Context #1 (Spring,
1999, pp. 16-19); copyright 1999 by The Nature Institute
Programming the Universe: Are Animals Robots?
Stephen L. Talbott
Artificial intelligence researchers are fond of creating little robots
that scurry around on the floor and possess - or will soon possess, so
we are assuredthe wit of an insect. The assumption, often made explicit,
is that with each technical advance the intelligence of these devices
will ascend another step of the evolutionary ladder, finally approaching
the mental cleverness and versatility of man.
But this gets things exactly backward. The truth of the matter is that
it is much easier to program human intelligenceor, at least, certain
aspects of itthan to program anything at all of an insect's intelligence.
After all, we're the ones who have invented computers. Clearly we
can learn to pattern our thinking in step with the mechanisms of a computer;
we do this every time we write a programand therefore the computer
so programmed is patterned after our thinking. The execution of a computer
program must reproduce our thinking in some sense. But it's quite
a different matter with the beetle, who is presumably a long way from
being able to hire on as a software engineer.
What artificial intelligence researchers like about insects is their
supposedly simple, rule-based intelligence. This is odd, however, since
it is the human being, not the insect, who has gained the ability to think
in a rule-driven manner. We are the ones who not only speak, but derive
syntactic rules of speechand then admonish our children to "obey
the rules". We are the ones who not only analyze our experience, but tease
out of our analytic activity the logic of analysis. We alone formulate
volumes of legal code, rules of etiquette, protocols of office, organizational
procedures, New Year's resolutions.
"But even if the insect doesn't consciously formulate the relatively
simple rules of its behavior, isn't it obvious that it obeys rules?"
No, it's not. Here several things need saying.
The Commonsensical Beetle
In the first place, no one would be so foolish as to claim that an insect
obeys rules in any literal sense. Certainly there is no conscious obedience
going on, nor even a conscious apprehension of rules. At most one might
say that the insect has rules of behavior somehow "built in" to it, which
it must follow.
But this, too, is a misrepresentation. Yes, an insect often displays
a kind of rigidity in its behavior, and, yes, when we apply our narrowly
focused, abstracting intellects to this behavior, we may all too easily
reduce it to rules of thumb. Rules of thumb, however, are not the behavior
they summarize and reduce, nor do they adequately describe the behavior.
We learned this during the early days of artificial intelligence work,
when researchers confidently applied the same reduction to human intelligence.
Only slowly did they realize that our common sensewhich lies at
the opposite pole from the more highly conscious activity whereby we explicitly
formulate and obey ruleswas hopelessly beyond any imaginable collection
There is every reason to think that this must hold true all the more
for the intelligence of the insect, which surely lies even further toward
the "organic" and implicit end of the spectrum than does our own common
sense. Only the abstract, rule-bound activity of our intellects
allows us to think otherwise. We see the fly expending the last energies
of its life banging its head against a window pane, and we think, "What
a simple-minded program it is following!"
Do Reflexes Exist?
But the fly was not made for a world of window panes. Its powers of adaptation
are extremely limited. A lack of adaptability to a wholly alien world, however,
is not the same as simple-mindedness or rule-mindedness. Within its own,
infinitely complex and ever-changing world, the fly demonstrates an intelligence
that in many respects is more subtle and efficacious than our own. (For
an example of insect intelligence, see "The
Obscure Wisdom of the Potter Wasp" in this issue.) Even the behavior
at the window pane would reveal endless subtleties if we were to observe
it carefully enough: how does it vary with different light and reflective
conditions, different temperature conditions, time of day, the presence
of other insects or spider webs, different chemical gradients in the air,
and so on? The simple-mindedness of its "rules" is really the simple-mindedness
of our observation.
All this was put into clear relief by Kurt Goldstein in his several-decades-old
and decisively important book, The Organism (recently reprinted
with a foreword by Oliver Sacks). A neurologist, Goldstein looked at the
various ways we analyze organisms into rule-based, mechanical parts and
then try to reconstruct the whole from these parts. It never works. He
assesses the idea of the "reflex" in animals and shows in exhaustive detail
that the "simple-minded" reflex mechanisms we so easily imagine actually
don't exist. For example, slight changes in the intensity of a stimulus
can often reverse a reflex; a reflex in one part of a body can be altered
by the position of other parts; an organism's exposure to certain chemicals
such as strychnine can reverse a reflex; other chemicals can completely
change the nature of a reflex (a decerebrated cat will immediately swallow
any water placed on its pharynx, but will produce wiping movements of
the tongue instead if there is a little alcohol in the water); fatigue
can have the same effect; consciously trying to repress a reflex can accentuate
it (try it with your "knee-jerk" reflex); and so on without end.
Goldstein showed that the reflex is an artifact of our own stance as
researchers, whereby we conceptually and experimentally isolate one part
of an organism, cutting the part off from its whole. Moreover, he finds
that higher organisms, including human beings, are much more likely
to show approximations of reflexes, because it is we who can allow parts
of ourselves to become isolated and de-centered. (That's what many procedures
of medical assessment are all about.)
Human beings are able, by assuming a special attitude, to surrender
single parts of their organism to the environment for isolated reaction.
Usually, this is the condition under which we examine a patient's "reflexes"....But
[regarding the pupillary reflex] it certainly is not true that the same
light intensity will produce the same contraction when it affects the
organ in isolation (as in the reflex examination) and when it acts on
the eye of the person who deliberately regards an object....one only needs
to contrast the pupillary reaction of a man looking interestedly at a
brightly illuminated object with the reaction of an eye that has been
exposed "in isolation" to the same light intensity. The difference in
pupillary reaction is immediately manifest. (Goldstein, 1995, p. 144)
"I Am the World"
If we would approach the intelligence of the insect in its own terms, without
imposing our artificial requirements on it, then the fly on the window pane
usefully reminds us that the insect is inseparable from its environment.
The naturalist E. L. Grant Watson speaks of this when describing a newly
emerged mason-bee taking flight for the first time and alighting on some
chance object. It has no experience of the world as yet; how will it find
I ask it: "What do you know, young bee, of the world and the
universe?" It answers: "I am the universe." (Watson, 1995, pp. 142-43)
Our temptation is to take this as poetic shorthand for "the bee comes equipped
with a set of built-in rules that prepare it for what it will meet in its
normal environment." As we have seen, this is a highly problematic view.
What are the alternatives?
One alternative is to take Watson's remark at face value. The bee is
a whole, integrated within a larger whole; its intelligence is at the
same time an expression of the world's intelligence. What it meets in
the world is not something foreign, but itself.
Such statements will doubtless offend conventional scientific sensibilities.
And yet we are driven to them as soon as we try to understand conscious
intelligence, whether an insect's dull and "sleeping" intelligence or
our own focused and self-aware intelligence. You cannot speak of the unity
or coherence of any activities of consciousnessin fact, you cannot
speak of consciousness at allwithout speaking of qualities. Take
any state of your own consciousness and subtract all its qualities, and
you will have nothing conscious left. Without the qualitative experience
of, say, a tree, you won't even have anything from which to abstract quantities.
Now, the peculiar thing about qualities is their double nature. They
are "in here"part of the interior of the self (which is why science
has shied away from them). But they are also "out there" in the world
we share. That is, once you have discounted hallucinations and other purely
subjective contents, you are left with the phenomenal world we hold in
common. It remains qualitative through and through. Only our habits of
abstraction, and our willingness to mistake abstract residues for the
reality they came from, could convince us otherwise.
Qualities that are both in here and out there raise the question whether
the task of programming a beetle's consciousnessprogramming the
qualitative texture of its experienceis also a task of programming
its environment. It may be less true to think of the beetle as a centered,
executive agent responding by-the-rule to stimuli from the world than
to think of the world gathering itself (in all its qualitative beetleness)
to a focus in the insect.
That focus may be diffuse, perhaps vaguely analogous to the soft focus
of our dreams. In the human being, on the other hand, the gathering of
interior light has become intense and sharply defined, finally igniting
a flame within us, so that we, unlike the beetle, can throw light back
upon the world. We can understand and bring ever new things to pass. It
is not only that the world gathers and expresses itself in us; we express
ourselves through the world.
What Is It Like to Be a Program? (Not Much)
So the spectrum of conscious intelligence, running from "simple" organism
to complex, can be seen to reflect a shifting balance between a diffuse,
world-intelligence raying into the organism from without, and a bright,
focused intelligence partly raying back out into the world. This is a very
different matter from seeing the spectrum as reflecting an ever more complex
assemblage of rules.
What makes the latter view possible is the abandonment of the attempt
to understand conscious intelligence altogether. It is replaced by the
search for mechanisms that can more or less approximately reproduce particular
behaviors. And with this narrow, isolating focus come the pathologies
of understanding that Goldstein documents so devastatingly. Once you have
abandoned the only ground upon which unities and wholes can be recognizedand
that ground is irreducibly qualitativeyou will inevitably find separate,
well-defined mechanisms where there are nonecreating them if necessary.
And you will no longer care very much for the kind of qualitative observation
that alone can show you the limitations of your viewpoint.
It's no surprise, then, that the abiding problem at the center of all
efforts to create artificial intelligence remains as untouched today as
it was forty years ago: how do we program the qualitative texture and
content of consciousness? How, that is, do we program a consciousness
as experienced "from the inside" (and, of course, the experience of things
from the inside is what consciousness is.)
Is any organism ever precisely bound by rules in the strict, computational
sense? It would be a healthy thing if those attempting to program animal
behavior and intelligence would attempt to answer this question through
direct observation of animals in their natural habitat. What I think they
would find, if they were good observers, is that their rules became ever
more compromised, more qualitative, more artistic in form as the range
of their observation expanded, until finally they realized that they were
striving to experience the inner being of the animal, its meaning or expressive
unity, more than a set of rules.
It was long ago pointed out (by Thomas Nagel, 1980) that artificial
intelligence researchers have not been able to approach the qualitative
side of consciousness. They have not satisfactorily answered the question,
"What is it like to be a bat?" This is sometimes viewed as the "last frontier"
of artificial intelligence. The truth of the matter, I think, is that
it is the entire unsolved and unsolvable problem of artificial intelligence.
In the end, there is no other question to ask about an organism, except
"What is it like to be that organism?" All the "mechanisms", all the "reflexes"
and "drives"if we were to explore them fully enoughwould be
recognized as nothing more than stranded fragments of the answer to this
question. The organism is a unity, and the only way for anything to be
a unity is for it to be an interior shining throughintegrally expressing
itself throughan exterior (Barfield, 1977).
The interior is given in consciousness, and this consciousness can,
at one extreme, possess the soft focus that is more an expression of the
world than of the "separate" organism. At the other extreme, it can exhibit
the sharp, self-sustaining focus that enables usunlike the fly and
the beetleto detach ourselves from the world and thereby to make
both rules and mistakes. Many of our mistakes, it turns out, have to do
with projecting our own, lately achieved, rule-following consciousness
upon the living dynamic of the world from which we have detached ourselves.
Stephen L. Talbott is a senior researcher at The Nature Institute and author
of The Future Does Not Compute: Transcending the Machines in Our Midst.
Barfield, Owen (1977). "Science and Quality,"
The Rediscovery of Meaning and Other Essays. Middletown, Conn.: Wesleyan
Goldstein, Kurt (1995; originally published in 1935).
The Organism. New York, N.Y.: Zone Books.
Nagel, Thomas (1980; originally published in 1974).
"What is It Like to Be a Bat?" In Readings in Philosophy of Psychology,
vol. 1, edited By Ned Block. Cambridge, Mass.: Harvard University Press.
Watson, E.L. Grant (1995; originally published in
1964). The Mystery of Physical Life. Hudson, N.Y.: Lindisfarne
Original source: In Context (Spring, 1999, pp. 16-19); copyright
1999 by The Nature Institute
Steve Talbott :: Programming the Universe: Are Animals Robots?