems - not just technological hurdles...

 

[PixyMisa]

1. If you are referring to a dynamical simulation of a steam engine in, say, a computer, then it is not a steam engine.

I wasn't, actually. I was talking about working models of steam engines.

And as far as the simulation is concerned, it is a steam engine. If we analyse the work it can do, the answer is the same as for a physical steam engine. Of course, it's not a physical steam engine, so the work is simulated work, not physical work.

It is a simulation of a steam engine using the most basic principles required

Yes.

Steam engines, compared to brains, are simplistically and idiotically simple constructs by comparison. Bad analogy.

It's not my analogy. However, in its badness it serves a very useful purpose.

We distinguish between the work done by a simulated steam engine and the work done by a physical steam engine on precisely the basis of physicality.

However, the work done by a brain is information processing, and there is no difference whatsoever whether this is done by a physical brain or a simulated one.

2. To build an 'accurate model' of a brain would be to be a brain (a biological organ consisting of neurons and so on).

Baloney.

An accurate model is an accurate model. We have, quite specifically, modelled the brain at the biological level - read the thread; that's what we are talking about.

A mechanical simulation would not be a brain, it would be a simulation.

As far as the simulation is concerned, it is a brain. Indeed, as far as the simulation is concerned, it is a biological brain.

And since brains are vastly more complex than steam engines I'd say that simulating one accurately will take another several centuries (if not longer).

And you have a reason for this? I've already shown that we could build a sufficiently powerful computer today.

3. To your point on my statement: It is completely relevant. You want a model that simulates a human brain.

Yes.

Without predictability at some stage, there is no comparative analysis between them.

Entirely false.

The human brain is a chaotic system. The simulated brain will be a chaotic system. If the simulated brain wasn't chaotic, the entire project would be a failure.

Furthermore, we don't need predictibility, because this is not a predictive model. It is a working model, and we judge its success or failure by how well it works.

If you wire a 100 billion pseudo-neurons together, you don't have a human brain simulation.

If you wire them together as they are in the human brain, a human brain simulation is exactly what you have. It is impossible for it to be anything else.

You have a 100 billion neuron simulation. You must be able to make predictions upon the structures of the pseudo-brain that correlate to a real one in order to show simulation accuracy (or are you going to judge it on superficial response alone?).

Nonsense.

IT IS NOT A PREDICTIVE MODEL.

NO-ONE EVER SAID IT WAS GOING TO BE A PREDICTIVE MODEL.

IT IS A WORKING MODEL.

The real brain is not a random or chaotic system

Randomness cannot be eliminated, but it is probably not significant.

However, the brain is most definitely a chaotic system.

but it is complex enough that in modeling it, chaos and randomness may crop up (almost certainly, as a matter of fact).

Necessarily, because both are present in the real thing.

Unless you are using specially tailored (and awfully slow) arbitrary precision floating point mathematical calculations, you are going to run into the time-tried and well-known wall of limited precision floating point numbers on computers as well as precision loss (roundoff and truncation) if not countered by an almost AI-level monitoring and restoration system to go along with it. You cannot circumvent such limitations easily.

Wrong again.

The precision of the Universe itself is limited; we know that from Quantum Mechanics. We only have to simulate to that precision - which is not really all that great - to eliminate not only any likely side-effects but any possible side-effects of precision loss. And that's worst case. Very likely we don't need to even approach that precision to make things work; standard double-precision is probably plenty.

ETA: And there is still the 'halting problem' with which to contend. Your multi-billion dollar brain may end up in an infinite loop (i.e.: bloody mad!). These are NP-hard problems - not just technological hurdles...

Baloney.

There is no additional difficulty raised in the simulation; it's a perfectly straight-forward massively parallel number-crunching task. The simulation is subject to the halting problem because the brain itself is subject to the halting problem.

 

[PixyMisa]

Sorry, duplicate post.

 

[kuroyume0161]

I wasn't, actually. I was talking about working models of steam engines.

Right. But it is still the same thing - a human-built representation of a human-built object. Boy, that's difficult. A brain is not a human-built object. It is a biological system. There is a difference! Again, bad analogy - no useful purpose.

And as far as the simulation is concerned, it is a steam engine. If we analyse the work it can do, the answer is the same as for a physical steam engine. Of course, it's not a physical steam engine, so the work is simulated work, not physical work.

It is a simulation of a steam engine since it does not model every aspect thereof (unless you can show models that account for quantum affects and relativity). And you can't observe it in all ways physical unless you simulate it completely. It is an approximation (that's what I mean by simulation).

However, the work done by a brain is information processing, and there is no difference whatsoever whether this is done by a physical brain or a simulated one.

The brain does more than 'information processing'. It also controls involuntary functions, responds to chemical inbalances, emits chemicals and compounds as stimulus response (pheromones, dopamines, etc.) - and not just sensory.

An accurate model is an accurate model. We have, quite specifically, modelled the brain at the biological level - read the thread; that's what we are talking about.

As far as the simulation is concerned, it is a brain. Indeed, as far as the simulation is concerned, it is a biological brain.

Until you include subatomic interactions and energy transmissions (so-called Alpha waves, for instance), it is still not a brain. It is an approximation.

And you have a reason for this? I've already shown that we could build a sufficiently powerful computer today.

Only to yourself. As any good scientist would do, I won't agree that something can be done until it has been actually done. This is why 1950's AI was woefully laughable - they boasted to themselves that they'd have AI in ten years. But they never produced (not yet even) anything even approaching their goal. Still waiting, 60 years later!!!

Entirely false.

The human brain is a chaotic system. The simulated brain will be a chaotic system. If the simulated brain wasn't chaotic, the entire project would be a failure.

Furthermore, we don't need predictibility, because this is not a predictive model. It is a working model, and we judge its success or failure by how well it works.

If the human brain was truly chaotic, we could do nothing. More precisely, the human brain is a complex of ordered systems which display chaotic behaviour at certain levels. But, overall, the brain is ordered - it performs cyclical functions and allows for precise actions (even if they are complex).

But you still misunderstand me. I'm not saying that the simulation is a predictive model. I'm saying that you need a predictive model with which to do comparative analysis between the simulation and the real thing. But that is hard to achieve for such a complex system with tinges of chaos. For instance, we have "Working models" of weather systems, not to predict weather, but to see how these systems work. Show me a working weather system model that is even one-hundredth the 'reality' of the real system being modeled.

If you wire them together as they are in the human brain, a human brain simulation is exactly what you have. It is impossible for it to be anything else.

Is is possible that perhaps some physical structures of the brain (at whatever levels) play an important role in brain function? If the answer is even a meager 'yes', then your simplistic statement above falls apart.

The precision of the Universe itself is limited; we know that from Quantum Mechanics. We only have to simulate to that precision - which is not really all that great - to eliminate not only any likely side-effects but any possible side-effects of precision loss. And that's worst case. Very likely we don't need to even approach that precision to make things work; standard double-precision is probably plenty.

I'm still skeptical of this. I agree that some of the precision may work, but there is a possibility that it may still be a hindrance. Again, no working models to evidence the claim.

There is no additional difficulty raised in the simulation; it's a perfectly straight-forward massively parallel number-crunching task. The simulation is subject to the halting problem because the brain itself is subject to the halting problem.

IF you can build it correctly and to the required accuracy, the halting problem MAY not be a factor. As soon as you get a set of pseudo-neurons cyclically firing to each other, you have a problem (hint: 'Halting problem'). It has been a problem since the beginning of AI over 60 years ago. I haven't read anything that proclaims that it has been solved (and I have TEXTBOOKS on AI and Neural Networks which were read and used in my own actual neural network simulations)! There are partial solutions, but none are simple or fool-proof (see my statement about requirements of an AI monitoring system again). This, of course, would just raise the Godel level a notch.

Postscript:

You seem to be of the usual 'artificial brain' types who don't consider the fact that a brain is not something you turn on and you have a human being. Real brains are constructed during the gestation period, change and grow during life, are the products of billions of years of evolution, and are very dynamic, plastic systems which are interdependent upon the other systems engaged along with it. You think you can fire it up and, voila!, you have your new pal Abby Normal to talk to. I think you are dreaming.

 

[Soapy Sam]

the work is simulated work, not physical work- PixyMisa

Ah! Simulated work. I'm an expert on that!

Well, I've re-read the thread. (And will again later). This may seem trivial, but I do get the feeling we need to specify exactly what we mean by certain terms- (model, emulate, simulate and related terms), because I think we are disagreeing (in part) due to subtle differences in usage.

If something wired exactly like a brain operates exactly like a human brain, the banal question of why dead brains don't work must be answered.
Nobody here, certainly not me, is making a case for elan vital or magic, but there are people in long term comas whose neural wiring appears to be exactly the way it was before the coma. I guess Pixy- and probably everyone here-would say that some aspect of neural hardware is not the same- choose any of numerous levels of object from the molecular to large scale cellular level- How is this to be mimiced in wiring? What role does software take in this hardware brainclone?
(How do you spell mimiced anyway? That doesn't look right. Damn Christmas sherry).

I agree you can mimic an ant's neural hardware and probably produce "X"-ulated output which looks , to a human, pretty much like ant behaviour. Whether you could fool an ant is another Turing Test, I suppose; one we non-ants may be neurally unable to answer.

Pixy- I have the greatest respect for your knowledge of computers, but I think you overestimate what hardware / software modelling is capable of.
I simply do not see that a model of a brain will necessarily produce a mind, which I use here as short for "whatever it is brains do that other brains recognise as human".

It might- but I see no evidence in your argument to support the assertion .
That does not make minds magical. My own suspicion is that mind is an emergent property of life and that said life must be above a certain level of complexity. I fully accept that we can backtrack down through neurology, biology, chemistry and physics to ultimate explanations. I postulate no magic. But I still see no evidence at all that computers, phone systems , telegraphs or watermills serve as anything other than a verbal metaphor- (a model, if you will) for the complexity required to generate consciousness.
(I actually suspect electronics can become aware, but that's pure techno-bias on my part. I just don't think it will be anything like human awareness. We have only one sure sample of awareness. There may be many other kinds.)

 

[Xeriar]

Each cell in the human brain operates at about one megaflop, and has a similar amount of bandwidth (a megabyte per second). This is extremely rough, but that's the scale of the human brain. In addition, it has plasticity, which is a bit more difficult to grasp. What does that mean exactly? Sure there are ways to simulate it, but it throws a wrench into a raw model.

We'll be building petaflop machines soon? Sure. Will they cost a thousand dollars? Somehow I doubt it, the 2000-era fabs cost some four billion dollars, and are getting more expensive fast. Maybe early next century, and even then, they need bandwidth to compete in order to build such an AI, which is a significantly more difficult problem.

That said, it's also difficult to imagine such a system running on 25 watts of power.

I'm sure that, by the end of this century, we will have begun enhancing our own brains to compensate.

 

[kuroyume0161]

the work is simulated work, not physical work- PixyMisa

Ah! Simulated work. I'm an expert on that!

Well, I've re-read the thread. (And will again later). This may seem trivial, but I do get the feeling we need to specify exactly what we mean by certain terms- (model, emulate, simulate and related terms), because I think we are disagreeing (in part) due to subtle differences in usage.

If something wired exactly like a brain operates exactly like a human brain, the banal question of why dead brains don't work must be answered.
Nobody here, certainly not me, is making a case for elan vital or magic, but there are people in long term comas whose neural wiring appears to be exactly the way it was before the coma. I guess Pixy- and probably everyone here-would say that some aspect of neural hardware is not the same- choose any of numerous levels of object from the molecular to large scale cellular level- How is this to be mimiced in wiring? What role does software take in this hardware brainclone?
(How do you spell mimiced anyway? That doesn't look right. Damn Christmas sherry).

I agree you can mimic an ant's neural hardware and probably produce "X"-ulated output which looks , to a human, pretty much like ant behaviour. Whether you could fool an ant is another Turing Test, I suppose; one we non-ants may be neurally unable to answer.

Pixy- I have the greatest respect for your knowledge of computers, but I think you overestimate what hardware / software modelling is capable of.
I simply do not see that a model of a brain will necessarily produce a mind, which I use here as short for "whatever it is brains do that other brains recognise as human".

It might- but I see no evidence in your argument to support the assertion .
That does not make minds magical. My own suspicion is that mind is an emergent property of life and that said life must be above a certain level of complexity. I fully accept that we can backtrack down through neurology, biology, chemistry and physics to ultimate explanations. I postulate no magic. But I still see no evidence at all that computers, phone systems , telegraphs or watermills serve as anything other than a verbal metaphor- (a model, if you will) for the complexity required to generate consciousness.
(I actually suspect electronics can become aware, but that's pure techno-bias on my part. I just don't think it will be anything like human awareness. We have only one sure sample of awareness. There may be many other kinds.)

mimicked. Rum is beter...

Yes, this is exactly my point and in agreement with what you have been arguing. Unless one can construct an indistinguishable brain analog to a human brain (which might require an entire anatomy to boot!), what you have is an intelligent artificial lifeform which is more than likely not analogous to its human counterpart.

And Xeriar echoes my convictions: that plasticity and flexibility (such as the ability to reassign function to other brain areas in brain-damaged children) will be damned-near impossible to implement into an octillion-part system of nearly unimaginable complexity.

Think about this. You spend billions and years to build this so-called 'first artificial human brain' and then you must (by definition) spend some more years teaching it (just as any human requires). If this prototypical system is not accurate in any of multitudinous ways, you must start again from scratch - probably requiring approximately four years to reach the inevitable conclusion. Presupposing that the 'third time is the charm', you are already encroaching upon maybe fifty years of endless financial support, dedication, and popular support (if that were even possible). Without such fervor, triple that. That is at least 150 years. Poof! There goes your 'we can do this TODAY' mantra. Human brains are not simply complex logic gates. They are highly complex systems - so complex that we understand less of our brain than we understand of the Earth.

Point me to the complete and vast information that will allow us to build this artificial brain.

ETA: Let me pose this final analog to Pixy's idea of silicon-based replicas of brains. Could one build a bird in iron (or aluminum) - with mechanical systems - so precisely that it would not only fly like a bird but act like one? Pixy, you are making the mistake of equivocability between disparate architectures. A snowman is not a human made in snow. It is a crude human-like form approximated in snow. Do you suggest that a sufficiently complex snowman could ever be indistinguishable from a human being in all conceivable ways? I doubt that. So, what bolsters your notion that silicon and electricity are any more sufficient towards this idea?

 

[kuroyume0161]

To be conciliatory, I'm not suggesting that modeling a human brain is an impossibility (although my rebuttals may emote that sympathy). I'm suggesting that grandiously stating that our capabilities are sufficient before evidence of even basic principles is an inappropriate stance to take.

You may not think that my computer knowledge is as acute as Pixy's, but I've been at this game for nearly twenty years. And I've done the footwork into robotics and AI, all the way into neural networks and sensory apparatus. It's more of a hobby and I don't claim anything near the greatest minds involved in the process. Still, the literature is repleat with 'this is the way' and 'that is the way' and do we have anything near even the intelligence of a rat in computer intelligence? Unequivocably: NOOOOO.

We've 'mapped' the entire human genome (at great expense and time) and it is only a map of the base pairs. It is nothing near a complete functional explanation of the meanings of all of the pairs and their consequences to the organism. That may take a century or many to categorize. We only have a binary dump of our genome, but little idea of the instructions and operations. To boast that we can build a human intelligence today is naive and foolish at the most optimistic and fanatical. We cannot even achieve human intelligence in humans at a sufficient quantity and quality.

So, please, continue the endeavors. But do not lull yourself into a false sense of achievement where there is none achieved yet. Let the brash arrogance of the early AI 'experts' be a warning to boasting results that are not yet forthcoming. Make strides and boast when there is sufficient reason to do so!

Merry Christmas and Happy New Years!

 

[daenku32]

My simplisitic reply to this "singularity" is that, no matter how powerfull the JREF server is, the intelligence stored in it's files is only as good as the people posting on it.

You can take a 20 year old computer and make it appear and act "smarter" than a bran new computer. All you have to do is give it a programming edge.

 

[hammegk]

Pixy's problem is the one all philosophical materialists must face. If their worldview is correct, what he says on the brain model MUST be True. Nothing else is possible.

 

[Xeriar]

And Xeriar echoes my convictions: that plasticity and flexibility (such as the ability to reassign function to other brain areas in brain-damaged children) will be damned-near impossible to implement into an octillion-part system of nearly unimaginable complexity.

I'm not saying it's impossible, or even near impossible, it just means that the current quantification of computer power in question (floating point operations per second, addition in particular) does not translate well to actual brainpower. The connections, their presence, formation, destruction and regeneration of the brain have actual meaning. This sort of thing can be simulated, and is being used on a smaller scale in some current processors, but it's still another factor.

Making a perfect analogue of the human brain would require the full scope of human experience, and unless we want to simulate a quadripalegic there are going to be differences. Still, to respond to you in particular, I am pretty sure that a computer will be made in the next few decades that will be observably sentient, and difficult, though not impossible, to distinguish from being a human.

Granted, there are people who can't tell Eliza from a human, but that sort of thing is a factor as well. And yes, such an AI would need to be 'raised', and the organization of the human brain properly deciphered. Ultimately, though, it's a finite problem that nature has solved before. Saying that we won't solve it, and better, is like saying heavier than air flight is impossible.

----

Even still, as I mentioned, call me when they can build such a machine that draws only 25 watts on full load.