phaedrus@eneevax.UUCP (04/26/84)
<Why do you eat this line if you're so artificially intelli-
gent?>
Hi there, I would like to start a new topic for discussion.
Here, at the Univ. of Maryland, we have a course on the book
by Douglas R. Hofstadter, "Godel, Escher, Bach: An Eternal
Golden Braid." In this book there are some really fascinat-
ing topics on which we could get into some really good fra-
cases (sp?). I want you experienced AI'ers, philosophers,
anybody that is interested in AI out there in netland to let
your hair down and start voicing your opinions on the fol-
lowing excerpt from the book which in turn is an excerpt
from the article by J.R. Lucas entitled, "Minds, Machines,
and Godel".
I have no idea what I need to do to follow copyright laws,
so my apologies to the afore-mentioned authors in advance,
if I do something to offend them.
Here goes....
At one's first and simplest attempts to philosophize, one
becomes entangled in questions of whether when one knows
something one that one knows it, and what, when is thinking
of oneself, is being thought about, and what is doing the
thinking. After one has been puzzled and bruised by this
problem for a long time, one learns not to press these ques-
tions: the concept of a conscious being is, implicitly,
realized to be different from that of an unconscious object.
In saying that a conscious being knows something we are say-
ing not only does he knows it, but he knows that he knows
it, and that he knows that he knows that he knows it, and so
on, as long as we care to pose the question: there is, we
recognize, an infinity here, but it is not an infinite
regress in the bad sense, for it is the questions that peter
out, as being pointless, rather than the answers. The ques-
tions are felt to be pointless because the concept contains
within itself the idea of being able to go on answering such
questions indefinitely. Although conscious beings have the
power of going on, we do not wish to exhibit this simply as
a succession of tasks they are able to perform, nor do we
see the mind as an infinite sequence of selves and super-
selves and super-super-selves. Rather, we insist that a
conscious being is a unity, and though we talk about parts
of our mind, we do so only as a metaphor, and will not allow
it to be taken literally.
The paradoxes of consciousness arise because a cons-
cious being can be aware of itself, as well as of other
things, and yet cannot really be construed as being divisi-
ble into parts. It means that a conscious being can deal
with Godelian questions in a way in which a machine cannot,
because a conscious being can consider itself and its per-
formance and yet not be other than that which did the per-
formance. A machine can be made in a manner of speaking to
"consider" its performance, but it cannot take this "into
account" without thereby becoming a different machine,
namely the old machine with a new part added. But it is
inherent in our idea of a conscious mind that it can reflect
upon itself and criticize its own performances and, no extra
part is required to do this: it is already complete, and
has no Achilles' heel.
The thesis thus begins to become more of a matter of
conceptual analysis than mathematical discovery. This is
borne out by considering another argument put forward by
Turing. So far, we have constructed only fairly simple and
predictable artifacts. When we increase the complexity of
our machines, there may, perhaps, be surprises in store for
us. He draws a parallel with a fission pile. Below a cer-
tain "critical" size, nothing much happens: but above the
critical size, the sparks begin to fly. So too, perhaps
with brains and machines. Most brains and all machines are,
at present, "sub-critical"-they react to incoming stimuli in
a stodgy and uninteresting way, they have no ideas of their
own and can produce only stock responses-but a few brains at
present, and possibly some machines in the future, are
super-critical, and scintillate on their own account. Tur-
ing is suggesting that it is only a matter of complexity a
qualitative difference appears. So that super critical
machines will be quite unlike the simple ones hitherto
envisaged.
This maybe so. Complexity often does introduce quali-
tative differences. Although it sounds implausible, it might
turn out that above a certain level of complexity, a machine
ceased to be predictable, even in principle, and started
doing things on its own account, of, to use a very revealing
phrase, it might begin to have a mind of its own. It would
begin to have a mind of its own when it was no longer
entirely predictable and entirely docile, but was capable of
doing things which we recognized as intelligent, and not
just mistakes or random shots, but which we had not pro-
grammed into it. But then it would cease to be a machine,
within the meaning of the act. What is at stake in the
mechanist debate is not how minds are, or might be, brought
into being, but how they operate. It is essential for the
mechanist thesis that the mechanical model of the mind shall
operate according to "mechanical principles," that is, we
can understand the operation of the whole in terms of the
operation of its parts, and the operation of each part shall
be either determined by its initial state and the
construction of the machine or shall be a random choice
between a determinate number of determinate operations. if
the mechanist produces a machine which is so complicated
that this ceases to hold good of it, then it is no longer a
machine for the purpose of our discussion, no matter how it
was constructed. We should say, rather, that he had created
a mind, in the same sort of sense as we procreate people at
present. There then be two ways of bringing new minds into
the world, the traditional way, by begetting children born
of women, and a new way by constructing very, very complex
systems of, say, valves and relays. When talking of the
second way we should take care to stress that although what
was created looked like a machine, it was not one really,
because it was not just the total of its parts: one could
not even tell the limits of what it could do, for even when
presented with the Godel type question, it got the answer
right. In fact we should say briefly that any system which
was not floored by the Godel question was eo ipso not a Tur-
ing machine, i.e. not a machine within the meaning of the
act.
Ibbidy-Ibbidy-Ibbidy that's all folks!!!
--
Without hallucinogens, life itself would be impossible.
ARPA: phaedrus%eneevax%umcp-cs@CSNet-Relay
BITNET: phaedrus@UMDC
UUCP: {seismo,allegra,brl-bmd}!umcp-cs!eneevax!phaedrusdavidson@sdcsvax.UUCP (Greg Davidson) (04/28/84)
This is a response to the submission by Phaedrus at the University of Maryland concerning speculations about the nature of conscious beings. I would like to take some of the points in his/her submission and treat them very skeptically. My personal bias is that the nature of conscious experience is still obscure, and that current theoretical attempts to deal with the issue are far off the mark. I recommend reading the book ``The Mind's Eye'' (Hofstadter & Dennett, eds.) for some marvelous thought experiments which (for me) debunk most current theories, including the one referred to by Phaedrus. The quoted passages which I am criticizing are excerpted from an article by J. R. Lucas entitled ``Minds, Machines, and Goedel'' which was excerpted in Hofstadter's Goedel, Escher, Bach and found there by Phaedrus. the concept of a conscious being is, implicitly, realized to be different from that of an unconscious object This statement begs the question. No rule is given to distinguish conscious and unconscious objects, nothing is said about the nature of either, and nothing indicates that consciousness is or is not a property of all or no objects. In saying that a conscious being knows something we are saying not only does he know it, but he knows that he knows it, and that he knows that he knows that he knows it, and so on .... First, I don't accept the claim that people possess this meta-knowlege more than a (small) finite number of levels deep at any time, nor do I accept that human beings frequently engage in such meta-awareness; just because human beings can pursue this abstraction process arbitrarily deeply (but they get lost fairly quickly, in practice), does not mean that there is any process or structure of infinite extent present. Second, such a recursive process is straightforward to simulate on a computer, or imbue an AI system with. I don't see any reason to regard such systems as being conscious, even though they do it better than we do (they don't have our short term memory limitations). we insist that a conscious being is a unity, and though we talk about parts of our mind, we do so only as a metaphor, and will not allow it to be taken literally. Well, this is hardly in accord with my experience. I often become aware of having been persuing parallel thought trains, but until they merge back together again, neither was particularly aware of the other. Marvin Minsky once said the same thing after a talk claiming that the conscious mind is inherently serial. Superficially, introspection may seem to show a unitary process, but more careful introspection dissolves this notion. The paradoxes of consciousness arise because a conscious being can be aware of itself, as well as of other things, and yet cannot really be construed as being divisible into parts. The word ``aware'' is an implicit reference to the unknown mechanism of consciousness. This is part of the apparent paradox. Again, there's nothing mysterious about a system having a model of itself and being able to do reasoning on that model the same way it does reasoning on other models. Also again, nothing here supports the claim that the conscious mind is not divisible. It means that a conscious being can deal with Godelian questions in a way in which a machine cannot, because a conscious being can consider itself and its performance and yet not be other than that which did the performance. Whatever the conscious mind is, it appears to be housed in a physical information processing system, to wit, the human brain. If our current understanding about the kind of information processing brains are capable of is correct, brains fall into the class of automata and cannot ultimately do any processing task that cannot be done with a computer. The conscious mind can scrutinize its internal workings to an extent, but so can computer programs. Presumably the Goedelian & (more to the point) Turing limitations apply in principle to both. no extra part is required to do this: it is already complete, and has no Achilles' heel. This is an unsupported statement. The whole line of reasoning is rather loose; perhaps the author simply finds it psychologically difficult to suppose that he has any fundamental limitations. When we increase the complexity of our machines, there may, perhaps, be surprises in store for us.... Below a certain ``critical'' size, nothing much happens.... Turing is suggesting that it is only a matter of complexity [before?] a qualitative difference appears. Well, its very easy to build machines that are infeasible to predict. Such machines do not even have to be very complex in construction to be highly complex in behavior. Las Vegas is full of many examples of such machines. The idea that complexity in itself can result in a system able to escape Goedelian and Turing limitations is directly contradicted by the mathematical induction used in their proofs: The limitations apply to <<arbitrary>> automata, not just to automata simple enough for us to inspect. Charlatans can claim any properties they want for mechanisms too complex for direct disproofs, but one need not work hard before dismissing them with indirect disproofs. This is why the patent office rejects claimed perpetual motion machines which supposedly operate merely by the complexities of their mechanical or electromagnetic design. It is also why journals of mathematics reject rediculously long proofs which claim to supply methods of squaring the circle, etc. No one examines such proofs to find the flaw, it would be a thankless task, and is not necessary. It is essential for the mechanist thesis that the mechanical model of the mind shall operate according to ``mechanical principles,'' that is, we can understand the operation of the whole in terms of the operation of its parts.... Certainly one expects that the behavior of physical objects can be explained at any level of reduction. However, consciousness is not necessarily a behavior, it is an ``experience'', whatever that is. Claims of consciousness, as in ``I assert that I am conscious'' are behavior, and can reasonably be subjected to a reductionist analysis. But whether this will shed any light on the nature of consciousness is unclear. A useful analogy is whether attacking a computer with a voltmeter will teach you anything about the abstractions ``program'', ``data structure'', ``operating system'', etc., which we use to describe the nature of what is going on there. These abstractions, which we claim are part of the nature of the machine at the level we usually address it, are not useful when examining the machine below a certain level of reduction. But that is no paradox, because these abstractions are not physical structure or behavior, they are our conceptualizations of its structure and behavior. This is as mystical as I'm willing to get in my analysis, but look at what Lucas does with it: if the mechanist produces a machine which is so complicated that this [process of reductionist analysis] ceases to hold good of it, then it is no longer a machine for the purpose of our discussoin, no matter how it was constructed. We should say, rather, that he had created a mind, in the same sort of sense as we procreate people at prsent. If someone produces a machine which which exhibits behavior that is infeasible to predict through reductionist methods, there is nothing fundamentally different about it. It is still obeying the laws of physics at all levels of its structure, and we can still in principle apply to it any desired reductionist analysis. We should certainly not claim to have produced anything special (such as a mind) just because we can't easily disprove the notion. When talking of [human beings and these specially complex machines] we should take care to stress that although what was created looked like a machine, it was not one really, because it was not just the total of its parts: one could not even tell the limits of what it could do, for even when presented with the Goedel type question, it got the answer right. There is simply no reason to believe that people can answer Goedelian questions any better than machines can. This bizarre notion that conscious objects can do such things is unproven and dubious. I assert that people cannot do these things, and neither can machines, and that the ability to escape from Goedel or Turing restrictions is irrelevant to questions of consciousness, since we are (experientially) conscious but cannot do such things. I find that most current analyses of consciousness are either mystical like the one I've addressed here, or simply miss the phenonmenon by attacking the system at a level of reduction beneath the level where the concept seems to apply. It is tempting to thing we can make scientific statements about consciousness just because we can experience consciousness ourselves. This idea runs aground when we find that this notion is dependent on capturing scientifically the phenomena of ``experience'', ``consciousness'' or ``self'', which I have not yet seen adequately done. Whether consciousness is a phenomenon with scientific existence, or whether it is an abstract creation of our conceptualizations with no external or reductionist existence is still undetermined. -Greg Davidson (davidson@sdcsvax.UUCP or davidson@nosc.ARPA)
pem1a@ihuxr.UUCP (Tom Portegys) (05/04/84)
Phaedrus' article made me think of a story in the book "The
Mind's Eye", by Hofstadter and Dennett, in which the relationship
between subjective experience and physical substance is explored.
Can't remember the story's name but good reading. Some other
thoughts:
One aspect of experience and substance is how to determine when
a piece of substance is experiencing something. This is good to
know because then you can fiddle with the substance until it stops
experiencing and thereby get an idea of what it was about the
substance which allowed it to experience.
The first reasonable choice for the piece of substance might be
yourself, since most people presume that they can tell when they
are having a conscious experience. Unfortunately, being both the
measuree and measurer could have its drawbacks, since some experiments
could simulaneously zap both experience and the ability to know or not
know if an experience exists. All sorts of problems here. Could you
just THINK you were experiencing something, but not really?
What this calls for, it seems to me, is two people. One to measure
and one to experience. Of course this would all be based on the
assumption that it is even possible to measure such an elusive
thing as experience. Some people might even object to the notion
that subjective experiences are possible at all.
The next thing is to choose an experience.
This is tricky. If you chose self-awareness as the experience, then
you would have to decide if being self-aware in one state is the same
as being self-aware in a different state. Can the experience be the
same even if the object of the experience is not?
Then, a measuring criterion would have to be established whereby
someone could measure if an experience was happening or not. This
could range from body and facial expressions to neurological readings.
Another would be a Turing test-like setup: put the subject into a
box with certain I/O channels, and have protocols made up for
measuring things. This would allow you to REALLY get in there and
fiddle with things, like replacing body parts, etc.
These are some of the thoughts that ran through my head after reading
the Phaedrus article. I think I thought them, and if I didn't, how
did this article get here?
Tom Portegys, Bell Labs, ihlpg!portegys
(ihlpg currently does not have netnews, that's why this is coming from
ihuxr).