Firstly, there is absolutely no evidence to support the statement that neurons are an optimal architecture for their functionality. It is almost certainly the case that other structures can compute and not necessarily in the way the neurons do. This really gets down to the fundamental question of ''what is computation/information processing''.

quote:

---

Original post by aboeing
I believe there definately is, just look at how limited the AI/learning field really is, its basically just neural nets, evolutionary algorithms, and fuzzy logic.

---

AI and Learning is much, much more than just artificial neural networks, evolutionary algorithms and fuzzy logic. Perhaps if you had read and studied more before making this comment, you''d realise how naive it is.


Timkin

God, what a beligerant prick you are Mr.Moderator.

Care to elaborate a bit more instead of releasing the angst on us?

geeeeezz.......

---

[Hugo Ferreira][Positronic Dreams][Colibri 3D Engine][Entropy HL2 MOD][Yann L.][Enginuity]
Stop reading my signature and click it!

Timkin,

I like the comment on what is computation/information processing as opposed to what is intelligence. I''m sure no one will argue about the "intelligence" of bacteria/viruses, but they certain do an incredible job in information processing and finding some very nice minima within their fitness landscapes, don''t you think? Adaptation in general is an interesting study in information processing.

Just my $0.02.

-Kirk

quote:

---

Original post by Timkin
Firstly, there is absolutely no evidence to support the statement that neurons are an optimal architecture for their functionality.

---

Of course not. Nobody said such a thing.

Let me point out that there is absolutely no evidence to support the statement that other "organic" forms of processing are better than neurons.

quote:

---

Original post by pentium3id
God, what a beligerant prick you are Mr.Moderator.

---

There is absolutely no need for this kind of language and attitude in this, or any other forum. If you disagree with something I have written, then say so. If you have a personal problem with me, then take it up with me privately. There is nothing I said above that warrants such abuse.

Timkin

If I have ofended you I apologise, but you did jump in "in the middle" of discussion, and instead of puting anything on the table all you did was try to invalidate anything that was being put forward.

If you see my posting history you''ll see im not the type to start flame wars.

quote:

---

AI and Learning is much, much more than just artificial neural networks, evolutionary algorithms and fuzzy logic. Perhaps if you had read and studied more before making this comment, you''d realise how naive it is.

---

If you had read your own post, you would see how beligerent it is. BTW, care to put forward another 3 AI areas not mentioned there?

quote:

---

Firstly, there is absolutely no evidence to support the statement that neurons are an optimal architecture for their functionality.

---

And who was defending such a statement? I was just making the observation that neurons appear to have cornered the market for "high-level" AI. No one here is defending they are the most developed structure for what they do. Had you cared to read my post, you would know that that is exactly what I''m looking for, "alien" forms of AI, that don''t use neurons, or use similar structures, but not exactly the same...

I would also like to lodge a complaint, because I don''t think you are fit to hold the responsability of Moderator of this, or any other forum, and therefore would like to know the emails of the persons I should forward that complaint to.

Also, congratualtions for your baby daughter, may she live long and be healthy.

---

[Hugo Ferreira][Positronic Dreams][Colibri 3D Engine][Entropy HL2 MOD][Yann L.][Enginuity]
Stop reading my signature and click it!

Interesting concept:

A human being can comprehend thousands of complex concepts, recognize a virtually infinate set of objects, retain hundreds of names and dates and sequences, but how many people can add two 20 digit numbers insantaniously?

Neural nets sacrifice speed for flexability, not only in computer science, but in the real world also.

There has got to be a better way!

-SniperBoB-

Well, maybe we can design and manufacture a better way, but we will allways base it off what nature has show us that works.

It will be a mix between our "brute-force-attack" methods and nature''s adaptation methods.

If the universe, or at least the environment arround you can be reduced to a (complex) formula, then IMO neurons aproximate that function pretty well.

So, intelligent behaviour seems to be born out of "systems" that try to replicate/aproximate the "environment" they''re in as much as possible, either by genetic mutation, which gives an immediate "bonus" to the creature once it is born, because its genetic makeup already deals better with the environment, or using the neuron aproach, in which the subject must be presented with the situation, and if he survives said situation, he has "evolved" into a more adapted being.

And we can''t really get away of intelligence being an aproximation to the environment equation, can we?

Other creatures, even alien lifeforms might have come up with other chemical/organic structures to deal with the environment in "real time" (non GA ways), but once having successfully dealt with a new situation, if the creature doesnt store how it solved it, then there is no evolution.

So the only other aproach is to "fake it", by actually archiving all your interactions with the environment. Something like V''GER from Star Trek 1. A being that actually stores all the situations it has been in, and then, when needed, retrieves them to deal with said situation once more.

We Humans, and most other multi-celled organisms are incapable of storing *every* living experience we have, so nature came up with a less expensive way...

sorry for the long rant...

---

[Hugo Ferreira][Positronic Dreams][Colibri 3D Engine][Entropy HL2 MOD][Yann L.][Enginuity]
Stop reading my signature and click it!

quote:

---

Original post by pentium3id
If you had read your own post, you would see how beligerent it is. BTW, care to put forward another 3 AI areas not mentioned there?

---

My statement does not indicate belligerence on my part. It is simply a statement of fact about the naivete of the statement that AI was just neural nets, fuzzy logic and evolutionary algorithms. If you care to disagree with my assessment of this statement, then by all means do so.

quote:

---

Original post by pentium3id
And who was defending such a statement?

---

It has nothing to do with anyone defending it. The statement was made in this thread. I stated that there was no evidence to support the statement; not within this thread and not within the scientific community at large. If there is literature that I am unaware of that supports this statement, please direct me to it. I would be most interested to read it.

quote:

---

Original post by pentium3id
I was just making the observation that neurons appear to have cornered the market for "high-level" AI.

---

Why do you believe my initial statements were directed at you? They clearly were not. As for the statement that neurons are ''high-level AI'', do they actually have anything to do with AI at all? Most would say no. Perhaps you meant artificial neural networks though? It isn''t clear from your statement which exactly you mean.

quote:

---

Original post by pentium3id
No one here is defending they are the most developed structure for what they do.

---

I did read your post. Perhaps you should read the first response to your post again...

quote:

---

Original post by owl
neurons are the optimal solution for organic processing of information

---

quote:

---

Original post by pentium3id
I would also like to lodge a complaint, because I don''t think you are fit to hold the responsability of Moderator of this, or any other forum, and therefore would like to know the emails of the persons I should forward that complaint to.

---

You are well within your rights to lodge a complaint if you feel that is appropriate. You can send your complaint to Dave Astle at dave@gamedev.net.

As to whether I am fit to moderate this, or any other forum, I''ll let history (and Dave) be the judge of that, rather than your opinion.

Timkin

AI and Learning is much, much more than just artificial neural networks, evolutionary algorithms and fuzzy logic.
Yeah, I'll back up (some of) pentium3id's comment on this. Enlighten us, tell us about any other major fields of AI that are floating about.

A human being can comprehend thousands of complex concepts, recognize a virtually infinate set of objects, retain hundreds of names and dates and sequences, but how many people can add two 20 digit numbers insantaniously?
Well I'm not sure of the exact number, but iirc, the eyes contain a comparable number of neurons to the brain? (feel free to tell me i'm full of $#!7). Still some people can add two 20 digit numbers instantly, just goes to show you the amazing flexibilty of our brains.

There has got to be a better way!
Definately. We barely even understand how neural nets work. Which is why most of the time you just slap on a GA and let it figure out the weights.

So the only other aproach is to "fake it", by actually archiving all your interactions with the environment.
Yeah but I dont consider being a big database of previous conditions as AI. One could argue that a book is AI in that case.

And its not the only other approach, (isnt that what this discussion is about?), we just dont know of any others. (damn stupid neural net brain, if only it worked better I could invent something better than neural nets. maybe if I invent something better than neural nets, i can use that to invent something better than neural nets... ha. ha. ha...)

EDIT: timkin replied during my reply.
My statement does not indicate belligerence on my part. It is simply a statement of fact about the naivete of the statement that AI was just neural nets, fuzzy logic and evolutionary algorithms. If you care to disagree with my assessment of this statement, then by all means do so.
I agree, there is more to AI than just nn,fuzzy and EA's, but nothing truely generic in its learning like the 3 I mentioned, as far as I know. (In fact, I would probably be willing to reduce my statement to only NN&EA) Like I say, you obviously had something in mind which contradicted my statement, please let us know what it was.
PS: Just to be pedantic, I did say "basically", not "only".

[edited by - aboeing on December 3, 2003 11:26:27 PM]