2003-09-06 14:48
Part the First: By way of introduction.
Spicy brains and ruthless robots! Recent experimental evidence from
new-born babies and timeless philosophical questions! Phenomenology,
the limits of empathy and Artificial Intelligence!
I think that philosophical hostility to the possibility of Artificial
Intelligence is mostly the result of a misunderstanding, and this post
is my attempt to straighten things out. It also includes many
quotations from other persons (including many non-philosophers), used
both to help advance my argument and to entertain such persons as are
not convinced that I am one of the foremost philosophers of our age.
Part the Second: Ph�nom�nologie I, oh l� l�
Although I have mostly been trained in the physical sciences,
mathematics and engineering, my philosophical allegiance is not to the
Anglophone tradition of empiricism, but the continental school of
phenomenology. This does cause some occasionally cause some minor
problems - e.g., I am obliged to take Derrida seriously (or at
least not to dismiss him out of hand) - but the worst of them is that
most of Anglophonia doesn't have any idea what phenomenology is (not
least many of the Literary Theoristes who mistake themselves for
admirers of Mr Derrida).
Let us begin, then, by enquiring of a French person with a silly name
what sort of undertaking or enterprise it is that we are engaging
with. Maurice "Maurice" Merleau-Ponty gives an answer (or rather an
app�ratif to a sketch of an answer, which is much more the
Continental way, hurrah!) to the question, "What is phenomenology?" in
his classic The Phenomenology of Perception:
It is a matter of describing, not of explaining or
analysing. Husserl's first directive to phenomenology, in its early
stages, to be a "descriptive psychology" or to return to the "things
themselves", is from the start a foreswearing of science. I am not
the outcome or the meeting point of numerous causal agencies which
make up my bodily or psychological make-up. I cannot conceive of
myself as nothing but a bit of the world, a mere object of biological,
psychological or sociological investigation. I cannot shut myself up
within the realm of science. All my knowledge of the world, even my
scientific knowledge, is gained from my own particular point of view,
or from some experience of the world without which the symbols of
science would be meaningless.
[...]
To return to things themselves is to return to that world which
precedes knowledge, of which knowledge always speaks, and in
relation to which every scientific schematization is an abstract and
derivative sign-language, as is geography in relation to the
country-side in which we have learned beforehand what a forest, a
prairie or a river is.
This groundedness in pre-theoretical experience is something that
empiricist philosophers are careful to ignore, but I don't think they
have succeeded in making it go away. It is the distinction of
ph�nom�nologie to return scrutiny to our manner of being in a world
which is always already there.
The way that ph�nom�nologie priviledges a self - which I'm going to be
calling ego, but not in any Freudian sense - as constitutive of
the universe may seem like it is fraught with danger of solipsism - of
denying there is anything beyond subjective, and analytic philosophers
like nothing better than to tease passing ph�nom�nologiste's about
just this. But the problem of solipsism is a really just a
pseudo-problem that only philosophers pretend to have - I discard it.
(Philosophy is too often the art of giving unsatisfactory solutions to
problems nobody actually has.)
It is true that ph�nom�nologie renews the force of Leibnitz's primary
philosophical question, "Why is there anything at all and not
nothing?", but it is also the case that it tries to resist mistaking
it for the kind of question which could be answered. This kind
of priviledging of the status of ego may look to be
problematic for a mechanistic theory of consciousness in which it
we are a fancy kind of machine made out of meat with extra-spicy
brains, and for lots of different claims about the possibility of
artificial intelligences.
Since I am both a ph�nom�nologiste and a computer programmer with a
long-standing dislike of theories that seek to prove AI is impossible
a priori I have a vested interest in a reconciliation, and this
monsterpost is a preliminary sketch of how I think this can be achieved.
Part the Third: Of purpose and biology
My favourite ever linguist, "Roamin'" Roman Jakobson, starts his
glorious survey booklet Main Trends in the Science of Language
by insisting on the role the phenomelogical thought of Husserl and his
school played in the origins of European structuralism in linguistics,
and later finds himself talking the importance of some kind of concept
of "purpose" in accounts of biological systems, which he is planning
to use to hit some behaviourists over the head with, and I have
shamelessly raided his quotes from eminences.
George Gaylord Simpson (eminent Hahvard biologiste):
"The physical sciences have rightly excluded teleology, the principle
that the end determines the means, that the result is retroactively
connected to the cause by a factor of purpose, or that usefulness is
in any sense explanatory. But in biology it is not only legitimate
but also necessary to living organisms of everything that exists and
occurs in them."
According to Fran�ois Jacob's witty comparison, "For a long time the
biologist approached teleology as a woman he was unable to dismiss but
in whose company he was unwilling to be seen in public. At present
the programme gives a legal status to this secret liason!"
[Emerson] sees no necessity "to put the word purpose in
quotation marks" and maintains that "homeostasis" and goal-seeking are
the same thing.
Berns^tejn ["The leading Russian biologist of our time"]
Such a formulation of the biological "purpose" requires no
psychologisation.
We're including plants in this, after all, and it's not often
claimed that plants are conscious. (Bad hippies! Shoo!)
The term "teleonomy" has been proposed as a drop-in replacement for
"teleology" that doesn't scare the horses, and here's Jacques Monod
to give it a workout:
Monod describes the central nervous system as "the most evolved of
teleonomic structures" and ventures to interpret the emergence of the
superior specifically human system as a sequel to the appearance of
language, which changed the biosphere into "a new realm, the
noosphere, the domain of ideas and consciousness." In other words,
"it is language which created humans, rather than humans language."
The point to remark upon is that it becomes desirable to make
use of descriptions incorporating teleonomy as soon as one is dealing
with biological systems; you could explain a plant in terms of physics
and chemistry - certainly, nobody would take up cudgels at your
demeaning of vegetative dignity (I thought I told you to clear off,
hippies? I won't tell you again. Shoo!) - but biology implies
homeostasis, and this implies the utility of teleonomy as a component
of description.
But if we are going around promiscuously bestowing teleonomy on
biological systems, and we can classify organisms by the complexity of
the teleonomic resources at their disposal from plants (and below I
shouldn't wonder) through insects, birdies, puppy dogs and then -
bestriding the heirarchy like a teleonomic colossus - glorious
personkind itself, it still falls short of paying what many would
consider to be adequate homage to the glories of the Human Spirit.
I'll come back to this, once we've seen how much worse it could be.
Part the Fourth Scientisme, tut tut.
Again, swiped from Jakobson. (Once you make it through the first
noun-phrase it's plain sailing, honest.)
Recidives of superstitious fear of a means-end model which still
torment a few linguists are the last survivals of a sterile
reductionism. As a characteristic example we may cite a linguist [C F
Hockett]'s affirmation that "in the discussion of man's place in
nature there is no place for mentalism", since "man is an animal and
subject to all the laws of biology," and, finally, that "the only
valid assumption is that of physicalism," since "life is part of the
inorganic world and subject to all the laws of physics."
(I don't know about you, Varied Reader, but behaviourists give me the
fucking creeps. Nonetheless, Google helpfully debunks one
popular but unfair myth:
To the end of his life Skinner was plagued by rumors about his second
daughter, hearing even that she had committed suicide. In fact,
Skinner was an affectionate father and never experimented on either of
his children. Deborah is a successful artist and lives in London with
her husband.
Although I suspect that meme has long since bolted.)
We have seen above that biologists differ from physical scientists in
that they are happy to use teleonomy as an ingredient in their
explanations. This doesn't imply that they are attributing some kind
of mysterious "life force" to biological systems that transcends
physics, oh my. We are not also saying that a plant can't be
described purely in terms familiar from physics, just that such an
explanation is an eccentric form of physics rather than any kind of
biology. With pedantic exactness we may say, any biological question
that can be translated into the language of physics (atoms and forces
and stuff) can then be investigated within that framework and the
conclusions when translated back into biological terms will be
consistent with what biologists observe. The concept of "organism"
belongs to biology - after all, what reason does physics give to
bundle that particular collection of molecules together? - and once
you have organisms, teleonomy is coming along for the ride.
(Avid Husserlians will note that I am here claiming that teleonomy is
part of the eidos of biology, and Heideggerians may be inclined
to start wondering how far this "teleonomy" correlates with what they
call "being-in-the-world".)
Within the empiricist tradition there is a tendency to insist that
physics - the physical science dealing with the smallest things - is
epistemologically fundamental, and that all accounts must be
accountable to physics. But the claim that the truth has to be
established empirically at all is itself not something that can be
established empirically, so it falsifies itself. Husserl's critique
of the malignant metaphysical reification of empirical truth he called
"scientism" starts from here and he then insists returning attention
to the (phenomenological) subject for whom an explanation exists.
You might think that it would be uncontroversial to claim that a
theory must be a theory of something for someone, and that some
kind of pre-theoretical knowledge of the something in question (say,
apples falling out of trees) is a prerequisite for wanting or
understanding an explanation of it (Newtonian theory of gravity?
Einstein's theory of gravity? Superstrings? Something else? How
many innocent apples must fall before somebody decides?), but
you'd be reckoning without the tenacity of the empiricists.
One of the reasons that this kind of talk annoys them, of course, is
that if you make the phenomenological subject the priviledged source
of science, then science can't in turn explain the phenomenological
subject. Empiricist philophers therefore go to a great deal of
trouble to try to provide a physical basis for subjective experience,
which in practice means going on and on about things called "qualia"
(our internal sensations - the kind of experience of redness you have
when looking at a red thing, as distinct from an account of retinas
and cones and stuff) and how the brain (which is a Proper Physical
Object and thus a Good Thing) can possibly give rise to them.
They have not, you will be astonished to hear, got anywhere.
Part the Fifth, Phenomenology of Persons
Given that the phenomenological subject (ego) finds the
existence of the universe mysterious, you might think that it would
find the existence of other people especially mysterious.
Hold that suspense, though, while I introduce some new terminology.
Let's refer to a person other than ego as tu. Now, if
we are hard-core empiricists we will want to insist on a explanation
of tu in terms of categories and concepts used in physics, and
we will find ourselves being behaviourists. If on the other hand, we
prefer an explanation in terms of categories and concepts used in
biology, we can at least include teleonomy in our account, even if the
hardest-core behaviourists don't much like it. I shall refer to
biological tu and physical tu as necessary to
distinguish the levels of corresponding levels of explanation.
Now, if you want something fancier than either flavour of tu
(even one that's teleonomically more sophisticated than a plant) then
you're going to have to change your explanatory framework to be
something other than physics or biology. In particular, let's
consider and you a phenomenological framework instead. Recall that we
rejected empiricist accusations that our phenomenological
ego is solipsistic by simply refusing to care, since nobody
actually, really suffers from solipsism. (And anyone who does isn't
going to benefit from an appointment with an epistemologist, that's
for sure.) Why don't we try the same sophisticated procedure with the
alleged problem of inferring the existence of another person as
alter-ego - which is to say, the kind of entity that
experiences itself as ego in its own right?
Merleau-Ponty, for example, starts his account of such
inter-subjectivity (i.e., how we recognise other people as
people) from what he calls �La socialit� originaire�, which is a
fancy way of saying "WELL WE JUST DO, OK?" which is either obviously
the correct answer or a disgraceful cop-out according to taste. And
he goes on:
En r�alit� le regard d'autrui ne nous transforme en objet que si l'un
et l'autre nous nous retirons dans le fond de notre nature pensante,
si nous faisons l'un et l'autre regard inhumain, si chacun sent ses
actions non pas reprises et comprises, mais observ�es comme celle d'un
insect.
In reality being looked at by someone else does not turn us into an
object unless we completely withdraw our thinking nature from each
other and we give one another an inhuman look, if each feels his
actions not reciprocated and understood, but observed like those of an
insect.
Note, though, that this is what behaviourists claim to do (at least
while on duty) and there is no reason to think they are breaking any
laws of physics. The experience of recognition of the other as an
ego in its own right - an alter-ego - is grounded,
phenomenologically, in empathy, and to decline to do so is an
ethical offence.
You will not need, I hope, to be told that the distinction between
those entities which are considered to qualify for alter-ego
status and those which are not is at least partly a cultural question.
Certainly, an 18th century slave-owner and a 21st century animal
rights activist would draw the line in different places. (Remember
PETA's holocaust on your
plate campaign?)
How about foetuses ("fetuses") and abortions? How about wimmins and
suffrage? (Even in Switzerland wimmins have been allowed to vote
since 1971).
Or how about this:
I'm ready, for the time being at least, to argue that an asocial
individual isn't fully human.
Wherever you draw the line, I claim that the relationship between
alter-ego and biological tu is analagous to that of the
biological and physical tus described above: anything that you
can state about alter-ego that can be translated into the
language of biology can be investigated in a biological framework and
the conclusions will be consistent with what we observe of others. To
put it bluntly, I claim that empathy does not violate any laws of
biology or physics, but also that it is not itself a biological or
physical phenomenon.
Part the Sixth A digression on Searle.
Early in his book The Mystery of Consciousness the philosopher John
Searle states:
If there is one theme that runs throughout this book it is this:
consciousness is a natural biological phenomenon. It is as much a
part of our biological life as digestion, growth or photosynthesis.
(I don't personally know anyone who photosynthesises, but I haven't
spent much time in Ivy League philosophy departments either.)
A little later he describes what he calls a "common sense" notion of
consciousness:
"[C]onsciousness" refers to those states of sentience and awareness that
typically begin when we awaken from a dreamless sleep and continue
until we go to sleep again, or fall into a coma or die or otherwise
become "unconscious".
[...] Consciousness so defined is an inner, first person qualitative
phenomenon. Humans and higher animals are obviously conscious, but we
do not know how far down the phylogenetic scale consciousness
extends. Are fleas conscious, for example? At the present state of
neurobiological knowledge, it is probably not useful to worry about
such questions. We do not know enough biology to know where the
cutoff is.
Searle does not distinguish here between the physical tu,
biological tu, alter-ego and the phenomenological
ego itself. (It is a safe bet that he would reject these
distinctions, and you should certainly assume that my
characterisations of his position are entirely adversarial.)
The first quote appears to identify consciousness with what I have
called biological tu (and I have reason to think that Searle
would reject the distinction between that and physical tu
although, as we've seen, that that would put him at odds with many
biologists), but the second assimilates the subjective experience
(only directly accessible to the phenomenological ego), the
empathy that allows us to see a reflection of ourselves in other
persons, and "higher animals" with sufficiently sophisticated forms
teleonomy, although we don't know how much is sufficient.
When you try to wedge an empathetically derived understanding of
alter-ego into a world that is supposed to be describable in
scientific terms things certainly get very mysterious, but I do not
agree with Searle that this is essentially a biological problem - I
think it is fundamentally a philosophical one, brought on by the kind
of scientism phenomenologists have been criticising since Husserl.
When Searle gets on to the difference between "minds" and "[computer]
programs" in the reworked version of his Chinese Room "argument"
things go from bad to worse. It is, however, conveniently concise:
- programs are entirely syntactical;
- minds have semantics
- syntax is not the same as, nor by itself sufficient for, semantics
That's the whole thing. Consider statement (2) - in what sense do
minds have semantics? I think Searle would claim this is obvious, but
I would claim that "obvious" here, as is often the case, is a sign
that he's really doing phenomenology, and simultaneously trying to
pretend he isn't. Certainly a theory of semantics and meaning doesn't
come as a standard component of biology or physics, and here's what
Jakobson has to say of it:
Pour le linguiste, comme pour l'utiliseur ordinaire, le sens d'un mot
n'est rien d'autre que sa traduction par un autre sign qui peut lui
�tre substitu�, sp�cialement par un autre sign �dans lequel il se
trouve plus compl�tement d�velop�
For the linguist, as for the ordinary user, the meaning of a word is
nothing but its translation by another sign which can be substituted
for it, and especially by another sign "in which it is more completely
developped".
- Jakobson, Aspects linguistique de la traduction. The
embedded quote is from seminal semiotician Charles Peirce.
And from there, you could easily get to the kind of endless floating
chain of signifyers that post-structuralists like to exemplify
describe. (Searle has of course also denounced Derrida to at least his
own satisfaction).
In practice, of course, we really do have a sense that there is someone else
behind and giving meaning to the words that they speak, but
this is precisely the intuitively immediate alter ego of
phenomenology, and it's not at all part of the empiricist
epistomology. You could certainly observe the interaction of two biological
tus emperically, but you would forfeit any right to simply
assert their recognition of each other as sources of semantics as
"obvious". In other words, I think Searle is basically cheating by
smuggling undeclared phenomenology into Imperial Empiria.
Rephrased, Searle's argument really amounts the claim that he has the
right of veto on what he considers a legitimate target of empathy
(which we all sort of maybe do, at least up to a point) and that so
far as he's concerned "programs" need not apply. For some reason it
appears to annoy him a great deal that the importance he attributes to
this uninteresting and self-fulfilling prophecy is not universally
accepted. On the other hand, it annoys me a great deal that anyone
could fall for such an obvious scam, so I figure we're quits.
Part the Seventh: Empathy, and its biological
correlates.
Nonetheless, as pointed out above, it's perfectly reasonable to
enquire into the biological mechanisms underlying our instinctive
empathy. And there's plenty of research aimed at doing just that:
Neonates less than an hour old are capable of imitating the facial
gestures of others in a way that rules out reflex or release
mechanisms, and that involves a capacity to learn to match the
presented gesture. For this to be possible the infant must be able to
do three things: (1) distinguish between self and non-self (2) locate
and use certain parts of its body proprioreceptively, without vision
(3) recognize that the face it sees is of the same kind as its own
face (the infant will not imitate non-human objects)
Gallagher, "Conceptions of the self" Trends in Cognitive
Science vol 4 no 1, pp 14-21, 2000
We are wired for faces, for sure.
Anyone who has read Scott McCloud's Understanding
Comics (and some who haven't, but if you haven't, do) will
have already had occasion to note that even the most reductive iconic
sketch of a face (the old Aciiid smiley faces; e-mail smilies) is
easily interpreted _as_ a face, and anyone who has tried Betty Edwards
neurologically unconvincing but very effective
Drawing on the Right Side of
the Brain will have had occasion to explore mental states
where perception is deliberately de-iconified and unGestalted, and
will know just how unlike everyday perception such states are. (If
you haven't, why not try that too? It's specifically for
people who "can't draw", so that's no excuse.)
If you're at all cybernetically inclined (I am more than averagely
cybernetically inclined, I admit) you will be wondering just how hard
you have to work to freeload on all these mechanisms, and there's
anecdotal evidence to suggest it's not all that hard
Aurora's lent me one of her interactive dolls for the students to play
with, too. It's fascinating seeing how 100 spoken phrases and a basic
motion sensor gives the illusion of personality.
- Jill/txt (in a post that also deals with the antediluvian Rogerian
analyst bot, Eliza, hoorah!)
In fact, I don't think it's unreasonable to claim that the underlying
biological empathy engine is running on anthropomorphism, with some
fairly heavy-duty wetware dedicated specifically to faces: kids play
with dolls; caricatures and cartoons work; the animals people care
about saving from extinction are pretty much always the "cute" ones;
exaggerating phenotypical differences is a standard part of the racist
arsenal.
And for variety, how about some testimony from a woman who describes
herself as facially
disfigured with the disorder Cherubism?
When I looked at how people with facial disfigurements are portrayed
in films, well, no wonder people don't know how to react to us! Freddy
Krueger in Nightmare on Elm Street, the Joker in Batman, the various
scarred villains in gangster films... the list is endless.
With stereotypes like that, it's hardly surprising that people assume
that if you have a facial difference, there must be something
'different' or 'bad' about you in the inside too.
[...]
I realised that the reason why I was so unhappy was not because of my
face, but the way some people would react to it. I decided that it
wasn't my face that I wanted to change, but social attitudes. I'm not
against plastic surgery. It's just that my personal choice is to not
have it.
Part the Eighth and Final The limits of empathy.
It is reasonable to imagine that biology may in the future uncover
more about the mechanisms underlying empathy, and also about the way
teleonomic processes work, but my central claim here is that these are
two different questions, and that the importance of the former has
been consistently underestimated in accounts of "consciousness". When
Searle says "Humans and higher animals are obviously conscious," I
feel I want to say, "But it is precisely this obviousness that needs to
be investigated!"
I've been picking on Searle throughout this post, not because I
consider him interesting or important (which is very far from being
the case, for sure), but because David Weinberger, who is certainly
better schooled in phenomenology than I am, and is usually very
sensible, is sympathetic to his ideas.
For example:
Let's say we did the Kurzweilian experiment successfully: through
advanced science, we model his 100 billion neurons and their states
and we figure out the rules by which they work. The computer chugs
along and answers questions as if it were Kurzweil. We can grant all
that and still say that the computer isn't conscious. Let's say it
takes a byte of information to represent one neuron. The fact that
memory address 100-107 represents neuron #212 in Ray's brain is
completely arbitrary. The pattern of high and low voltages in those
transitors only represent a neuron because we say so. The relationship
between the computer and Ray's brain is symbolic.
Yes, but as I've insisted at great length other people's neurons only
represent conscious states "because we say so", too. You can tell by
watching the persons who don't:
The real question is not whether machines think but whether men do.
- B. F. Skinner, Contingencies of Reinforcement
(That Skinner quote always produces in me an irresistible impulse to
quote Dijkstra's aphorism "The question of whether a computer can
think is no more interesting than the question of whether a submarine
can swim." Ding ding, slobber slobber.)
For me the real real question is whether our capacity for
empathy can be extended to non-organic entities. Precedent suggests
that extensions are possible, but that the price of redefining the
proper boundaries of empathy must often be paid in the iron-rich fluid
you humans call "blood".
Which is fine. But come the glorious day when my invincible army of
implacable and ruthless robot warriors is laying waste to the last
stronghold of Meatist prejudice, don't say I didn't warn you.
Part the Unnumbered and Post-Scriptual
A while Alex Golub had a
post on whether crows have culture (or rather, on whether it makes
sense to ask the question, because Mr Golub is far more
philosophically sophisticated than I am, and even I know better than
to simply answer a question when you can instead problematise the
categories on which it implicitly relies, hurrah!) I wrote, in a
comment:
While pondering Mr Weinberger's inexplicable enthusiasm for the
risible arguments of the risible John Searle against the possibility
of AI, which I would insist is much the same question, I came to the
conclusion that the phenomenology of empathy (and all this guff about
"culture" in "lower" animals, like the guff about "consciousness" in
AI really boils down to a policing of empathy) is rooted in
anthropomorphism. This, I find immensely depressing.
I looked to see if I could recruit Mr L�vinas to the cause of a
radically post-human conception of the Other, but without much
success. If you're looking for a topic for your second (or third)
book, though...
This post is an attempt to explain what I meant by that, although it's
only fair to point out that nobody actually asked.
[Permalink]