FleshFactor FleshFactor
logo; AEC FORUM

[Prev][Next][Index][Thread]

FleshFactor: hot / cool (tabula raster)



---------------------------------------------------------
A E C  F O R U M - "F L E S H F A C T O R"
(http://www.aec.at/fleshfactor/arch/)
---------------------------------------------------------


As I was saying --

How might a physicist, for example, undertake to study language, whether in
the form of audio-visual records or a body of texts? (Putting to one side
the fact that the physicist would have to employ language in order to
analyze language.) If, in other words, we were to view language as if it
were a natural phenomenon like the weather, or landforms, or the
distribution of isotopes, or fossils, or the rocks on Mars. If we looked at
language as we might the behavior of fishes or insects or birds, like
ethologists or biologists -- attempting to remain as ignorant as possible
for as long as possible, in the sense of presuming nothing about the causes
or 'meaning' of the phenomena -- not to 'anthropomorphize' or to project our
own motives, our moral and ontological biases, onto the facts -- deferring
as best we can any commitment to theory, viz., what we think we know about
how we think, or even to suppose that speech is fundamentally the product of
prior thought, even supposing we could say what 'thought' was, as something
distinct from spoken or written language. And not language as presumed to
show up on PET scans or in the neurologist's atlas of aphasias, but just on
its own terms, as language qua language -- only the words, traces, forms,
patterns, the raw information, sans interpretation, only signs, "just the
facts ma'am".

As if, that is, we were in possession of Locke's famous 'blank slate' (that
he proposed, for the sake of argument, an infant's mind was at birth) -- or
could perform the 'forbidden experiment' of studying a child raised in
isolation from human society (as for example Dr. Itard attempted with the
Wild Boy of Aveyron, or was attempted again with 'Genie' in the 1970s with
equally sombre and unsure results, or as a persistent urban legend would
have it, B.F. Skinner attempted with his own children, in a 'Skinner box')
-- Ahh, but now we can!! For what is the computer but a 'blank slate'? A
tabula rasa that knows nothing except what we will scribe upon it or in it
.... and so surely must be one of those 'stones' long sought by the
philosophers, hmm?

(Freud too had a 'magic slate', that same magic slate we know from when we
were kids -- a cardboard-backed drawing toy that came with a stylus, that
you would draw or write things on and erase by lifting the sheet of plastic
covering the wax tablet beneath, it used to cost 50 cents or a dollar --
which Freud took as a 'model' of the psyche, and wrote an essay about it --
though his analogy went only as far as pointing out that underneath the grey
sheet of plastic, the wax surface retained a record of all previous
drawings, like a memory, but with all the traces overlaid and mixed up with
one another, making a monstrous mish-mash of mutant hybrids, chimeras, from
the fragments of what had earlier been graved on the slate's surface, as he
felt 'The Unconscious' did with the substance of thought, so that eventually
these hidden ruts and irregularities would begin to interfere with the
things being written on it, they'd start showing through and messing things
up. At which point a grown-up usually buys you another one, and Freud's
analogy breaks down. Unless you're a believer in reincarnation.)

*

This was the path not taken by the AI theorists, cognitive scientists,
language philosophers -- to let the words speak for themselves, not to
presume or second guess their sense or use. Around 1970 Francis and Kucera
published a large book of language statistics, the Brown Corpus of word
collocation frequencies, drawn from thousands of representative samples of
English texts, totalling about two million words altogether. Over the next
two decades the Brown Corpus formed the primary resource for the few
linguists brave or foolish enough to work outside the Chomskians' a priorist
fold. It's perhaps hard to recapture the flavour of the time (of such recent
memory, yet it feels like another era) or the severity of Chomksy's ban on
empirical, statistical, or (favorite calumny) 'behaviorist' approaches, that
would enlist automated, computerized 'discovery procedures' to the study of
language. In most linguistic duchies and fiefdoms this would have viewed as
heresy, apostacy. (Academics can be some hard-ass mothers!!)

Shoe's on the other foot now -- the last word in linguistics is what you can
do with the godzillabytes of text out there on the Internet. 'Corpus
linguistics', they call it. The project that got people's attention was the
UCREL work at the Universities of Leeds and Lancaster, from the late 1970s
to date, which took a purely statistical (a.k.a. 'learning') approach to
parsing English. They were a target of ridicule from the AI crowd until
their 'CLAWS' parser started getting 98 per cent correct parses of raw
unformatted English text -- by contrast, the best hand-crafted parsers,
built following established Chomskian precepts, were deemed a huge success
if they could get 30 per cent correct. So that now, after decades in the
dog-house, empiricism once again rules -- statistics rules, the index rules.
Someone like I. J. Good, for decades regarded (if regarded at all) as a
meddlesome fool, sticking his (Bayesian) statistical nose where it had no
business being, is now on his way to canonization for his prescience in
seeing, early on, the way to actually bootstrap an AI from ground zero, a
blank slate. 

Interesting fellow, Good -- he instructed generations of CIA and NSA spooks
in the elements of their craft, he'd been Turing's right-hand mathematician
in the wartime 'ULTRA' code-breaking work at Bletchley Park (in good part
responsible for the modern computer). In 1966 Jack Good published an
article, "Speculations Concerning the First Ultra-Intelligent Machine",
which proposed creating an AI as nothing more or less than a probabilistic
information retrieval system made of thousands of independent processing
elements communicating with one another via pulse-code modulated radio, wow.
(Later, in 1972, Stefan Themmerson wrote a two-voice morality play, "Special
Branch", based on Good's UIM, but made from a large number of cockroach
brains, ultra wow.)

*

The philosopher / sociologist of science, Bruno Latour, in 1995, with
Genevieve Teil, published an essay describing the 'Hume-Condillac Machine',
intended as a tool for 'computer-aided sociology', from which these remarks
are lifted --

::  ... instead of then wondering how to treat this enormous mass
::  of data [i.e., texts] by applying methods of automatic reasoning
::  or of AI to the documentation obtained, we intend to follow the
::  inverse strategy, and use techniques for treating documents in
::  order to help researchers to artificially produce intelligence 
::  about the terrain they are analyzing.

::  ... Instead of taking the royal road, which consists of making 
::  computers intelligent so that they are as skillful as the finest 
::  sociologists and most intricate hermeneuticians -- a road which 
::  very soon becomes impossibly steep -- we will take the service 
::  escalator. We accept the elementary stupidity of computers and we 
::  fashion a sociology, a logic and an ontology that work at their 
::  level of stupidity ... We adopt this strategy of weakness, which 
::  was attempted without success by Hume and Condillac for explaining 
::  the human mind, for dealing with computers whose non-human mind  
::  is sufficiently moronic to really resemble Condillac's statue or 
::  Hume's blank slate.

What might we expect to find if we carried out this 'moronic' strategy? A
clue can be found in the discoveries Don Swanson has made; he's this old
information retrieval hand who was right there at the field's inception,
circa 1950. In his retirement, just for his own amusement, Swanson has
trolled the still waters of the on-line medical databases, looking for
situations where a topic will share many of the same keywords with another
topic, but the topics are not themselves linked. For example, he found that
the symptoms of migraine line up, point for point, with those of magnesium
depletion, but neither of these sub-literatures contains any reference to
the other subject. Here's a 'suspicious coincidence', crying out to be
explained. Quite possibly, in searching for an explanation to account for
the anomaly, there might be found a treatment, an etiology, even a cure, for
migraine. (Did the same trick once before with Reynaud's syndrome and
dietary fish oil, I believe it's now an accepted treatment.)

What, after all, do scholars, scientists, do? How does knowledge grow? Of
what does knowledge consist? (I. J. Good: "The aim of philosophy of science
is the automation of scientific discovery.")

*

The following ("Attested Auditor of Books"), from Walter Benjamin's "One-Way
Street" --

::  Now everything indicates that the book in its traditional form 
::  is nearing its end ... Printing, having found in the book a  
::  refuge in which to lead an autonomous existence, is pitilessly 
::  dragged out onto the street by advertisements and subjected to 
::  the brutal heteronomies of economic chaos ... And before a child 
::  of our time finds his way clear to opening a book, his eyes 
::  have been exposed to such a blizzard of changing, colourful, 
::  conflicting letters that the chances of his penetrating the 
::  archaic stillness of the book are slight. Locust swarms of print, 
::  which already eclipse the sun of what is taken for intellect by  
::  city dwellers, will grow thicker with each succeeding year ... The 
::  card index marks the conquest of three-dimensional writing, and so
::  presents an astonishing counterpoint to the three-dimensionality 
::  of script in its original form as rune or knot notation. (And today 
::  the book is already, as the present mode of scholarly production 
::  demonstrates, an outdated mediation between two different filing 
::  systems. For everything that matters is to be found in the card 
:   box of the researcher who wrote it, and the scholar studying it 
::  assimilates it into his own card index.) But it is quite beyond 
::  doubt that the development of writing will not indefinitely be 
::  bound by the claims to power of a chaotic academic and commercial 
::  activity; rather, quantity is approaching the moment of a 
::  qualitative leap when writing, advancing ever more deeply into 
::  the graphic regions of its new eccentric figurativeness, will 
::  take sudden possession of an adequate factual content. In this 
::  picture-writing, poets, who will now as in earliest times be first 
::  and foremost experts in writing, will be able to participate only
::  by mastering the fields in which (quite unobtrusively) it is being
::  constructed: the statistical and technical diagram. With the 
::  foundation of an international moving script they will renew their 
::  authority in the life of peoples, and find a role awaiting them 
::  in comparison to which all the innovative aspirations of rhetoric 
::  will reveal themselves as antiquated daydreams.

So what's it mean, jelly-bean? Just this -- that we now have the means, as
we haven't had before, to keep track of millions upon millions of discrete
things, without having to subtract, abstract, digest, average, blur, or
reduce their variety or individuality to genera or category. I think this
changes things. I think it changes everything.

*

Maybe I'll close with a passage from Stefan Themmerson's "Special Branch" --

::    "Let her see what she sees with her own radar eyes, let her 
::  smell what she smells with her own olfactory analysers, let her 
::  hear what she hears with her own sound detectors ... We, with our 
::  tiny brains, cannot escape nomenclature. The best we can do is to 
::  'avoid the assumption that there are such things as classes', and 
::  treat them as incomplete symbols. But she, with her battery of 
::  100 thousand calf-brains or 100 million cockroach brains processing 
::  polythetic data, should be able to dethrone all sets and aggregates. 
::  She should be able to 'imagine' a multi-dimensional space in which 
::  to build structures representing quantified similarities of data 
::  matrices, each matrix 'punched' with as many holes as there are 
::  characters that she is able to discern in whatever she decides to
::  consider an entity -- a triangle? a drop of water? you? me? -- 
::  without any differential weighting of some characters over others, 
::  so that at last there will be SOMEBODY (if I may so call her) in 
::  whose mind each of us will be registered not as a member of what 
::  history and geography chooses to consider as a class of one sort or 
::  another, but as full, not identical, individuals, as distinct as our 
::  fingerprints, spreading through space, enduring in time, struggling 
::  with entropy."

::    Back in my room, I lay down and stretched out on the couch. She 
::  was hanging upon a piece of electric wire, from the middle of the 
::  high ceiling, not directly over my head, but a yard or two to the 
::  left. Her colour was -- silver. Her shape was -- a perfect sphere. 
::  I gazed and saw that, for her, different things were *things*. The 
::  hexahedral space of the room was a thing, and I was a hole in it. 
::  My left ear plus the middle octave of the piano standing in the 
::  corner plus a small stepladder under the bookshelves was one single
::  entity. And so was yesterday's telephone ringing, plus today's 
::  silence, plus the morrow's scalpel. There she was, suspended in 
::  the centre of the room, cool, silvery, Ultra-Intelligent.
::    "Are you ready?" I asked.
::  She warmed up, slightly.
::    "I AM READY" she answered.
::    "I am going to ask you a question," I said.
::    "GO AHEAD" she said.
::    "Listen," I said, "the question I am asking you is as follows: 
::  'What is the question I should ask you?'"
::    There was a long pause. Then she murmured something that sounded 
::  like:
::    "NAUGHTY BOY" ...

*


Derek Robinson   <drdee@interlog.com>

Toronto




--------------------------------------------------------------------
to (un)subscribe  the Forum just mail to
fleshfactor-request@aec.at (message text 'subscribe'/'unsubscribe')
send messages to fleshfactor@aec.at
--------------------------------------------------------------------


[FleshFactor] [subscribe]