Re: a priori (was Re:Contradiction in Rucker?)

Eric Watt Forste (arkuat@pobox.com)
Tue, 15 Jul 1997 16:07:28 -0700


Roderick A. Carder-Russell writes:
> I'm not certain that "hardcoded into our brains by the genes" is
> quite correct. Certainly there are some types of knowledge that
> can be had only by the intention of evolution (I use that loosely,
> I'm not promoting a deterministic view), such as walking. We have
> the knowledge, we could say, a priori, and just have to learn to
> utilize it. There are some bodies of knowledge, however, that
> arise in us naturally and not through sense data simply by the
> "structure" of the brain. Now given, the brain is determined by
> biology, but the information that it percieves can in some instances
> arise by the _very nature_ and structure of the brain, not by the
> genetic information composing it. Mathematics is one of these
> bodies of knowledge, I don't believe that it was written in our
> genes to discover mathematics, but rather it arose naturally as a
> benefit of the way in which our minds percieve. Seeing in three
> dimensions is another example, if somewhat more "physical". It
> isn't "hardcoded", but rather is by the very structure of our eyes
> and brain to percieve this way (imagine percieving naturally in
> other dimensions...).

"Hardcoded into our brains by the genes" was, perhaps, an overstatement.
I certainly did *not* mean to imply that I think that the Human
Genome Project is going to stumble across the genes for "structural
perception of space". That's just barely possible, only because it
seems we still know so little about human ontogenesis, but seems
unlikely to me.

But our brains do have a physical structure that allows for the
development of a neurocomputational machine with properties similar
to, but much more sophisticated (and kludge-ridden) than the
artificial neural networks people have been playing with since the
perceptrons, and much more since the late 1980s. And I suspect
that many of the "a priori" information structures are going to
turn out to be implicit in that style of computation. What our
genes code for is the production of tissue that can carry out that
kind of signal analysis, computation, and output motor-vector
generation. It codes for the neural architecture, indirectly, via
the proteins and the control genes that determine the sequence in
which proteins are constructed and the stimuli which will produce
the construction of other proteins. And the neural architecture
has certain "best ways" of dealing with the surrounding environment,
many of which were probably "discovered" and propagated culturally,
possibly even before the rise of language. Sorting out the genetic
from the cultural takes us right into the messiest possible issues
in animal ethology.

It's all a jolly mess, and I bet we're going to have a lot of
fun sorting it all out.

> This is a very materialistic view, or perhaps I should say
> biologically dependent view.

I guess I'm a methodological materialist. If something is to be
explained by an information structure, I still want to have some
clue about which atoms encode that information structure. In
most philosophical questions, the atoms that encode the
interesting information structure are in the human nervous system.

> A good deal of knowledge is not dependent upon the biological
> entity percieving it.

Well, remember what I was saying about how a posteriori
knowledge can be seen as the fruit of perception by the
organism, and a priori knowledge can be seen as the fruit of
"perception" by the species gene-pool. It's a very odd way to
use the word perception, certainly, and if I restricted myself
only to standard usage, I'd have to agree with you. But given
this genetic-algorithmish semantic extension of the word
"perception" (which is due to evolutionary epistemologists such
as Donald Campbell), I'd want to claim that all knowledge
(actual physically encoded information structures in people's
nervous systems that we use as knowledge) has its causal origins
in "perception" either in the conventional "small" sense or in
the new, unconventional "large" sense.

I tend to be of the opinion that no knowledge existed at the
time the first replicator spontaneously formed. Knowledge is
what is left over after gross falsehood gets stomped by the
environment. Or at least, that's a hypothesis I entertain from
time to time. ;) Here I'm using a rather strict, small,
physicalistic sense of "knowledge" as something that is encoded
within or can be communicated into (as in, reading a book or
something) a physical organism. I know it's odd for me to use
such a restricted sense of the word knowledge when I've just
gone and tried to drastically expand the sense of the word
perception. But if epistemology were a done deal, it wouldn't be
so interesting.

> But biological entities can have a priori knowledge of eternal
> truths, eternal throughout our universe, not by genetic intent
> (again, I'm not implying a telos) but by the very structure of
> their percieving. In other words, we can have knowledge of eternal
> truths by virtue of having the structure that we have, the genes
> evolutionary purpose isn't necessarily to enable an understanding
> of these truths.

I don't know what the phrase "evolutionary purpose" means. I'm not
saying no such thing exists, I'm just saying I'm unfamiliar with
it. The very structure of our perceiving is a causal consequence
of our ontogenesis. Perhaps I'm being a "causality dogmatist"
here... But purposes, it seems to me, are properties belonging to
the individual organisms that arise within the Big Genetic Algorithm
Soup, not properties of the Soup itself. I really am groping in
the dark here and merely spouting the opinions that get past my
filters, so don't make too much of this.

> Indeed, there _must_ be some innate or a priori knowledge,
> to enable initial perception.

If you're talking about a human individual's perception, yes. But
if you're talking about the first physical event of perception in
the cosmos, this doesn't square with our knowledge of biological
evolution. At some point, the first animal perceived something.
The "a priori" knowledge that it had at that time, which enable
its individual perception, lay in the structure of its nervous
system and its sense organs, which grew by an incredibly kludgy
and delightful process from its genes and its encounters with its
environment during its growth. Its genes evolved by a process of
random mutation, accurate replication, and selective elimination,
from those of the first (presumably "nonperceiving" in the narrow
sense) replicators. Again, the whole argument hinges on Campbell
and Waechterhaueser's extension of the word "perception" to mean
something that gene-pools can do to their environment over long
periods of time too, and *not* *only* something that animals can do
to their environment over short timespans.

To take it further and make the metaphor more concrete, I'd have
to have a lot more programming experience with genetic algorithms.
But to someone trying to make a GA robot that can navigate an
environment using a GA perceptual algorithm (I'm not sure whether
anything like this has been attempted by anyone yet), I suspect
that these ideas would seem quite concrete indeed.

--
Eric Watt Forste ++ arkuat@pobox.com ++ expectation foils perception -pcd