Artificial Neural Networks <=> Animal Brain?

[svnarax]

Hi all,

I have been searchig the net for a satisfactory answer to my question which is

Are ANNs computationally equivalent to animal brains?
(ie) Can we simulate animal behaviour using an ANN?
(ie) Consider a newly born dog and an ANN. Add to the ANN all information that the newly born dog has and assume that there is some means of providing the ANN the sensory information that the dog gets. In such a case will the two behave in the same way or will they differ.

If they differ in what way?

Can any of you guys clarify my doubt or atleast give me some links which might be related to this question.


-svnarax

 

[Blindman]

Neural nets are not multi purpose machines. You cant just pop a pile of neurons in and expect it to behave. They require very specific structures to do there jobs. Thus to simulate an animal brain you need to create a neural net with the same structures and pathways as the animal you are trying to copy. Also ANN's do not grow or die and thus lose a lot of the adaptability that biological systems have.

Are ANNs computationally equivalent to animal brains?

No, they are similar conceptually but functionally they differ greatly. Organic NN's use neurotransmitters that are not just created at the synapse but are created in glands through out the body and picked up from the environment. The liver can also plays a role in the breakdown of many neurotransmitters, altering the function of the neurons and synapses through out the body. These systems are not normally included in ANN's. There are many more differences which I will not go into now.

(ie) Can we simulate animal behaviour using an ANN?

Yes we could but a simulation would only be a subset of the animals overall behavior.

 

[svnarax]

"Yes we could but a simulation would only be a subset of the animals overall behavior.[/QUOTE]"


Can you say the same thing about the human brain?

I'm not talking about the various complex things that humans can do but just about the computational ability.

For eg: Finite State automata(machine) cannot solve all the problems that a Turing Machine can. So Turing Machines are superior to FSAs.

But a Turing Machine is computationally equivalent to our modern day computer though both follow different paradigms.

So is the human brain <=> a neural net in the sense similar to
a Turing machine <=> a RAM machine (our modern day comp.)

-svnarax

 

[Blindman]

So is the human brain <=> a neural net in the sense similar to
a Turing machine <=> a RAM machine (our modern day comp.)

Yes you could use that same analogy but the brain is more then a neural net it is also part chemical reaction, a bio computer.

I think that the true future of AI lays in bio computers and ultimately in quantum computing.

 

[kmguru]

And do not forget that the brain is a multi-layered snowflake program with part binary computation, part fuzzy logic, part basic neural, and part pseudo random activity among others. A neural net as understood today would have a heard time mimicing all form of computation.

 

[Rick]

Neural Network essentially follows everything i would say.You have to keep a threshold Function related to it.
For example in prediction,you could have inputs being given to the nodes and prediction node calculating and finally identifying pattern to give prediction.We can therefore decide what threshold function can be,rejection or acceptance might form a criteria,or may be VERY,SOMEWHAT etc might be another form of criteria...or a Pseudo Random mutating Threshold might be created...isnt it?


bye!