2
$\begingroup$

First of all, I would like to say that I know nothing about cognitive sciences. I am just posting this question because it interested me and I stumbled across this group in stack exchange.

I was recently speaking to a cognitive scientist who had just finished modelling a certain function of the brain for his thesis (using a simulation program).

He soon told me that this can be done because the neurons are an analogue system and this analogue system can be modeled using math (like any analogue system that I know of).

My first question was: Does his task in "modelling the neurons" have a fixed set of rules or steps logical that he follows.

He said that there are clearly defined rules even though many rules are still undefined.

Then my second question was: Could it be possible to automate the task of "modelling the neurons" (using IT) in order to make larger systems more quickly and to test the disputable arguments more quickly.

Now my question to the forum is: Is this a realistic approach and has anyone here already tried to do this? Before waving it aside please consider the fact that IT has been used to develop software that can solve complex problems. Ie. Mathematica, Sage, Matlab etc..

I am aware that it is much more complicated than math, but if it is possible to define rules it should also be possible to apply those rules automatically...

I'm sorry if I used bad or confusing terminology. I have no clue about this field and I am just asking out of curiosity..

$\endgroup$
  • $\begingroup$ I'm glad Seanny123 managed to provide such a clear answer (and hence I reopened this post). An edit of this question to make it more focused would be most appreciated though. $\endgroup$ – Steven Jeuris Sep 1 '15 at 16:59
8
$\begingroup$

This is an incredibly broad question, but I admire your spark of curiosity so I'm going to give you a quick answer anyways.

Firstly, there is no defined steps for modelling parts of the brain. The laboratory I was a member of created Spaun. Spaun is the world's largest functioning brain model and uses biologically plausible neurons as the foundation for it's computation. The methodology for creating the most recent iteration of this model varies wildly from part-to-part. The vision system was adapted from Deep Learning Convolutional Neural Networks, the basal-ganglia and thalamus came from an anatomical model and the working-memory model came from engineering analysis of dynamical systems. None of these could have come from evolutionary algorithms or any other sort of algorithm known to humanity at this point.

Another way of modelling the brain (mostly the human behaviour part of it) is ACT-R, which doesn't use neurons at all. Once again, there's no algorithm for evolving this system and interestingly it's the foundation behind a lot of the intelligent systems you mentioned.

The search for evolution and learning methods for groups to gain intelligence is generally called Artificial Life (ALife).

You'll notice that there's a bunch of different levels to approach the problem of modelling the brain. Your friend was at the biological bottom-up level, ACT-R is at the behavioural top-down level and Spaun is the weird hybrid. Although all of these systems use IT in some way (Spaun was built with a Python library called Nengo, ACT-R is written in LISP), the processes for creating these systems cannot be automated and would require the creation of a general artificial intelligence.

$\endgroup$
  • 1
    $\begingroup$ And I admire your courage in answering this question despite its scope and gathering of close-votes. +1 and I retracted my close vote. $\endgroup$ – AliceD Aug 21 '15 at 14:00
  • 1
    $\begingroup$ I did not expect such a concise yet interesting answer to a broad question. $\endgroup$ – Mike de Klerk Sep 1 '15 at 17:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.