josh@cs.rutgers.edu (06/11/91)
[There is a small nanotech connection, but if you're not interested
in the AI/philosophy debate you can skip this and not miss anything..
--JoSH]
Roger Penrose's "The Emperor's New Mind" appears to have opened a new
chapter in the AI'ers vs philosophers debate. Unfortunately, this
appears to be a development of form and not of substance. Penrose's
book, if you will remember, was written in something of a departure
from the format of previous philosophical tracts. Most of the book
was devoted to overviews of different scientific and mathematical
fields, and in the closing chapter, Penrose pitched his personal
convictions on the matter of AI. Now Penrose is a world-class expert
in at least some of the subjects he was writing about, and his
opinions at the end were clearly labelled as such. Even though there
are some things he seems to have missed about AI, and his opinions are
wrong, the book as a whole is worth reading and an honest exposition
of his point of view.
Not so "Mind, Machines, and Human Consciousness" by Robert L. Nadeau
(Contemporary Books, 1991). Nadeau also gives us a peripatetic tour
of vaguely related items from Eliza to Langmuir-Blodgett films, but
he is not an expert in any of them (he is a philosopher). Moreover,
when he comes to the stinger on the tail end, rather than label his
opinions as such, he retreats into turgid jargon and claims he is
proving them.
Though numerous mistakes creep into the "grand tour," they are self-
contained as Nadeau appears simply to have written each page or two
as a condensation of some popular exposition or another. The shallowness
of understanding that this approach reveals, however, is an altogether
more serious defect. Penrose's gut feelings on varous subjects, honestly
labelled, are worthy of consideration; Nadeau's "proofs" are not.
A case in point is Eric Drexler and nanotechnology. Although Nadeau
purports to explain Drexler's ideas, from assemblers to AI by brain
simulation, he has not even read "Engines of Creation"; aside from
one National Academy paper, everything Nadeau knows about nanotechnology
comes from reading Fjermedal's "Tomorrow Makers"! (This is itself a
popularization of a bunch of such ideas; entertaining light reading.)
Nadeau also appears to believe that Eric is still at MIT.
If this were all there were to it, MM&HC would simply be shoddy and
shallow. However, it happens that if Nadeau actually had read EoC,
he would have learned about the concept of memes. And it is the
lack of knowledge of memes that makes Nadeau's central argument,
what there is of it, one of the more blatant strawman attacks I've
seen in published philosophy.
The argument follows what seems to be the latest fad in AI-bashing:
"I admit that intelligent machines can be created that act just like
humans; but they won't have true feelings, or consciousness, or some
other undefineable quality that is part of the essence of being human."
... and therefore AI machines can never be more, morally speaking,
than adjuncts to Truly Human people like Us Philosophers. To a
memetic analysis, this meme is quickly seen as consisting almost
entirely of its own hook: a sense of moral superiority is conferred
on the believer. It is the business of philosophers, of course,
to clothe such memes in appropriate mental camoflage: Nadeau's
version of this camoflage is called the Evolution of Consciousness
Hypothesis.
Nadeau's strawman is constructed as follows: The AI community has come
up with this doctrine that he calls the Evolution of Consciousness
Hypothesis. This hypothesis claims that the human mind is simply
one stage in a grand upward sweep towards ever more complex information-
processing systems. If this hypothesis ever comes to be believed,
people will want to make themselves extinct to make way for the
machines. This would be terrible, since intelligent machines would
not really be conscious, nor would they be moral agents. Nadeau goes
so far as to say:
"A world inhabited by our "descendants" in the form of AI systems[,]
and a world reduced to dust and ash by global thermonuclear war
would be quite similar in two important respects. Both would be
barren of conscious beings..."
Luckily, Nadeau has the answer to the AI Armageddon: the Evolution
of Consciousness Hypothesis is false! Since by this point the
argument has gotten a bit emotional, it gets a little harder to
pick out the logical strand, but it seems to go something like
this:
a) Darwin was wrong because he got some of his ideas from Malthus,
who got them from Spencer, who was wrong because Enlightenment
economists were too fond of physics. In contrast, Nadeau can
show, that "we are now aware, as Darwin could not be," that
things evolve by being nice to each other instead of being
mean to each other.
b) Nadeau cites Gould to show that whole organisms, rather than
individual genes, are selected for in evolution. Thus, he
claims, there is no evolutionary pressure towards ever more
complex organisms.
c) Thus, after our brains became complex enough to handle symbolism
and language, we didn't evolve, rather, we "invented" ourselves
consciously.
What Nadeau has actually managed to understand is a triviality: that
biological evolution of the human genome is not responsible for the
production of artificially intelligent computers. Nadeau appears to
believe that the concept of AI systems as our descendants, a la
Moravec in "Mind Children", is the modern-day equivalent of Social
Darwinism, and much of his rhetoric is aimed at the concept in those
terms.
It is almost incredible, given the catalog-like nature of the "AI,
This Is Your Life" section of the book, that Nadeau would never have
heard of memes or genetic algorithms, nor would have seen anything
like Calvin's "Cerebral Symphony". It is as if he had specifically
decided to avoid reading anything about the relationship between
evolutionary theory and AI and consciousness, in writing a book
whose crucial point turned on exactly that relationship!
And yet it seems to be the case. I can find no mention of memes,
or genetic algorithms, or of recent neurophysiological theories
of what Calvin referred to as "Darwin machines". The closest he
came was a passing reference to AM, which he calls "interesting
and innovative", period. No mention of mechanism. Eurisko isn't
mentioned at all.
Given his total innocence of evolutionary theory, it's unsurprising
that Nadeau arrives at his silly conclusion.
a) It doesn't really matter whether evolution occurs by niceness
or meanness (in fact both pressures operate). It does occur.
b) In the memetic ecology, the unit of selection can vary from
the individual meme to the whole meme complex.
c) The concept of evolution by variation and selection is broad
enough to encompass a host of mechanisms for each of those two
components. Just as when we invented airplanes, we "conquered
gravity" in such a way all physical law, including gravity,
continues to hold true everywhere, it seems that we have
"invented ourselves" in a process that is quite validly
described by the Darwinian laws of memetic evolution.
Even taken at face value, the argument is a non sequitur--after all,
if we "invented" what we are today, how then can it be wrong for us
to invent what we'll be tomorrow?
After demolishing the Evolution of Consciousness Theory, Nadeau turns
to prove the ineffable essence of humanistic humanitude by reference
to quantum mechanics, relativity, Godel's theorem, philosopher of science
Errol Harris' conjecture that the universe as a whole is conscious,
and Kierkegaard's Leap of Faith (I kid you not). I somehow missed
understanding how, if the universe as a whole is conscious, and
humanity is conscious by virtue of partaking in this All Permeating
Holistic Consciousness, that computers couldn't also partake in the
A.P.H.C as well.
But then I'm just one of those evil "scientist-engineers" who is
deficient in being without an "active moral and philosophical
imagination." I guess I have to agree to this extent: if the
mind of Robert L. Nadeau were uploaded into a computer that met
*my* standards for a wise, knowledgeable, perceptive, intelligent
system, something essential would, indeed, have been lost.
--JoSH