Standards in Artificial Intelligence

L.A. Loren lloren at mitre.org
Wed Nov 8 09:43:55 EST 2000


Nick Maclaren wrote:
> James Hammerton <james at tardis.ed.ac.uk> writes:
[snip]
> |> I disagree, the aim of neural networks is to employ abstractions of
> |> how the brain works in computing, e.g. using architectures that are at
> |> least broadly speaking inspired by the organisation of neurons in the
> |> brain.
> 
> A few people may have that aim, but the majority use the techniques
> for general optimisation and analysis and have little or no interest
> in how the brain actually works.  

There is no incompatibility here, one can maintain both claims
simultaneously. The majority of people can use "architectures that are
at least broadly speaking inspired by the organization of neurons in the
brain" for "general optimization and analysis and [yet] have little or
no interest in how the brain actually works".

> Also, remember that the current
> fad for 'neural network algorithms' PREDATES most of even our limited
> understanding of how the brain actually analyses data!

It is probably true that the preponderance of information about the
functioning of the brain has come after the advent of neural network
architectures. I am unclear about what you think follows from this fact.
 
> |> The idea of using
> |> networks of units with modifiable connections is itself an abstraction
> |> from the organisation of neurons in the brain anyway.
> 
> Sorry, but it isn't.  It predates that by a long way.  I can't tell
> you when the first self-modifying network systems were invented,
> but it was before my time - i.e. not after the mid-1960s, and I am
> pretty sure that it was in the 1950s.

Early neural network research did start in the 1950's (and I think
someone in this thread posted a date from the 40's). But I'm pretty sure
(and I could be wrong here so feel to correct me) that even in the 40's
we knew that the brain was comprised of neurons. I suspect this all
boils down to what one means by the claim "neural networks are inspired
by the architecture of the brain", especially with regard to the word
"inspired." How closely do neural networks have to be to brains before
we can say they were "inspired" by them, and how much do we have to know
about the functioning of the brain to make this same claim? Finally, I'm
not sure this line of inquiry counts against neural network research
anyway. Even if you are correct, and neural networks predate our
knowledge of the functioning of the brain, doesn't this mean that early
neural network research is even MORE impressive? If it predated it then
it was an implement able hypothesis of the functioning of the brain that
turned out to be kind of right (albeit painted in broad brush strokes).
This seems even more impressive to me than the typical stance, i.e. that
neural network research was inspired by the limited understanding of the
human brain that we had 50 or so years ago.

Just my 2 cents (and btw why isn't there a 'cent' sign on computer
keyboards? I used to own a typewrite that had one, so where the hell did
it go?)

Cheers,
Lew






More information about the Neur-sci mailing list