[FoRK] On the unreality of bottom-up brain simulation

Jebadiah Moore jebdm at jebdm.net
Wed Aug 18 13:59:27 PDT 2010


On Wed, Aug 18, 2010 at 9:04 AM, Russell Turpin <russell.turpin at gmail.com>wrote:

> What I would ask is two questions. First, how did you validate the
> model? Second, having made and validated the model, would you then say
> you are STILL ignorant of plant hormone dispersal?


1) I validated it against empirical data (along the way doing quick checks
based on a visualization, based on what someone more knowledgeable than me
said it ought to look like, to make sure there weren't obvious bugs).

We could presumably check a model of the brain against brain scans of some
variety, but I can't think of any easy intermediate checks you could make
(but I am not a neuroscientist).  And there's the obvious check against
human intuition; put it in a robot and see if it works.  Though that one
would take a while.

2) I did learn something about plant hormone dispersal, of course, but I
could still tell you very little about it, and couldn't have even at the
time.  Some things I learned in the process of building the model, but these
were mostly the specifics of the math we were using for active/passive
transport.  But I couldn't have guessed what the thing would act like before
I ran it, except that I was told roughly.  (Well, if I had stared at the
math long enough, or did some research, I probably could have figured it
out.  But I didn't need to to make it work.)

Isn't that one of the main points of computer modeling?  That by figuring
out the underlying laws and running the thing, we can see what happens
without having to know in advance?  (The other main points are ease of
experimentation, and a sort of "(psuedo)proof via visualization" that our
math is right.)

-- 
Jebadiah Moore
http://blog.jebdm.net


More information about the FoRK mailing list