Getting Exaflop computers by 2020:

Originally posted by googlefudge
You are making a positive claim, that it's impossible for a Turing Machine to emulate
a biological brain.

Positive claims require positive evidence, proof.

Show me the PROOF that it's IMPOSSIBLE for a Turing Machine to emulate a biological
brain.

Human's can take on NP hard problems and get it right.

So?

We also very f ...[text shortened]... e ever heard of that a
sufficiently powerful and suitably programmed Turing Machine cannot do.

No, I'm not going to prove anything. I was stating an opinion, this is perfectly reasonable behaviour on an internet forum. What's more I gave some reasons and I gave an example of something that humans still beat machines at (Go).

For one thing I'd have to show that NP ≠ P, which there's a one million dollar prize for. I'd have to show that P ⊂ BQP. These are hard problems. Since a brain is a neural network there's a good chance that the relevant complexity classes are EXP, NEXP and BQEXP.

Both you and twhitehead seem to think that the universe is deterministic or, if it is not, that it makes no difference. I differ in that, but as we know that's not easily proved. Although, if that were the case all nuclear decays should happen at the same time, for nucleii produced at the same time. The nucleii either exhibit Fermi or Bose-Einstein statistics, which implies that the states are identical. So you are relying on Bohmian mechanics to be rigorously correct as as far as I know that's the only deterministic version of quantum theory [1].

My favoured interpretation of quantum theory is the ensemble interpretation - sort of Cøpenhagen lite - it doesn't deny the reality of things like the wavefunction and particles between measurements. If you are right and the universe is deterministic then quantum machines will be no faster than conventional machines with the same number of internal states. If I am right then they will be [2]. Although the various sides in the interpretation of Quantum Theory wars always claim their version is right, it is nowhere near settled. But since various structures within neurones work at quantum scales (microtubules) it is not beyond the bounds of reason that thinking involves quantum computation, cf. quantum mechanics of chloroplasts. If the universe is not deterministic and neurones are little quantum computers then we are massively parallel quantum machines and can out think peta-, exa-, and whateva- scale conventional Turing machines.

Incidentally BQP means a problem solvable with bounded error on a quantum Turing machine. So quantum computers get it wrong too.

[1] Technically Everett's many worlds interpretation is deterministic, with the wavefunction evolving as |U(t)> = exp(iHt)|U(0)>. However from the point of view of any given observer it is indistinguishable from being in a Cøpenhagen universe.

[2] Bohmian mechanics may have some way round this, so I'm not making that statement with complete certainty.

Originally posted by DeepThought
Yes, I can build a computer. I can even do it from raw materials if you insist, although it'll require someone to turn the handle to make it work. I'm thinking of something on the lines of Charles Babbage's Analytical machine. All it needs to qualify is to be Turing Complete - it doesn't need to be electronic or particularly fast.

No you cannot, I don't believe you can do it. It's not anyone can do for himself. You always need the effort of other people, some of them old, some even dead. The collected knowledge of science you cannot repeat in a lifetime, not even you.

Originally posted by DeepThought
What's more I gave some reasons and I gave an example of something that humans still beat machines at (Go).

And when a computer beats humans at Go, what will that change? Will your claim be disproven?

For one thing I'd have to show that NP ≠ P, which there's a one million dollar prize for. I'd have to show that P ⊂ BQP. These are hard problems. Since a brain is a neural network there's a good chance that the relevant complexity classes are EXP, NEXP and BQEXP.
I think you first have to show that humans are actually capable of solving NP problems. I say they are not. Guessing the answer doesn't count, as computers are just as capable of solving them that way.

Both you and twhitehead seem to think that the universe is deterministic or, if it is not, that it makes no difference.
I think it makes no difference with regards to the human brain. I have no reason to believe the human brain does anything that cannot be emulated in a Turing machine. Pseudo randomness can be generated using Turing machines too. If you are referring to quantum effects (not quite the compliment of determinism), then again, I see no evidence that they are involved in human thinking (other than in the mechanics).

If the universe is not deterministic and neurones are little quantum computers then we are massively parallel quantum machines and can out think peta-, exa-, and whateva- scale conventional Turing machines.
No. We would potentially be able to out think them. It is not necessarily true that if we have quantum components then we use them effectively, nor is it a given that if the universe is non-deterministic then our neurons are quantum computers.

Originally posted by FabianFnas
No you cannot, I don't believe you can do it. It's not anyone can do for himself. You always need the effort of other people, some of them old, some even dead. The collected knowledge of science you cannot repeat in a lifetime, not even you.

He would look up the designs on his computer.

Originally posted by twhitehead
He would look up the designs on his computer.

Exactly. In order to build a computer you have to get help from a computer. No human being is able to build a computer without the help of a computer.

In order for a human to build a human, you must have help of another human. No human can produce another human without help from another human.

In order for a computer to build another computer, it must have help from a human. No computer can produce another computer without help from a human.

We all exist in symbiosis.

Originally posted by FabianFnas
Exactly. In order to build a computer you have to get help from a computer. No human being is able to build a computer without the help of a computer.

In order for a human to build a human, you must have help of another human. No human can produce another human without help from another human.

In order for a computer to build another computer, it mu ...[text shortened]... o computer can produce another computer without help from a human.

We all exist in symbiosis.

Well, Babbage did. He was well on his way anyway. It was finished a couple hundred years later but that design worked. Slow compared to electronic computers but it did the job for its day.

Originally posted by sonhouse
Well, Babbage did. He was well on his way anyway. It was finished a couple hundred years later but that design worked. Slow compared to electronic computers but it did the job for its day.

No, he didn't. Without the works of his education (built of others), his experience (thanks of others), and so forth, he couldn't come far.

I say that not anything is built on one person only. Anyone can take others work and go further, but it is still built on other thoughts. Inclusive Babbage, Darwin, Einstein, Hawking and the rest of the giants.

Noone can built a computer from scratch without previous persons work.

Originally posted by FabianFnas
No, he didn't. Without the works of his education (built of others), his experience (thanks of others), and so forth, he couldn't come far.

I say that not anything is built on one person only. Anyone can take others work and go further, but it is still built on other thoughts. Inclusive Babbage, Darwin, Einstein, Hawking and the rest of the giants.

Noone can built a computer from scratch without previous persons work.

I really don't think your contribution is adding anything at all. By your criteria no human has ever done anything. Since I'm thinking in terms of something quite simple that would implement a programming language like brainduck [1] I can invent the design for myself and not use any electronic or electrical components at all. So stop pretending you're raising valid objections. You're just being a troll.

[1] d -> f

Originally posted by DeepThought
I really don't think your contribution is adding anything at all. By your criteria no human has ever done anything. Since I'm thinking in terms of something quite simple that would implement a programming language like brainduck [1] I can invent the design for myself and not use any electronic or electrical components at all. So stop pretending you're raising valid objections. You're just being a troll.

[1] d -> f

I on the other hand think he has a valid point. Your original claim was that humans can create computers but computers cannot create humans. Not only did you use a rather unfair challenge given that creating a human from scratch is far more difficult (even humans haven't cracked that one), but it is not true that a single human has been able to make a typical modern computer. Making a basic turing machine is a dodge. The challenge is for you to make a computer capable of Exaflops (since that is the scale of computer we are discussing) - something the whole human race is yet to achieve.

Originally posted by DeepThought
No, I'm not going to prove anything. I was stating an opinion, this is perfectly reasonable behaviour on an internet forum. What's more I gave some reasons and I gave an example of something that humans still beat machines at (Go).

For one thing I'd have to show that NP ≠ P, which there's a one million dollar prize for. I'd have to show that P ⊂ BQ ...[text shortened]... echanics may have some way round this, so I'm not making that statement with complete certainty.

What's more I gave some reasons and I gave an example of something that humans still beat machines at (Go)

So???

You are basically making the argument from ignorance fallacy. The fact that it hasn't been done YET doesn't mean
it cannot be done. It's at the very best VERY WEAK evidence that it cannot be done, and it probably doesn't even
get that far. People used to be much better at chess than computers, till computers got powerful and well programmed
enough that they now beat us.

For one thing I'd have to show that NP ≠ P, which there's a one million dollar prize for. ...

No, you are the one claiming that NP = P is a precondition for Turing Machines being able to simulate a human brain.
Another claim you have not substantiated.

Both you and twhitehead seem to think that the universe is deterministic or, if it is not, that it makes no difference.

You really need to stop guessing what I think, you get it wrong when you do, ASK don't assume.

I don't know if the universe is deterministic [and how are you defining deterministic?] but that is irrelevant to the
question at hand. The question is "Can a [non-quantum] Turing Machine of suitable design and programming emulate
a human brain?".
Given that we currently have no reason to believe that a biological brain is a quantum computer utilising entanglement and
state superstition to store and manipulate the biological equivalent of quibits, and plenty of reason to suppose that in
the messy noisy warm environment in the brain such things are nigh on impossible, I see no reason to suppose that the
behaviour of the neurons in the human brain cannot be adequately simulated on a sufficiently enabled Turing Machine.

Basically I concur with twhiteheads 3rd post on this page.

No, I'm not going to prove anything. I was stating an opinion, this is perfectly reasonable behaviour on an internet forum.

Ok fine... But you stated it in the form of statement of fact.

Oh yes it is. Human's can take on NP hard problems and get it right. No matter what the processing power of a Turing machine it simply cannot compete with a human on some sorts of problems.

That is a statement of fact, and I dispute that it is fact. So I ask for proof.

You have now dropped it down to an 'opinion', so, ok fine. I still think it's wrong and as yet unjustified.

Originally posted by twhitehead
And when a computer beats humans at Go, what will that change? Will your claim be disproven?

[b]For one thing I'd have to show that NP ≠ P, which there's a one million dollar prize for. I'd have to show that P ⊂ BQP. These are hard problems. Since a brain is a neural network there's a good chance that the relevant complexity classes are EXP, NEXP an ...[text shortened]... is it a given that if the universe is non-deterministic then our neurons are quantum computers.

The thing with go is that it's in the complexity class EXP, so all you have to do is double the board size and the computer can't cope, the human player can.

Guessing the answer doesn't count, as computers are just as capable of solving them that way.

No, this is the point with BQP, there is a bounded error. Consider a decision problem with a large space of possible answers. A quantum machine has a 1/3 chance of getting it wrong on any attempt. But it requires a fairly small number of runs to get the correct answer with a very high degree of confidence. Further, if one can use a conventional machine to check the answer then incorrect guesses can be ruled out quickly.

If you set a conventional machine guessing then it has to sample a significant fraction of the space of possible answers to have a hope of getting the right one (so your pseudo-random number generator had better not generate repeats). Consequently a conventional Turing machine is slower than a Universal quantum machine (aka Quantum Turing machine).

It's the quality of the guessing that is key to this.

Well, it's difficult to answer your point in the last paragraph. Since the science isn't well enough developed to rule it either in or out. However I'll draw your attention to this behaviour in photosynthesis in which chloroplasts exploit quantum effects for their function. This demonstrates that organisms are known to use quantum effects, beyond a basis for classical mechanics.

A phenomenon known as quantum walk increases the efficiency of the energy transport of light significantly. ... The particle's wave properties enable it to cover a wider area and try out several possible paths simultaneously, allowing it to instantaneously "choose" the most efficient route, where it will have the highest probability of arriving at its destination in the minimum possible time. ... quantum walking is only possible over very short distances ... These [interference effects] cause the particle to lose its wave properties for an instant before it regains them once again after it is freed from its locked position through a classic "hop". The distance towards the center is therefore covered in a series of conventional hops and quantum walks. [1]

I see no particular reason that neurones should not also exploit some sort of effect like this.

[1] http://en.wikipedia.org/wiki/Photosynthesis#Efficiency
The writer of the Wikipedia page was a little verbose so I cut out some waffle.

Originally posted by googlefudge

What's more I gave some reasons and I gave an example of something that humans still beat machines at (Go)

So???

You are basically making the argument from ignorance fallacy. The fact that it hasn't been done YET doesn't mean
it cannot be done. It's at the very best VERY WEAK evidence that it cannot be done, and it probably doesn't ...[text shortened]... w dropped it down to an 'opinion', so, ok fine. I still think it's wrong and as yet unjustified.

The word "seem" gets me out of that complaint. You do seem to regard the world as deterministic, I get the strong impression of that from the various threads in Spirituality touching on metaphysics you've posted in.

When people make statements of opinion they normally present them the same way as statements of fact. It's the way natural languages work.

No it's not an argument from ignorance. I have a good reason for saying that computers can't cope, you just need to make the board bigger as they catch up. A computer could play Go perfectly on a small enough board.

Fnabian Fnas always does this. He was denying that it's possible to write down an irrational number in a thread a while ago (so what? it doesn't invalidate Cantor's diagonalisation proof). He invariably takes positions of extreme pedantry. twhitehead's objections to my overall position actually form the basis for a debate.

The final point is that my "humans build computers but not the other way round" point was clearly not meant to be taken particularly seriously. Although, if you want to take it seriously then I'll accept "faithfully simulated on" as built. A virtual human will do for this. By faithful I'm borrowing a bit of language from the theory of group representations. A faithful representation of a group doesn't clobber any elements (e.g. all groups have the trivial representation where all elements are mapped to 1). So a faithful simulation of a human would be no different to a human in any meaningful sense.

Originally posted by DeepThought
The word "seem" gets me out of that complaint. You do seem to regard the world as deterministic, I get the strong impression of that from the various threads in Spirituality touching on metaphysics you've posted in.

When people make statements of opinion they normally present them the same way as statements of fact. It's the way natural langu ...[text shortened]... . So a faithful simulation of a human would be no different to a human in any meaningful sense.

The word "seem" gets me out of that complaint. You do seem to regard the world as deterministic, I get the strong impression of that from the various threads in Spirituality touching on metaphysics you've posted in.

Then you are misunderstanding my posts. Given you now know this, that should stop you
from assuming you do understand my posts.

When people make statements of opinion they normally present them the same way as statements of fact. It's the way natural languages work.

It's how they work if you are being sloppy and lazy. Which is a good way to get into trouble in a debate.
This is a debate, and it got you into trouble.

No it's not an argument from ignorance. I have a good reason for saying that computers can't cope, you just need to make the board bigger as they catch up. A computer could play Go perfectly on a small enough board.

Still an argument from ignorance fallacy. I understand exactly what your point is and it's wrong.

If you build a program that tries to calculate all possible moves and evaluate which is best,
which is a simple brute force method of solving the problem that is simple from a 'human
programming computer' perspective, then your program will require exponential increases in
processing power as the board size increases.

However that is not the only way to write a 'Go' solving computer program for a Turing Machine.
What you are arguing is that because nobody as yet written a good 'Go' playing bot, that it's
not possible to build a good 'Go' playing bot. Which is classic argument from ignorance.

The final point is that my "humans build computers but not the other way round" point was clearly not meant to be taken particularly seriously.

Clearly it wasn't clear.

Originally posted by DeepThought
The thing with go is that it's in the complexity class EXP, so all you have to do is double the board size and the computer can't cope, the human player can.

Guessing the answer doesn't count, as computers are just as capable of solving them that way.

No, this is the point with BQP, there is a bounded error. Consider a decision probl ...[text shortened]... esis#Efficiency
The writer of the Wikipedia page was a little verbose so I cut out some waffle.

I see no particular reason that neurones should not also exploit some sort of effect like this.

However for the purposes of this discussion what matters is if neurons computing
ability is quantum, and there are really strong reasons for thinking it's not.
ALL the fundamental underpinnings of reality are quantum [or whatever may or may not lie under that].
However that doesn't mean that any macro system cannot be classically simulated.

Even if that were not true, I can write the corollary of your statement...
I see no particular reason that neurones should also exploit some sort of effect like this...
And have it be just as valid.

This is not any kind of convincing argument.

You are basically saying that "it's impossible for the human brain to be simulated on a Turing
Machine because the human brain MIGHT work in a way impossible to simulate on a Turing Machine"

The correct form is "It MIGHT not be possible for the human brain to be simulated on a Turing
Machine because the human brain MIGHT work in a way impossible to simulate on a Turing Machine"

Originally posted by DeepThought
The thing with go is that it's in the complexity class EXP, so all you have to do is double the board size and the computer can't cope, the human player can.

That depends on how the computer is programmed.

No, this is the point with BQP, there is a bounded error. Consider a decision problem with a large space of possible answers. A quantum machine has a 1/3 chance of getting it wrong on any attempt. But it requires a fairly small number of runs to get the correct answer with a very high degree of confidence. Further, if one can use a conventional machine to check the answer then incorrect guesses can be ruled out quickly.
Who cares what quantum machines can do. We are discussing the human brain (which is not known to be a quantum machine). Can your human brain give accurate answers to the class of problem we are discussing? Answer: no, it can't.
So why are you demanding that a computer do so?

Well, it's difficult to answer your point in the last paragraph. Since the science isn't well enough developed to rule it either in or out.
There are very good reasons to think it is out.

However I'll draw your attention to this behaviour in photosynthesis in which chloroplasts exploit quantum effects for their function.
Well light diffraction is a quantum effect and we use that to see. That doesn't suggest we use quantum effects computationally. As I already said early, how the brain works at the smallest level is actually fairly well known. What we lack is computation power to simulate it and learn how the larger scale effects emerge. But there is no reason whatsoever to think quantum effects are involved (beyond the fact that everything is a quantum effect 🙂 ) or necessary.