death of the mind.

Sergio Navega snavega at intelliwise.com
Tue Jul 20 15:45:15 EST 2004


"David Longley" <David at longley.demon.co.uk> escreveu na mensagem
news:C$r7w1IXxV$AFwwZ at longley.demon.co.uk...
>
> You haven't *really* addressed my question. You suggested that this
> technology should unsettle Glen. But people like Glen and I have spent
> time "peeking" at the brains of animals by inserting cannulae or
> electrodes like many other behavioural neuroscientists and recording
> animals' operant behaviour (neurosurgeons have done it quite directly
> for decades). What's new about fMRI etc is that it makes some of this
> easier, especially for those who don't know enough about brain and
> behaviour to look at it with circumspection. It misleads a lot of
> people, especially those cognitivists who are undisciplined mentalists
> or intensionalists in my view.  The technology is just a new technology
> in a long line of technologies and one of my tacit points was that when
> others here have said things similar to what you said in your post, it
> has just betrayed their (all too common) naive conception of radical or
> evidential behaviourism. When a person makes a verbal or other response
> when someone is looking at an fMRI image etc, how does that
> fundamentally differ from when someone records what a field electrode
> etc picks up in a freely moving animal on some schedule or other
> 'behavioural assay'. It  doesn't unless one is a closet mentalist
> looking for "meaning" in what the subject says or does. This is what
> almost all "cognitivists" including "cognitive neuroscientists".  People
> (including themselves) are often just seduced by their proximity to
> physical "brain talk"!

My comment that follows will perhaps use an often repeated argument,
but I see no alternative at the moment. An AI researcher may
eventually be interested in understanding human behavior, but
what is essential to such a person is to comprehend the *mechanisms*
behind such behavior, because his/her task is to devise an
artificial mechanism capable of performing (behaving) similarly.
Thus, this leads to what I think is an overexposed argument: the
hardware/software question. For the sake of improving the idea,
let's consider a group of aliens which steal my computer. They
will do anything to understand how it works, because they want
to build their own.

The "hardware aliens" will analyze the boards, chips, cables
and connectors, trying to look for essential principles of
operation. The behaviorist aliens will annotate all reactions
of the computer to given stimuli. The "abstract aliens" will
notice patterns of behavior of that machine. But instead of
just annotating these patterns of behavior, the abstract
aliens will, after some time, start to hypothesize that the
windows showing in the monitor of the computer seem to be
individual instances of programs (which are, by themselves,
abstractions). From this simple hypothesis (that was developed
because of experimental observation) they will *deduct* that
it is necessary to have a central program which coordinates
how much processor time is spent in each of the windows.
The real value of such a deductive model is to be capable of
giving us some *predictions*. One prediction of this abstract
model is that if one opens too much windows, the speed of
the program in each window will be reduced (because of
the shared processor hypothesis). One can experimentally
check this prediction, and if this doesn't correspond to
what is measured, then the model is *wrong* and should be
rejected.

To the behaviorist aliens (a funny idea indeed...) it will
reject the notions of "operating systems", "time slices",
"protected data spaces", "interrupt service routines",
"high level languages", and a lot of other abstract concepts
created by the cognitive aliens. They dismiss these ideas
based solely on the fact that such things don't show directly
as behavior, being just figments of a cognitive alien mind
(sorry, brain).  However, these concepts are essential for
anyone doing creative development in software engineering.
Without entertaining such notions, one would hardly get a
deep vision of all one can do with software. In other words,
if Bill Gates was a behaviorist, he would be selling orange
juice at fifth avenue.

Although the software/hardware distinction is a bad analogy
for the mind/brain distinction, the idea is to have different
levels of analysis, provided that one obeys basic scientific
practices in all these levels.

Sergio Navega.


P.S: As an aside, let me mention a passage that I find hilarious,
although it is a bit against my prior argument. I read this as
an introductory quote to a paper about the use of metaphors and
analogical reasoning (and, of course, its misuse). This is the
situation: a physicist was invited to give a speech to a group
of simple farmers interested in improving the yield of milk of
their cows. Here's how the physicist started his speech:
"Let's start by considering a spherical cow..."






More information about the Neur-sci mailing list