[Comp-neuro] Re: Attractors, variability and noise
bwyble at gmail.com
Fri Aug 15 19:23:54 CEST 2008
On Wed, Aug 13, 2008 at 7:23 PM, Robert Cannon <robert.c.cannon at gmail.com>wrote:
>> As I understand the other end of the spectrum, we construct increasingly
>> realistic models and end up with a simulated brain without a real
>> understanding of how it works, which makes no sense to me. Understanding is
>> what we're after, and that understanding can only reside in the brains of
>> the population of scientists, not in their models.
> Brad's point is fascinating - not least because I couldn't disagree more.
> I do like the notion of understanding, but I suspect it is also somewhat
> self-indulgent, because there may not be a level on which it can be shared
> above that of working models.
But It is still the people, and not the models, that possess the
understanding. If I were to email you my theory, encapsulated in a
functioning model of the hippocampus using millions of 500 compartment
neurons, you would still need to execute the model and build an
understanding of its function in your head, in order to have any useful
thoughts about it. To do that, I would have to tell you my theory in
verbal form, and how to find evidence of it within the model's behavior.
For complex behavior, that verbal formulation has to be high enough level to
capture the dynamics of interest in the behavior. I mean, do channel
dynamics have much to say about language production? And if not, why use a
supercomputer to simulate them?
[snipping Astronomy analogy]
> My point is that for this particular problem, high-level theory is
> not much use. Some of it is epiphenomenal, and the rest is just plain
> wrong. The models work fine but they are too complicated to run in
> your head. The simpler things that you can run in your head or on
> paper are too coarse to be any use.
I agree with you completely. One cannot run the model in one's head because
even a simple equation can have surprising dynamics. But abstract modelling
does not involve thinking in one's head, it just involves a level of
simulation that averages over low level details, much as you already average
over Brownian motion.
But one does need to *have* a high level theory to understand behavior. The
back and forth between the scientist and the model is what generates
progress, not the mere existence of the model.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Comp-neuro