| Transcript
from July 19, 2002 4-5 PM Eastern
Copyright
© by International Society for Complexity, Information, and Design
July 2002.
ISCID
Moderator
Welcome, everyone. ISCID is pleased to have several of the contributors
to the new book "Are We Spiritual Machines?" as guest speakers
in todays chat. Before we announce the speakers, I would like to give
everyone a heads-up on the protocol for today's chat.
ISCID
Moderator
The public (that is most of you) can type in questions and submit them.
The questions will not automatically be displayed. Rather, they will
be sent to a moderator who will then select questions for everyone to
view. The guest speakers will then have the opportunity to respond to
the questions that have been selected. When the guest speakers have
finished their comments, the moderator will approve another question.
This cycle will continue until 5PM Eastern. When submitting a questions,
please indicate which guest speaker it is addressed to. Please stay
on topic and be as brief and concise as possible.
ISCID
Moderator
Right now it looks like Ray K. and Jay Richards are the only two guest
speakers logged on.
ISCID
Moderator
I will go ahead and introduce them and the others we are expecting.
Bill
Dembski
I'm here as well
ISCID
Moderator
Tom Ray, who authored the chapter "Kurzweil's Turing Fallacy",
is not able to be with us today. He did, however, want to say a few
words. Here they are:
ISCID
Moderator
Sorry that I can't join the chat. I'm sleeping in Kyoto, Japan. But
I have prepared a few comments for those who are interested. You can
read them at: http://www.his.atr.co.jp/~ray/richards.html
ISCID
Moderator
Our first guest is Ray Kurzweil. Ray is an inventor, entrepreneur, and
author. His books include "The Age of Intelligent Machines"
and "The Age of Spiritual Machines, When Computers Exceed Human
Intelligence". The recently published "Are We Spiritual Machines?
Ray Kurzweil vs. the Critics of Strong AI" is a collection of critiques
and responses regarding Ray's vision of Strong Artificial Intelligence.
ISCID
Moderator
Welcome Ray.
Ray
Kurzweil
Glad to "be" here.
ISCID
Moderator
The next guest is Jay Wesley Richards. Jay received his Ph.D. in philosophy
and theology from Princeton Theological Seminary. He is the editor of
the collection "Are We Spiritual Machines? Ray Kurzweil vs. the
Critics of Strong AI."Jay is currently finishing up a book with
astronomer Guillermo Gonzalez entitled The Privileged Planet. He is
also a senior fellow and program director at the Discovery Institute
in Seattle, Washington.
ISCID
Moderator
William A. Dembski is associate research professor in the conceptual
foundations of science at Baylor University. He holds a Ph.D. in both
philosophy (Univ. of Illinois at Chicago) and mathematics (University
of Chicago) as well as an M.Div. from Princeton Theological Seminary.
Bill is the author and editor of several books including "The Design
Inference" and "No Free Lunch" which deal with his work
in information and probability theory. Bill's contribution to "Are
We Spiritual Machines?" is a chapter entitled "Kurzweil's
Impoverished Spirituality."
ISCID
Moderator
Ok. It looks as if, right now, Ray may be the only one who is actually
"here" with us.
ISCID
Moderator
So lets get started.
Danpech
While we can
have the idea that there are things outside ourself that are aware only
because we ourself are aware, it is granted by nearly every one as objective
fact that there are things outside himself that do not possess awareness
(that do not feel either sensorily or emotionally, that do not think,
etc.). In my opinion, central to the problem of multiple realizability
is the question of how, in the first place, we each get the very idea
that there are things that are not aware. How do you, personally, know
that there are things that are not aware?
Ray
Kurzweil
I don’t know that absolutely. Maybe I’m the only person
who’s conscious. That’s consistent with a dream. Even by
common wisdom, there seem to be both people and objects in my dream
that are outside myself, but clearly they were created in myself and
are part of me, they are mental constructs in my own brain. Or maybe
every other object is conscious. I’m not sure what that would
mean as some objects may not have much to be conscious of. It’s
hard to even define what each object or thing is that might be conscious
as there are no clear boundaries. Or maybe there’s more than one
conscious awareness associated with my own brain and body. There are
plenty of hints along these lines with multiple personalities, or people
who appear to do fine with only half a brain (either half will do).
We appear to be programmed with the idea that there are “things”
outside of our self, and some are conscious, and some are not. That’s
how we’re constructed. But it’s not entirely clear that
this intuitive conception
Ray
Kurzweil
Is there a limit on the length of responses, it seems the end of mine
was cut off?
ISCID
Moderator
When you cut and paste, sometimes it gets messed up when there are line
breaks.
Ray
Kurzweil
Here's the last paragraph again: Or maybe there’s more than one
conscious awareness associated with my own brain and body. There are
plenty of hints along these lines with multiple personalities, or people
who appear to do fine with only half a brain (either half will do).
We appear to be programmed with the idea that there are “things”
outside of our self, and some are conscious, and some are not. That’s
how we’re constructed. But it’s not entirely clear that
this intuitive conception matches perfectly to ultimate ontology.
Ray
Kurzweil
(end of response)
Micah
Sparacio
Ray, what
are your thoughts on the Frame Problem? Do you see it as a challenge
that requires a "paradigm shift" in software before human
level intelligence is attainable?
Ray
Kurzweil
The concept of frames as discussed by Minsky and others has a variety
of interpretations, but generally refers to issues of context in a knowledge
base that incorporates such concepts as inheritance of features. And
yes, I do think a paradigm shift is required from this kind of expert
system methodology. We’ll never get to anything with the suppleness
and subtlety of human intelligence with expert system rules. That’s
not how human intelligence works.
Ray
Kurzweil
Although the work on a system such as Cyc is impressive and worthwhile,
the it’s clear that this type of approach will not create human-level
intelligence. Human intelligence works through pattern recognition,
which uses a paradigm combining massive parallelism, chaotic self-organization,
holographic-like methods of information storage.
Ray
Kurzweil
Human thinking is very slow, typical interneuronal connection reset
times are 5 milliseconds, so we don’t have time in real time to
think through complex sequential rules. But we do have on the order
of a 100 trillion connections working more or less simultaneously to
provide powerful and subtle pattern recognition capabilities. Pattern
recognition approaches have softer edges to their expertise. We need
to bring pattern recognition methodologies to domains such as natural
language understanding.
Ray
Kurzweil
(end of response).
Brandon
Watson
Your replies
to your critics often seem to involve what might be called a moderate
skepticism about other consciousnesses. That is, you seem to proceed
in something like this way: There is no objective test for consciousness;
so it cannot be decided with certainty whether any given thing is conscious,
although we do assume it in certain cases based on analogy with ourselves;
so we can concern ourselves entirely with the objective correlates of
consciousness; we can build functionally equivalent versions of these;
so we will reach a point at which our reasons for calling the computer
conscious will be as good as our reasons for calling anything else conscious.
Do you regard your conclusions as dependent on this line of thought?
Do we have any knowledge of consciousness beyond its subjectivity?
Ray
Kurzweil
I do believe that the line of thinking that you aptly summarize is more
than just a philosophical slight of hand. We really do have problems
penetrating the subjective reality of other entities. Our inability
to do that with other humans is behind much human conflict. Humans feel
deeply the suffering of their friends and allies, and easily discount
/ dismiss the comparable experience of their enemies.
Ray
Kurzweil
How about animals? Basic disagreement on whether or not animals are
capable of subjective experience, e.g., suffering, or whether they are
just operating by “instinct,” i.e., like a simple machine,
lies behind our disputes on animal rights.
Ray
Kurzweil
This issue will be even more profound and difficult to resolve with
machines. On the one hand, one might argue that machines are less like
humans than animals because at least animals are biological and have
many similar organs and structures and behaviors. On the other hand,
since at least some machines will be based on reverse engineering of
humans, these machines may be more similar to humans and human behavior
than animals.
Ray
Kurzweil
Also relevant here is a rather important slippery slope. We’re
already replacing portions of our biological neurons with machine equivalents.
It’s early yet in this process, but this will ultimately pick
up speed. No one says that the current Cyborgs among us (e.g., people
with cochlear implants, or the recently FDA approved neural implant
for Parkinson’s Disease) are not fully human. There’s no
clear line between these beginning steps in replacing biological circuits
with nonbiological ones to a fully nonbiological equivalent to human
brains.
Ray
Kurzweil
(end of response).
Iain
Strachan
How do computers get to develop their own smart software, when it's
likely that to do such things involve complex decisions that, for a
formal mathematical system such as a computer, involves NP hard problems?
Ray
Kurzweil
We use Bayesian nets in one of our projects, and at the risk of oversimplification,
we can think of it as a clever way to get probabilities into a network
that provides more flexibility than rule-based knowledge bases. I don’t
believe that humans solve problems in real time involving multiple variables
using anything like a mathematically appropriate solution.
Ray
Kurzweil
We apply a pattern recognition type of methodology to our real-time
estimates. Think of a child catching a fly ball. At some point, she
sense to take a few steps back, and then raises her hand in about the
right place (maybe). But she is not computing all the differential equations
to do this even in an unconscious way. We appear to have an ability
to develop models of how certain curves will evolve which utilize our
self-organizing methods, so it requires training and experience. That’s
why she won’t do a very good job of catching the ball until she
becomes experienced at doing it. But it’s not a direct computational
process. It’s an ability to apply certain curve fitting abilities
that approximate expected behavior.
Ray
Kurzweil
And humans are not particularly optimal at doing this. The ability of
a well designed “quant” investment system to anticipate
short term market trends better than human analysts (something I’ve
been working on) is a good example of the limitations of human trend
or expectation analysis abilities. Most human decision making is based
on very flawed and incomplete models. It’s machines that can combine
this type of anticipatory modeling with more powerful mathematical techniques
when appropriate.
Ray
Kurzweil
Keep in mind that our 10^11 neurons and 10^14 synapses are characterized
by a genome with only 6 billion bits, which is 800 million bytes, which
after routine compression is only about 24 million bytes, half of which
characterizes the brain. We get from this very small genomic information
(which specifies the brain’s initial conditions) to a brain that
contains much more information because the genome specifies stochastic
processes (e.g., semi random wiring in certain regions according to
certain constraints) followed by a self-organizing process using essentially
an evolutionary process (i.e., learning).
Ray
Kurzweil
(end of response).
Chris
Science does very well as long as observers and observed phenomena are
assumed to be spatiotemporally separate, and as long as physical observables
are taken as “given”, but can be seen to fall flat when
it comes to putting the relationship between subjects and objects on
a logical basis. In fact, in its adherence to principles of scientific
objectivity, science excludes subjectivity from the theorization process
as a form of “contamination”. Might it therefore be possible
that the ontological medium of reality, whatever that may be, imposes
“hidden constraints” on emergence, and that these constraints
differentiate between natural and mechanical processes with respect
to the consciousness attribute? What basis is there for assuming otherwise?
Ray
Kurzweil
It is a good insight that there is a separation between the objective
reality of science and the subjective reality that is consciousness.
That is why there is room for philosophy and religion (as a form of
philosophy) to address questions that fall outside the domain of objective
observation and analysis that is the province of science. However,
ISCID
Moderator
Any questions for our other guests (Jay Richards, William Dembski)?
Ray
Kurzweil
the "natural" world is observable by objective methods, and
can be understood and modeled through objective methods, and most importantly,
its methods can be recreated using objective methods, i.e., technology.
In my own field of "pattern recognition," we use biologically
inspired paradigms, such as evolutionary algorithms, neural nets, etc.
These are far from perfect models, but as the pace of reverse engineering
(i.e., understanding the principles of operation ) of the human brain
picks up speed (as it is), our biologically inspired methods will become
closer analogues to the methods we find in nature.
Ray
Kurzweil
(end of response).
Nic
Machines cannot lie. A person can. Do you sappose a test for con. is
the capacity to lie? In other words, Machines are analytic domains while
the Con. mind is not an analytic domain.
Ray
Kurzweil
The Turing test is very specifically a test of lying. The machine has
to lie that it is a human. There is already a Turing test competition,
which my organization is entering (for the designation of "most
human" machine, as we are not yet close to "passing"
a Turing test). But there's no reason why a machine cannot lie. Of course,
successfully lying takes a lot of intelligence, actually more intelligence
than most humans have ("oh what a tangled web we weave....)
Ray
Kurzweil
(end of response)
Iain
I'd like to come back on one of Ray's comments about pattern recognition.
Working with neural nets, it's an interest of mine. I have for a long
thought that pattern recognition is a key component of intelligence.
But surely it is only one small part. One might get a creative idea
by noting the similarity to another known idea (e.g. in "brainstorming").
But the pattern recognition must also be supplemented by an intelligent
putting together of patterns, rather than just recognising similarity.
What kind of techniques do you see being developed for that?
Jay
Richards
Perhaps instead of focusing on consciousness, we could consider *freedom.*
One of the most readily apparent realities about ourselves as persons
is our capacity for free choice. We experience that capacity directly,
at least much more directly than we experience, say, the external world.
We choose between alternatives, and although we are shaped by our context,
biology, etc., we have the sense that these don't always determine our
choices. This capacity, and not simply "consciousness," seems
to me to be an essential property of intelligence. But it is very difficult
to see how a computer, no matter how advanced, which is governed by
algorithms (and perhaps randomizing functions), could ever enjoy such
freedom. I'm not sure what Ray's view is on this, but I know some advocates
of Strong AI just deny the existence of libertarian freedom. But my
direct experience of possessing such freedom seems much more certain
than any theory that might deny its existence.
Ray
Kurzweil
We do have an ability for sequential logical ("rational")
thinking, that apparently is a relatively recent evolutionary development
(in biological brain development).
Ray
Kurzweil
But most of our thinking is pattern recognition based. That's how a
human plays chess. We don't have much time to do real-time sequential
analysis the way a machine can.
Ray
Kurzweil
(end of response)
Chris
Let’s talk about the medium of emergence of consciousness. Phenomena
must be appropriate to the media in which they occur. For example, sound
can occur only in a medium capable of supporting transverse (compression-rarefaction)
waves, and winged flight can occur only in a gaseous atmosphere capable
of providing the necessary lift. It follows that there must be a relationship
between the emergence of consciousness and the medium in which it occurs.
Specifically, this emergence must involve a process parameterized by
attributes of the medium. Unfortunately, science is unable to characterize
the overall medium of reality with respect to the subjectivity attribute;
in fact, it is unable to characterize the process of attribution itself
in any but the most superficial of ways. What is it that actually relates
objects and their attributes…some kind of glue? Don’t we
need a complete account of this relationship in order to relate the
attribute of consciousness to various systems and devices?
Ray
Kurzweil
Jay has raised an important issue - free will, which is closely related
to consciousness, the apparent sense of making free choices. It's as
hard to pin down objectively as, well, subjectivity. My point is that
we will encounter machines whose complexity and depth of processing
is indistinguishable from that of humans, with all their anguished decision
making. Are they actually conscious? Are they actually deploying free
will? Or just appearing to? Is there a difference between such appearance
and reality? Because of the slippery slope argument I alluded to above,
and for many other reasons, I believe we will accept nonbiological intelligence
as human, i.e., conscious, i.e., responsible for its own free will decisions.
But that's a political and psychological prediction, not necessarily
an ontological one.
Ray
Kurzweil
(end of resposne).
tim
As a follow-up to Jay Richards last comment on freedom. Where does "motivation"
factor in?
Sulu
Jay, I do not believe even humans have free will the way you describe.
If the universe is deterministic (that is, if the state of the universe
is known at one time, physics will allow us to predict the state at
any other time), then so are humans. The only distinction we currently
make between humans and deterministic machines is that the complexity
of humans prevents any other human from having perfect knowledge to
predict future behaviour. It is possible that in the future, machines
will be complex enough to be able to view humans are predictable and
deterministic.
Ray
Kurzweil
To respond first to Chris, I think you've articulated another way of
saying that there is a barrier between the objective world of science
and the subjective issue of consciousness. Some people go on to say
that because the issue of consciousness is not scientific, it is, therefore,
not real, or an "illusion." That's not my view. One can say
that it is the most important question. It underlies, for example, morality,
and to the extent that our legal system is based on morality, then our
legal system. We treat crimes that cause suffering of a conscious entity
differently than damaging "propertly". In fact, you can damag
property if you own it. We only punish damaging property because some
other conscious person cares about it.
Ray
Kurzweil
With regard to "motivation," this is another high level attribute
like intuition, creativity, that are inherent charcteristics of entities
with the complexity and depth of organization of humans. We don't yet
have machines of that complexity, but we will.
Ray
Kurzweil
Often, an human can successfully predict the "free will" decisions
of others if that first human has a complete enough understanding of
the cultural model of the other.
Micah
Sparacio
Could you discuss how stasis plays a role in your Law of Accelerating
Returns? How does one get beyond stasis to a true "paradigm shift"?
Ray
Kurzweil
By statis, do you mean the approachong of an asymptote? If so, that
invariably happens with any paradigm. "Moore's Law" (the shrinking
of transistors on an integrated circuit) will approach such an asymptote
when we run out of room on two dimensional circuits (we'll then go to
the third dimension). Invariably as we approach the limits of a
Ray
Kurzweil
particular paradigm, pressure builds up to create the next. We see that
now with the increased intensity of work in three dimensional molecular
computing. Even Moore's law was not the first but the fifth paradigm
to provide exponential growth to computing.
Ray
Kurzweil
Generally speaking, the new paradigm already exists before the old one
dies. Transistors had a niche market before tubes reached their limit.
BUt once they could no longer shrink vacuum tubes any further and maintain
the vacuum, then the superior ability of transistors to maintain ongoing
price performance growth took over (end of response).
Chris
The spacetime intervals between machine components stand for causal
independence (within spatial cross-sections). This highlights the basic
distinction between the causal independence of mechanical components
and the coherence of consciousness, i.e. the parallel or “simultaneous”
mutual causal connectedness of spatially separate components. Reality
is constructed in such a way that spatial separations occupy a lower
rung of its ontological lattice; to build a conscious machine, one would
have to find a means of overcoming the componential independence of
classical reality. Far more likely is that the unity attribute is inherited
from a point up the lattice, and ultimately, from the global identity
of the lattice, which is distributed over spacetime and everywhere implicit.
But this means that unitary consciousness is, in a sense, derived from
the “unitary consciousness” of the universe. So the question
is, how do you propose to create that unity from an assemblage of parts,
and an associated set of laws
Jay
Richards
In response to Sulu. I would assume that if one assumes that the universe
is deterministic (though I don't think predictability is the same as
determinism), then one will deny the existence of free will. But such
determinism is hardly self-evident, even if lots of folks that it as
such. My point is that we experience our capacity for freedom directly,
and as a result should be much more certain of its reality than any
theory that entails determinism. Moroever, even determinists, when they're
not in a a philosophical mode, presuppose the existence of such freedom
when they make moral judgments about the actions of others. If some
theory, from, say, psychology or biology or physics makes such freedom
impossible, then so much the worse for that theory. Our experience and
knowledge of our own freedom will always be more certain than our certainty
of any such theory. This is relevant to to strong AI, because it seems
to me that if we had good reason to think that an AI, in say, 2059,
is exercising freedom (how we
Jay
Richards
(continued) (how we recognize freedom is a matter of debate), then we
would have good reason for inferring that the AI is exercising real
intelligence.
kuebler
It doesn't seem that being able to predict someone else's behavior denies
they have "free will". Knowing someone's preferences does
not give you any more insight into whether they are acting freely or
are "determined". Is there anyway in which a notion of free
will can be incorporated into a deterministic algorithm?
Ray
Kurzweil
Chris, you have a particular ontological model in mind. But making your
assumptions, there's nothing inherent in biological systems that cannot
be emulated with nonbiological systems, and no clear boundary between
biology and nonbiology. As our nonbiological systems become more biological
inspired, and as we merge more and more with our increasing biological-like
technology, it would have the same characteristics as a biological human.
It would share, then, in this unity of consciousness.
Ray
Kurzweil
Fredkin and
Wolfram's conceptions of cellular mechanics
Ray
Kurzweil
show that even simple processes may be deterministic but are unpredictable
by any process not complete equivalent in complexity and time as the
actual process. So deterministic does not necessarily imply predictable.
Ray
Kurzweil
(end of response).
Leonid_Andreev
How would you comment an idea that starting with a certain level of
information processing information field itself turns into information
processor(as was earlier discussed on Brainstorms: http://www.iscid.org/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=6;t=000093)
In this event, the term "spiritual machine" may literally
be understood as a result of biological evolution. In other words, spirit
processes spirit. If this is true, then reverse engineering of consciousness
should be absolutely impossible.
Ray
Kurzweil
In response to Jay
Ray
Kurzweil
I agree that free will is apparent to me, just as my consciousness is
apparent to me.
Ray
Kurzweil
But it may not be apparent to others. I assume that others are conscious
(or have free will), and others may assume it of me, and this may seem
obvious,
Ray
Kurzweil
but this obviousness (and this shared assumption) breaks down as we
leave
Jay
Richards
There seem to be three different categories: freedom (or agency), determinism,
and randomness. It's easy to imagine a computer "acting" according
to the latter two. It's very difficult to see how a computer could actually
exhibit freedom. In fact, it's difficult to see how *we* exhibit freedom.
We, however, experience the capacity for freedom directly, so any adequate
theory should accommodate its existence. Thus, if we had really good
reasons for thinking that an AI exhibited freedom, I would conclude
that we had reason to think there was something more going on than mere
determinism and randomness.
Ray
Kurzweil
shared human experience (e.g., animals, and now the emerging debate
on machines).
Ray
Kurzweil
(end of response).
Ray
Kurzweil
Further response to Jay:
Bill
Dembski
Ray, I'd like you to speak to the role of the first person within your
view of computation. In your review of Wolfram's book, you approvingly
cite Marvin Minsky's "society of mind," in which an intelligence
may result from a hierarchy of simpler intelligences. But that sounds
like the first person dissolves in a sea of modularity, and modularity
is exactly what you seem to want to avoid. How is it possible within
your view of computation to avoid thinking of the first person as anything
but an illusion (Hume's bundle of associations)? And if it is possible,
what computationally grounds the first person perspective?
Ray
Kurzweil
This freedom is only apparent to myself. There's no objective way of
demonstrating it, so there won't be a clear way to distinguish, say,
Ray Kurzweil from a very accurate simulation.
Ray
Kurzweil
Response to Bill:
Ray
Kurzweil
A human intelligence is obviously more than a loose association of processes,
Sulu
First year philosophy students are taught that liberty of desire (getting
what we want) is compatible with determinisim, but not liberty of spontaneity
(the possibility of things being otherwise than they could have been).
However, it is stronger to say that our morals and legal code not only
depend on consciousness but free will of the second kind (liberty of
spontaneity), and even if we can't reconcile it with determinism, it
is humanly infeasible to live without either determinism (the basis
of our science) or free will of the second kind (that we have a "soul"
that is independent of physical laws). This is similar to many other
problems, like justifying induction; we can't guarantee that reality
will stay the same, but it is humanly impossible to live without this
assumption. We attribute to luck or chance phenomena about which we
do not have complete information. It is possible that a machine will
be able to live without induction and be able to eliminate the idea
of "chance". It would be able to behave
Ray
Kurzweil
we do have a very coherently organized identity. But it's far from perfectly
coherent. It can be pretty incoherent at times, and we have all kinds
of different forces and feelings that we feel inside us in some chaotic
pattern. But there is a coherent organization to a human, it's not a
loose association.
Ray
Kurzweil
(end of response to Bill).
Chris
I understand, but rather than making assumptions, I’m trying to
get around certain assumptions already in place. For example, classical
spacetime has an ontological-qua-geometric lattice structure with dimensions
matching those of Minkowski space. The vertical (time) dimension of
this lattice corresponds to logical induction in the regressive direction
and logical deduction in the progressive direction. Progressing down
the lattice corresponds to moving forward in time, and involves a horizontal
(spatial) expansion which accommodates the expanding spatial distinctions
among the contents of spacetime and is thus characterized by entropy.
In this sense, the arrow of time is differentiative, and where (forward)
time is innately differentiative in the classical context, the establishment
of true unity constitutes a violation of entropy. Any deterministic
complexity emerging in the course of machine assembly and operation
is merely a function of distributed dynamical laws intersecting in specific
arguments wh
Brandon
Watson
For Dembski: You brought up the issue of the role of the first person
(and have just recently today posted on the subject on the discussion
site for this book). Would you briefly summarize your own view on this
subject, particularly as it relates to Kurzweil's patternism and strong
distinction between the objective correlates of consciousness and consciousness
in its subjectivity?
Jay
Richards
Ray make be correct that there will be no "explicit" way of
distinguishing his freedom from that of a very good simulation. That
is, we may not be able to come up with a list of necessary and sufficient
criteria that establishes that the entity is free rather than just a
good simulation. But it may be that we have an intuitive capacity of
some sort that cannot be made fully explicit, which, when properly functioning,
allows us to discern the existence of a free agent. I think this is
normally how we do recognize freedom: directly and intuitively, rather
than deductively or inferentially.
Ray
Kurzweil
Response to Jay:
Ray
Kurzweil
We may very well have intuitive capabilities that we have not yet identified
or articulated, but they do take place in our brains and bodies, characterized
as they are by a genome with only 23 million bytes of compressed information,
aided, of course, by humanity's exponentially growing knowledge base,
but a knowledge base accessible to nonbiological intelligences as well.
I don't think a suitably advanced AI will be distinguishable from human
intelligence. And there won't be a clear boundary - we are already putting
machines in our bodies and brains, and that will accelerate. There are
already
Chris
(end of question:) ... Any deterministic complexity emerging in the
course of machine assembly and operation is merely a function of distributed
dynamical laws intersecting in specific arguments which remain spatially
separate throughout (and thus causally independent in any spatial cross-section),
despite being subject to identical localistic laws. What, if anything,
would make you think that this particular kind of complexity could generate
unitary consciousness? And if you consider “unitary” consciousness
beside the point, then can you at least account for the fact that human
consciousness is introspectively and volitionally unitary, while there
is simply no way to effect this in a machine? I.e., on what basis are
machines being compared to human beings possessing unitary, volitional
consciousness with access to volition, emotion, qualia, and other “subjective”
properties?
Ray
Kurzweil
four major conferences on bioMEMS (biological micro electronic mechanical
systems) to put a first generation of nanobots in our blood stream.
These will evolve in increasingly intelligent machines meaning that
merging our brains with our machines will not require surgery.
Ray
Kurzweil
(end of response to Jay)
ISCID
Moderator
Chris' question was broken in two.
Ray
Kurzweil
Response to Chris:
Bill
Dembski
Responding to Brandon: My own view is that intelligence is a primitive
notion and that it finds its expression only through the first person.
It is persons that interpret patterns that are embodied in matter. Materialism
is content with matter and patterns but cannot make sense of the personal
in anything but a reductive way. That's why I asked Ray about how to
ground the first person perspective computationally. It's interesting
that the etymology of "Person" is thought to derive either
from the Persian "persu" (meaning mask) or Latin "personare"
(meaning to sound through). in either case, the person takes on a derivative
and, dare I say it, "emergent" role, not something fundamental.
Nic
Does a superposition of both Yes/No open the door to what could be called
free will?
Ray
Kurzweil
There is no way to access "qualia," we can only access neurological
correlates (i.e., objective correlates) of subjective experiences. But
there is no clear way to experience the link without being that entity
yourself. Consider that we will have machines that are as complex as
humans and indeed based on a thorough reverse engineering of the principles
of operation of humans. (end of response)
Ray
Kurzweil
Response to Nic
Ray
Kurzweil
Well, people have been inspired by the apparent duality of quantum mechanics
to suggest links to consciousness and free will.
Ray
Kurzweil
(end of response)
ISCID
Moderator
Last chance to get your questions in.
Sulu
Jay's freedom seems to be completely freedom of desire; this is still
freedom in that we get what we want, but our desire in the first place
is determined. It is also empirically determined as in "I want
to move my hand; I see my hand move; therefore I control my hand and
get what I want; therefore I have free will (of the first kind)".
I don't see any difference between this and the phenomenon of a machine
fulfilling its professed desires. Intuition is not needed; or if there
is intuition involved, it is also a determined process verified by experience
that machines could implement.
nanci
then how would you address the phenomonon of clarvoyance? Is that just
another type ob brainwave to be constructed-so to speak for a computer
Ray
Kurzweil
Sulu's argument is an old one (which is not to criticize it). We can
introduce some randomness to both humans and machines, but that as Jay
points out still falls short of the conception of free will. It's almost
impossible to articulate Jay's concept of free will, except in the first
person. But we will experience most machines (as we do most people)
Ray
Kurzweil
in the third person.
Ray
Kurzweil
(end of response)
Paul
Kisak
Asymptotes are the proof that God does not trust us;-) I have had experience
with pattern recognition models that base their design, or computational
ceiling, at 10^30 bits per second which many believe is an approximation
of the brains capacity. Correspondingly approximations have been made,
based on the numbers that Ray alluded to earlier (10^11 neurons and
10^14 synapses - which I think is higher now [on the order of 10^16];
yielding an approximate 10^16 bits per second limit. The discrepency
is several orders of magnitude. One model that I have read, attempts
to explain this gap by building on the discovery that all cell walls
have hollow protein microtubules and it is these structures that facilitate
a quantum mechanical computing capability. QUESTION: Have you heard
of this theory, and if yes, what is your opinion of it?
Jay
Richards
Just to be clear, I have been referring to what Sulu called "freedom
of spontaneity." I think it is subjectively clear to us that (at
least sometimes) we can choose between mutually exclusive alternatives,
that things could otherwise than they are.
Ray
Kurzweil
Response to Paul Kisak:
Ray
Kurzweil
The history of this is interesting. Roger Penrose suggested a possible
link of quantum computing (which is computing using quantum ambiguous
"qu bits") as a foundation of consciousness.
Ray
Kurzweil
It was pointed out to him that the computational structures in neurons
were too large to do quantum computing, so he came up with the tubules
as fine enough structures to do quantum computing.
Ray
Kurzweil
It is his objective to show that machines cannot do this because of
this quantum computing capability. There are some problems with this.
First of all, no one has ever shown any evidence for quantum computing
taking place in the tubules,
Ray
Kurzweil
nor any capbility of humans that requires quantum computing. Moreover,
if humans do perform quantum computing in the tubules or elsewhere,
that would not restrict quantum computing from machines.
Ray
Kurzweil
We're already making progress on quantum computing in machines, and
researchers claim to have a 7 bit quantum computer.
Ray
Kurzweil
(end of response).
ISCID
Moderator
Well, it looks like it is just about 5PM here on the East Coast. ISCID
would like to thank Ray Kurzweil, Jay Richards, and William Dembski
for their participation in today's chat. If you would like to continue
chatting after the event, feel free to move over into the General Discussion
room.
Ray
Kurzweil
Thanks, I enjoyed the dialogue. Look forward to reading it over again.
ISCID
Moderator
Thank you, Ray.
ISCID
Moderator
It was a pleasure to have you with us.
ISCID
Moderator
The transcript of this discussion will be posted on ISCID next week.
Ray
Kurzweil
Thanks for having me. Goodbye for now.
Copyright ©
by International Society for Complexity, Information, and Design July
2002.
|
