Autor Wątek: Summa technologiae  (Przeczytany 31754 razy)

Socrates

  • Gość
Re: Summa technologiae
« Odpowiedź #30 dnia: Maj 28, 2005, 07:57:52 pm »
Considering the fact that it is quite impossible currently to artifically expand human brains to have them acquire more information faster, but it is not impossible to do that with computers (it's simply a matter of costs), then once the barrier of AI is breached, I don;t see why computers wouldn't be able to think "better" than humans...But of course, the AI frontier is what all of this is about, as how does one make computers self-aware (And that is another topic somewhere in the archives)
Cheers, Socrates

peskanov

  • Gość
Re: Summa technologiae
« Odpowiedź #31 dnia: Maj 28, 2005, 09:17:55 pm »
Terminus,
Cytuj
Primo, I sincerely disagree with SoGo's statement, that machines would be able to think better than us. I agree with Lem, who stated (yet in his early Astronauts in 1951) that they will think faster. And they will have more data in disposition (because of no problems with forgetting, large hard drives, rom's, ram's, stuff...). So, to conclude - maybe I'd agree that they would be thinking more efficiently, but I don't think this means better.

Lem not only believes an AI could think faster, but also be more intelligent. Just check the text about Golem working on the metalanguages. And I don't think Lem has conserved exactly the same opinions about AI since the 50's; I can see a clear evolution in his ideas about super-computers along the years, being "Imaginary magnitude" his most elaborated opinion on the matter (despite the comical tone).
Also, in the Golem tale, the machine is capable to design physical extension of it's brain in order to work out problems which are out of it's reach. Golem clearly trascends human intelect in Lem's tale.

I have a hard time understanding your position, Terminus, You think an AI can be faster and more efficient, but no smarter. A few questions:
What qualities of the mind do you think makes a person be smarter than other? Or maybe you think that, at least potentially, everybody can reach the same intelectual heights?
Do you think the physical properties of the brain do not limit the mind in any way?

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #32 dnia: Maj 29, 2005, 03:08:38 pm »
Yes, I believe that everyone can ,,reach the same intellectual heights".

As for Lem's view of the matter, I am well aware of the evolution of his point of view on AI. He's not a blind man, and he sees and understands the evolution of computers. I'm not concerned about his point of view, though. I just used the example of his old book in order to emphasise the fact that the basic difference between us and machines is that they have bigger data processing capabilities. And that's all.

As for the those capabilities: I wrote quite a few times before, that I don't believe that fast data processing, ability to store it nor things like hyperthreading define intelligence.

So, peskanov, I will be able to tell if machines think 'better' than me or You, just after they start thinking at all, because nowadays they just don't. Everybody knows that the calculative efectiveness has nothing to do with AI, at least not much. There are, of course, scientists that believe that it has, but I'm not able to agree.

« Ostatnia zmiana: Maj 29, 2005, 03:08:58 pm wysłana przez Terminus »

peskanov

  • Gość
Re: Summa technologiae
« Odpowiedź #33 dnia: Maj 29, 2005, 09:08:19 pm »
Cytuj
As for Lem's view of the matter, I am well aware of the evolution of his point of view on AI. He's not a blind man, and he sees and understands the evolution of computers. I'm not concerned about his point of view, though. I just used the example of his old book in order to emphasise the fact that the basic difference between us and machines is that they have bigger data processing capabilities. And that's all.

You are quite wrong here. A common human brain has a huge  capacity for data processing (estimated by some neuroscientist in 100 teraflops), the problem is that maybe you are confusing our ability to perform calculus with data processing.
Let's take a simple example: visual recognition. Example: a person takes a look at his room and identify a chair.
This simple action, when performed by a current computer (it can be done), requires a huge amount of computations, requiring some of the best computers available to perform it in real time. And that's the case for a computer which knows a very small set of objects!
However, a human do it usually without effort, and he only uses part of his brain to perform the task.

As for Lem's view, I have not much doubts about his opinion. Both Golem XIV and DEUS (from Fiasco) are more intelligent than humans (although DEUS in a lesser degree) and capable of understanding more complex problems and make better hypothesis.
In Cyberiad and Tichy tales he always jokes about the nature of the brains. He has no prejudices about what kind of matter can constitute a thinking brain.
In "An imaginary magnitude" he jokes with his "Cogito" paradox, about the thinking machines which doubt about the capability of thinking attributed to humans!
Lem always downplays (in his most serious books) supossed transcendental attributes of mind, like telepathy or life after death.

I can search dozens of examples for you. To be fair, Lem is probably the most materialism oriented author I have read. He doesn't see the human mind as a special category in the world.
Cytuj
As for the those capabilities: I wrote quite a few times before, that I don't believe that fast data processing, ability to store it nor things like hyperthreading define intelligence.

Well, as Pascal said: "Memory is necessary for all the operations of reason".
Aside from that, you should learn more about neuroscience and the perceptron model. Modern computers are not made to reach intelligence, nor to work like a human brain. When thinkers start a discussion about "Could a computer think?", they don't refer to a modern computer, but a theorical one.
The real question, better formulated, would be: Can intelligence be correctly represented with a formal system?, and, "Can intelligence be deterministic?"
The perceptron model is the most succesful attempt to formalise the operations of a real neuron; the logical brick that would allow to build an artificial brain.
Cytuj
So, peskanov, I will be able to tell if machines think 'better' than me or You, just after they start thinking at all, because nowadays they just don't.

There is one famous sentence about that, said by Minsky. Minsky is one of the greatest minds in the XX century, and one of the fathers of AI. He said: "Of course machines can think, I am a machine and I think". I am sure Lem would agree :)
Cytuj
Yes, I believe that everyone can ,,reach the same intellectual heights".

Could you expand on this? In your opinion, can all the persons considered "slightly retarded" have a mind like Paul Dirac's one, for example?

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #34 dnia: Maj 30, 2005, 02:43:36 am »
Cytuj
You are quite wrong here. A common human brain has a huge  capacity for data processing (estimated by some neuroscientist in 100 teraflops), the problem is that maybe you are confusing our ability to perform calculus with data processing.
Let's take a simple example: visual recognition. Example: a person takes a look at his room and identify a chair.
This simple action, when performed by a current computer (it can be done), requires a huge amount of computations, requiring some of the best computers available to perform it in real time. And that's the case for a computer which knows a very small set of objects!
However, a human do it usually without effort, and he only uses part of his brain to perform the task.


Ok, I meant ability to do calculations. I am a mathematician and am fully aware of the fact, that my computer calculates faster than me, that's all I wanted to say. And if you mean recognition by neural networks, than I am not sure if I can compare it with human thinking.
« Ostatnia zmiana: Maj 30, 2005, 02:45:19 am wysłana przez Terminus »

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #35 dnia: Maj 30, 2005, 02:49:21 am »
Cytuj
As for Lem's view, I have not much doubts about his opinion. Both Golem XIV and DEUS (from Fiasco) are more intelligent than humans (although DEUS in a lesser degree) and capable of understanding more complex problems and make better hypothesis.
In Cyberiad and Tichy tales he always jokes about the nature of the brains. He has no prejudices about what kind of matter can constitute a thinking brain.
In "An imaginary magnitude" he jokes with his "Cogito" paradox, about the thinking machines which doubt about the capability of thinking attributed to humans!
Lem always downplays (in his most serious books) supossed transcendental attributes of mind, like telepathy or life after death.

I can search dozens of examples for you. To be fair, Lem is probably the most materialism oriented author I have read. He doesn't see the human mind as a special category in the world.


As I said before, I am not concerned about Lem's POV on this, I only gave the example from Astronauts. He stated therein, that a calculational machine was able to calculate faster than human, and that's all.
You don't need to give me any examples then.

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #36 dnia: Maj 30, 2005, 03:03:49 am »
Cytuj
Well, as Pascal said: "Memory is necessary for all the operations of reason".

Still, having large amount of memory and large calculational capabilities isn't equivalent to being intelligent.
Cytuj
Aside from that, you should learn more about neuroscience and the perceptron model.

I happily would, I just hope it doesn't come down to neural networks. I was greatly dissapointed when shown what they are....

Cytuj
Modern computers are not made to reach intelligence, nor to work like a human brain.

Tadaaa...

Cytuj

The real question, better formulated, would be: Can intelligence be correctly represented with a formal system?, and, "Can intelligence be deterministic?"

The perceptron model is the most succesful attempt to formalise the operations of a real neuron; the logical brick that would allow to build an artificial brain.


You say that someone tries to model the working neuron? Would You be so kind and tell me where I can learn some substantial things about it?  I wonder how does it look from mathematician's point of view...

I think I just about understood Your position... it seems that instead of quarelling You, I should try to learn something from You, hihi, why not  ::)

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #37 dnia: Maj 30, 2005, 03:20:39 am »
Cytuj
Could you expand on this? In your opinion, can all the persons considered "slightly retarded" have a mind like Paul Dirac's one, for example?


No, I am aware of the fact, that mentally handicaped persons are not able to attain certain levels. Even looking from Lem's materialistic point of view, it comes down to a fact that some parts of their brains are not working efficiently or something like that.

But I still think that if someone's brain is biologically normal, than he is capable of everything, if he/she tries. That's all I wanted to state: I don't claim, that I could be like Dirac only because I'd like to, but I claim that  it is technically possible.

I've read Achuthan's sentence about Dirac (about his works):
Each of the pieces not only is in praise of an exceptionally gifted intellect but also places on record how deeply and abidingly the human mind can delve into the realms of mathematical insight and modelling, keeping intact the spirit of beauty and clarity of a creative genius.


and am aware, that comparing 'regular men' to Dirac is sort of provocative:)  But so what. Let everybody have the chance.



peskanov

  • Gość
Re: Summa technologiae
« Odpowiedź #38 dnia: Maj 30, 2005, 11:27:40 am »
Cytuj
Ok, I meant ability to do calculations. I am a mathematician and am fully aware of the fact, that my computer calculates faster than me, that's all I wanted to say. And if you mean recognition by neural networks, than I am not sure if I can compare it with human thinking.  

I meant than "hability to perform calculation" is embedded in both the human brain and modern digital computers, no matter what the global activity is: moving a hand or solving a square root on the blackboard, both are extremely high level operations. However, It's true that you can use the calculation capabilities of a computer with total flexibility while neural networks are mostly "hardwired" to carry on specific operations (like recognition, yes).
And the total computing power of a brain, in terms of neural computing power, is estimated to be much greater than most supercomputers today.
Cytuj
As I said before, I am not concerned about Lem's POV on this, I only gave the example from Astronauts. He stated therein, that a calculational machine was able to calculate faster than human, and that's all.  
You don't need to give me any examples then.

Ok; I think I was confusing your opinions with the ones of other member of this board, sorry :D
Cytuj
Still, having large amount of memory and large calculational capabilities isn't equivalent to being intelligent.

Yes of course, but you still have to find which component of the human brain is lacking on a computer. At the moment we have a good amount of evidence about the brain being a computer made of meat.
Cytuj
I happily would, I just hope it doesn't come down to neural networks. I was greatly dissapointed when shown what they are....

Yes, neural networks. Everything in neural networks is impressive, from it's hability to learn ,to it's incredible power to solve fuzzy problems with a few neurons only.
What did you found dissapointing?
Cytuj
You say that someone tries to model the working neuron? Would You be so kind and tell me where I can learn some substantial things about it?  I wonder how does it look from mathematician's point of view...

You say you know neural networks, therefore you know the perceptron details and the mathematical formulation.
So shoot out, what's your problem with this area of scientific research? (I cross my fingers, I hope we don't have to discuss the Penrose-Hameroff lunacy).
Cytuj
I think I just about understood Your position... it seems that instead of quarelling You, I should try to learn something from You, hihi, why not  

There is no need to be sarcastic. Your arguments against AI were simplistic, so I assumed you were not well informed about the AI world.
But you are informed; ok, I beg your pardon for being condescending; elaborate more your arguments.
Cytuj
No, I am aware of the fact, that mentally handicaped persons are not able to attain certain levels. Even looking from Lem's materialistic point of view, it comes down to a fact that some parts of their brains are not working efficiently or something like that.

Not at all, from the point of view of materialism there is not a "ideal" brain which serves as reference to consider others as handicapped or not. From materialism POV, there exist strictly different brains, with differente properties and capabilities. There would be an statistical average about brain propierties, which could be used to label a brain as "normal", but that's all.
That's the reason I used the expression "slighty retarded" instead od something like "down syndrome victim".
It seems you see the human brain as an absolute value, a tool with universal habilities avalaible to all humans. A very optimistic view, to be fair.
Cytuj
and am aware, that comparing 'regular men' to Dirac is sort of provocative  But so what. Let everybody have the chance.

Yes, it is very provocative. It clashes with reality very strongly, I don't think you can support this opinion with evidence. However finding evidence in the opposite side is very easy.

SoGo

  • Gość
Re: Summa technologiae
« Odpowiedź #39 dnia: Maj 30, 2005, 12:43:56 pm »
Ok, there was a very interesting discussion, so here is my point.
I don't think capacity for operations is intelligence.
But after a very special point, these capacitys are able to copy human-acting in all points of thinking, speaking and so on very good.
A nearly perfect copy is as good as the original, if you can't see a difference between both.

The Future:
Maybe technology kill us all.
The humankind may be erased by an meteor, by war or by an enviromental collaps.
But the evolution still lives on.
And thats it, there will be life every time, just sometimes in some strange forms.

Theres a novel from, which telled about seed, that only spread by the cosmic catastrophe of an exploding planet.

Thats life.
Life goes on.
Maybe independent from us.
« Ostatnia zmiana: Maj 30, 2005, 12:44:23 pm wysłana przez SoGo »

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #40 dnia: Maj 31, 2005, 03:36:26 am »
Cytuj
I meant than "hability to perform calculation" is embedded in both the human brain and modern digital computers, no matter what the global activity is: moving a hand or solving a square root on the blackboard, both are extremely high level operations. However, It's true that you can use the calculation capabilities of a computer with total flexibility while neural networks are mostly "hardwired" to carry on specific operations (like recognition, yes).
And the total computing power of a brain, in terms of neural computing power, is estimated to be much greater than most supercomputers today.



Yes, I've read something about this recently... what was it... oh yeah... the article in ''Scientific American'' about Hans Moravec, the AI hothead. He compared the computational capabilities of computers with those of certain animals (like Bacteria, Mouse), and predicted that the 'human level' (of performance & speed) will be achieved at about 2040...  Nevertheless, he still meant   performance only.  

Cytuj
Yes of course, but you still have to find which component of the human brain is lacking on a computer. At the moment we have a good amount of evidence about the brain being a computer made of meat.

I won't object to that (regardless of how it sounds) but I does this mean, that after copying this 'meat' in silicon You get a properly working copy?
Cytuj
Yes, neural networks. [...]
What did you found dissapointing?

Uhm... I must admit, that maybe I just quit them to easily. (I specialize in different things).  I was shown the (mathematical) structure of them (during just one lecture), and what, I think, appears to You - the simplicity of the network - was for me rejecting. I just lacked respect for something that is not more complicated that a set of linear equations. I'm sure you know that... uhm.. that's kind of embarassing to say, mathematicians sometimes tend to ignore too simple things, which is of course wrong... But I understand that I shouldn't criticize the whole range of NN's applications, so maybe I should look it through again few times...
Cytuj
You say you know neural networks, therefore you know the perceptron details and the mathematical formulation.
So shoot out, what's your problem with this area of scientific research? (I cross my fingers, I hope we don't have to discuss the Penrose-Hameroff lunacy).

I just think that NNs cannot solve any problem that cannot be solved by, call them, 'classical' alghoritms. No revolution there.

Cytuj
There is no need to be sarcastic. Your arguments against AI were simplistic, so I assumed you were not well informed about the AI world.

I was not sarcastic at all. I really think I should read more about the efforts of AI-programmers, and I maybe I will. I don't think I am well infomed about the 'AI World', one can hold titles in mathematics and not be informed, I guess I'm the case. So no irony there. As I said before, the mathematics of Neural Networks seemed trivial for me, maybe I judged too hastily...

Cytuj
Not at all, from the point of view of materialism there is not a "ideal" brain which serves as reference to consider others as handicapped or not. From materialism POV, there exist strictly different brains, with differente properties and capabilities. There would be an statistical average about brain propierties, which could be used to label a brain as "normal", but that's all.

Well of course, I didn't mean anything else. That's all the meaning of the world "normal" You can really find anywhere (outside mathematics, where it means hundred different things, as You probably know).

This reminds me my former teacher's saing: There are no normal people, just the ones that haven't been yet tested.

It's of course simple, but still a bit funny.
Cytuj

It seems you see the human brain as an absolute value, a tool with universal habilities avalaible to all humans. A very optimistic view, to be fair.

Sure.
Cytuj
Yes, it is very provocative. It clashes with reality very strongly, I don't think you can support this opinion with evidence. However finding evidence in the opposite side is very easy.

Haha, too easy, unfortunately. Let's just say this is the expression of my hopes about what it means to own a brain. Don't treat it as a scientific theorem or even a hypothesis. It's just my belief.
I am well aware of the materialist point of view of this matter: that the 'geniuses' have some specific parts/regions of brain more active that "normal" ones, and so on...  But there's no discovered limit of activity (which, in my understanding, would have to be of the form of some quantitative characteristic of brain) , where "normality" ends, and the genius begins.  That's the simple consequence of the fact, that as You well know, the "normality" is just a statistical median, and nothing more. No limits are well-defined... At least I am not familiar with it.
So it's always good to hope that everyone (statistical majority) is "able & capable"... Even if that hope is somewhat insane.  

Back to NN:

I once again repeat, that I don't mean to quarell or be sarcastic. I feel I have some lacks of knowledge about the applications of Neural Networks, I'd be happy to change that.
« Ostatnia zmiana: Maj 31, 2005, 03:53:35 am wysłana przez Terminus »

dzi

  • Gość
Re: Summa technologiae
« Odpowiedź #41 dnia: Maj 31, 2005, 09:24:16 am »
Cytuj
I just think that NNs cannot solve any problem that cannot be solved by, call them, 'classical' alghoritms. No revolution there.
That's a fact, but, they are used to *find* algoritms that are unknown. So there is a basic difference between a "classical" way of making algoritms where you use your brain to make it and the "NN way" where you use a NN.
BTW it's same for genetic algoritms, heurystics and generally all AI algoritms.
« Ostatnia zmiana: Maj 31, 2005, 09:24:26 am wysłana przez dzi »

Terminus

  • Gość
Re: Summa technologiae
« Odpowiedź #42 dnia: Maj 31, 2005, 01:24:58 pm »
Well now when it comes to genetic algorhitms, than I must say I'm sort of specialist here (lucky me, at last I know what I'm talking about).
Thereby I feel sure to state, that as You know, they are based on trying, trying, trying, and trying... endlessly, unless the certain criterion is fulfilled or closely fulfilled. They surely act funny, and (sometimes) give good solvings of problems, but are they intelligent... Of course they're based on imitating evolution (mutations, crossbreeding, so on) which is very interesting, but still...
« Ostatnia zmiana: Maj 31, 2005, 01:26:34 pm wysłana przez Terminus »

Deckert

  • Gość
Re: Summa technologiae
« Odpowiedź #43 dnia: Maj 31, 2005, 01:38:58 pm »
Cytuj
...they are based on trying, trying, trying, and trying... endlessly, unless the certain criterion is fulfilled or closely fulfilled. They surely act funny, and (sometimes) give good solvings of problems, but are they intelligent...


Well, this reminds me of "learning from mistakes" - it's a sort of human bahaviour, however for genetic algorythm it's not a self-learned reaction, but already built-in one.

peskanov

  • Gość
Re: Summa technologiae
« Odpowiedź #44 dnia: Maj 31, 2005, 08:16:28 pm »
Cytuj
Yes, I've read something about this recently... what was it... oh yeah... the article in ''Scientific American'' about Hans Moravec, the AI hothead. He compared the computational capabilities of computers with those of certain animals (like Bacteria, Mouse), and predicted that the 'human level' (of performance & speed) will be achieved at about 2040...  Nevertheless, he still meant   performance only.

That's right, the numbers I mentioned (100 Teraflops for the simulation of a full human brain) are the same he writes in his homepage:

http://www.frc.ri.cmu.edu/~hpm/project.archive/robot.papers/2000/Cerebrum.html

BTW, the year 2040 is a prevision about common computers reaching 100 Teraflops. But today, a supercomputer can reach that speed. It's question of price.

However Moravec is more of an expert in robots than he is in neuroscience...He is interested in any technology which can make his robots more intelligent today, instead of working exclusively in the most promising one.
I say that because when we talk about the questions "could a computer think, be intelligent or be conscious?", I discard all the fields of the AI sciences (like expert systems, bayesian filters or heuristics) to talk exclusively about neuroscience and neural networks. This is the field closest to finding and answer for our question!
Cytuj
I won't object to that (regardless of how it sounds) but I does this mean, that after copying this 'meat' in silicon You get a properly working copy?

We already have an attempt to simulate a part of the brain, the hippocampus prosthesis:

http://www.newscientist.com/article.ns?id=dn3488

(it's fun to learn that Lem already wrote about this kind of prosthesis a lot of years ago, on his tale "Does Mr. Smith exist?").
There is also a company which allegedly tries to simulate a human cortex:
http://www.wisdomportal.com/Stanford/MarcosGuillen-CCortex.html
http://www.ad.com/
http://www.ravllc.com/robotic-blog/2004/07/artificial-development-news.htm

...but imho it seems a scam or a lunacy, not serius neuroscience. I don't think we have enough information about the brain to try something like this (yet).

This is a fast growing field of research, so everybody can start preparing their bets. Myself, I find dificult to believe that a brain (human or animal) has any kind of function aside from processing discrete information. Therefore, I think whatever a mind is, it should be computable.

Another link, more serious, about neuroscience and simulation:
http://www.rni.org/