Interview with
Vernor Vinge
Questions by Sander Olson.
Answers by Vernor Vinge.
Vernor Vinge is a renowned
science fiction writer who's works include A Fire Upon the Deep and
A Deepness in the Sky. He was the first person to articulate the concept
of a technological singularity in a talk in 1993. He has worked in the
Department of Mathematical Sciences at San Diego State University, but
now writes science fiction full time.
Question 1: When you
gave your talk on the singularity in 1993, you stated that you would be
surprised if it occurred before 2005 or after 2030. Do you still believe
this?
Yes.
What would you change if
you were to give an updated talk today?
In the 1993 talk and essay,
I described several possible paths to the Technological Singularity. I
still think those are the plausible approaches (with, as before, the probability
that progress along these paths will be intertwined). However, I think
the intellectual power of the civilization as a whole (a kind of emergent
"group mind") is also something to consider. This fits under the "network"
category of the essay, but progress in this area has become more convincing
to me (eg, see Gregory Stock's _Metaman_ non-fiction book or Bruce Sterling's
short story "Maneki Neko").
In the last few years, I
have become more impressed by how much our intelligence proceeds without
introspection. Much of creativity and insight seems to occur without our
conscious intervention. This, together with Marvin Minsky's society of
mind idea, raises the possibility of a superhumanly intelligent critter
whose self-awareness is a much like ours (a few top agents) backed up by
awesome intuition and creativity. (I don't know that this is any more likely
than godlike self-awareness, but this alternative is intriguing.)
Question 2: Let's talk
about some of the objections to the idea of a singularity. What are your
reaction's to Penrose and Searle's arguments against machine intelligence?
My contact with these arguments
has not been deep. I have read Penrose's _Emperor's New Mind_. The
closest I have come to reading Searle is via one of Dennett's critiques!
The Dennett critique seemed very reasonable to me<grin>, but that doesn't
give me grounds to respond about Searle on my own.
I take Penrose's main point
to be that there are physical aspects of mind (eg, quantum effects) that
we have not recognized. This may be. Of course, such complications could
slow down progress toward the Singularity -- but to me the quantum effects
would simply be another thing that we would learn to do with our artifacts.
(By the way, Hans Moravec has publicly debated Roger Penrose on these issues.
I don't know if the record of that debate is online, but it ought to be
very interesting.)
Question 3: In your
paper you speak of the potential of Intelligence Amplification (IA). Isn't
the main problem with IA that the organic "wetware" of the brain simply
isn't compatible with inorganic "dry" electronics?
That depends on how strongly
"incompatible" is meant. To get the bit rates we have between the different
parts of our brain would be very hard. Progress at that level might require
changes starting with the embryo.
You state that a team of
PhDs with a workstation could probably max out any intelligence test ever
written. As computer power and connectivity increases, couldn't this modest
progress continue indefinitely without a singularity?
Possibly. Or it might meant
that we'd get a "soft" takeoff compared to some of the other approaches
to superintelligence.
If the connected person can
quickly pose concrete problems and retrieve results, including images,
then the combo entity can look pretty good. Is it superhuman? I think as
the interface gets better and the computer side gets better this would
more and more look like a real superhuman. For practical purposes it might
be fair to consider the augmentation (after David Brin's terminology) a
kind of "neo-neocortex").
Conceivably, the right way
to look at such a combo would be as a human with a superhuman subconscious
or superhumanly effective intuition. (People often speak of insight coming
"from out of the blue" or "as a gift from above". More likely, the inspiration
came from processes that usurp (or don't use) the part of ourselves that
sustains self-awareness. Of course, self-awareness and motivation are good
for overall control.)
Question 4: What specific
computer technology do you think has the best potential to create intelligence?
Do you think that silicon integrated circuits can produce intelligence,
or do you believe we will need to move to some other technology, such as
carbon nanotubes?
Classical semiconductor may
be able to do it (this conclusion is just from looking at current hardware
trends and taking, say, Hans Moravec's estimate of human computational
power). There are people who think there is a lot of computation going
on
_inside_ neurons,
and that individual neurons are actually very powerful. This could mean
we need something more radical, like
nanoscale technology.
Similarly, if there is a quantum trick involved, that could also push back
onset of the Singularity.
Question 5: Speaking
of computer technology, if Moore's law dies out within the next decade,
wouldn't that essentially kill the concept of a Singularity?
Of course, if hardware progress
gets stymied early, then I agree that it's unlikely we would get the Singularity.
Moore's Law has come to cover a multitude of trends. The basic idea is
about photo-reduction lithography -- which has been a marvelously scalable
technology for a very long time. I could see that running into a wall in
the next few years, perhaps before the
Singularity is achieved.
The vaguer version of "Moore's law" has been applied to a much wider range
of technologies and hardware improvements. In this wider form, I
think hardware improvements with exponential trends can go on for at least
another twenty or thirty years.
Question 6: Even Ray
Kurzweil concedes that computer hardware alone is not sufficient to create
real intelligence. Has there been any real progress in AI software development?
To me the lack of progress
with "software complexity" or "AI software" is by far the strongest argument
against the possibility of the Singularity. If we don't have a Tech Singularity,
then I think in retrospect, "software complexity" will the the cause cited.
(So to me, the "software" issue is a much bigger imponderable than hardware
progress.)
There has been progress with
"software complexity", but it is comparatively pitiful. The strongest reason
to believe that this
problem can be conquered
is simply the existence of solutions in natural world.
In the past, we've seen that
new solutions get _inspired_ by the existence of powerful hardware. For
instance, people thought about numerical solutions for ordinary and partial
differential equations before there was hardware to effectively execute
such codes. When digital computers were developed, it turned out that the
numerical schemes previously imagined were often fatally flawed. Playing
with the new hardware and feeding those insights back to the theorists
and designers gave us effective algorithms. The same thing could
happen with "AI software". It is possible that we are only beginning to
have systems powerful enough to play on the scale of animal intelligence.
Are we making any progress
in understanding how a brain works? In certain respects, we are. I think
it was Francis Crick writing in _Science_ a few years back, who pointed
out that the great difficulty with mind/brain research compared to molecular
biology has been that the mind/brain researchers have had comparatively
miserable measurement technology. With the advent of high resolution (and
multi-aspect) imaging schemes, this problem is becoming less devastating.
Question 7: You argue
that we should see symptoms of the singularity before it occurs. Have we
seen any symptoms yet? If we haven't seen any symptoms, couldn't that be
seen as evidence that a singularity won't occur?
Quite possibly. However,
some symptoms have appeared and some have not (see also the hardware commentary
at the end of my answer to Question 6):
o In terms of group problem
solving powers, the Net constitutes a positive symptom.
o The use of networked PDAs
is a weak positive symptom.
o The proliferation of networked
embedded systems would be a positive symptom (though in 2001, that is still
mainly imminent).
o On the other hand, I see
very few positive symptoms in terms of conquering software complexity.
One thing to watch for would be _large scale_ systems that depend on probablistic
features and training.
Question 8: What is
your assessment of the prospects of Eric Drexler's molecular nanotechnology?
A number of articles in the
latest issue of Scientific American claim that it is completely infeasible.
I don't know how hard Eric's molecular nanotechnology would/will be. I'd
like to see people working on both pure and bio-based approaches.
Yet if it is infeasible,
can we really have groundbreaking changes, even with AI? My intuition is
that we can reach the Technological Singularity even if our nanotechnology
is still soft and mushy. (And if we got the Singularity, then I think there
would be extremely effective nanotechnology very quickly.)
Question 9: Describe
the relationship between AI and Molecular Nanotechnology. You state
in your singularity paper that if true AI cannot be achieved, it will be
the end of real progress. Are you saying that AI must be achieved before
molecular nanotechnology can be feasible?
I think we could get some
form of molecular nanotechnology first. However, I think the full form
of molecular nanotechnology would require that we solve the software complexity
problem.
If the bottleneck to achieving
the Singularity is the need for exceptionally good hardware, it's possible
that there would be an
interaction between nanotechnology
and computational support, and we'd get very good at both at the same time.
Question 10: Can we
have a true singularity without genuine machine intelligence?
Can IA by itself take us
to a singularity? Yes, although as time passes, the machine part of the
IA symbiosis would likely become more and more important -- so in the end
it would probably look a lot like the pure machine scenario. But
the word
"machine" in such a context
is demeaning to the entities involved. Their relation to us would more
accurately be the relationship of oneself to one's zygote, or the evolutionary
relationship of humanity to single-celled critters.
Question 11: Marvin
Minsky, who has spent most of the past half century studying AI, has stated
that with current research we will never achieve true AI. His argument
is that a 1 megahertz machine could become sentient with the right programming,
but that we aren't moving any closer to getting to that programming. What
is your response?
My intuition is that it takes
powerful hardware. However, Marvin is a brilliant guy. Unlike writers
like me, Marvin has actually _done_ great things and has been the advisor
to a crowd of people who have done great things. That makes what he says
about this worth extraordinarily careful attention from anyone thinking
about doing research. In any case, I feel that we need a lot more people
thinking about different lines of research. It would be cool if more of
the brightest new people don't simply follow up with incremental improvements
on what currently looks like the hottest thing. I may be more optimistic
than Marvin, in that even if there is not more "heretical" research, I
think the research enterprise will eventually muddle through (perhaps needing
super giant parallel hardware to finally recognize simple insights that
were there all along :-).
Question 12: What are
your plans for the future?
I want to write science-fiction
full-time, and to observe what is certainly the most interesting and --
potentially -- the happiest time in human history so far.
 |
| About Us | Advertise | Contact
Us By Email: calin [at] nanotech.biz |
|
| Copyright © 2005 Nanotech.Biz |
| Disclaimer: No content, on or affiliated with Nanotech.Biz should be construed as or relied upon as investment advice. While every effort is made to ensure that the information contained on Nanotech.Biz is correct, the operators of Nanotech.Biz make no warranties as to its accuracy. In all respects visitors should seek independent verification and investment advice.
Investing in nanotechnology, investing in nanotech, where do i invest in nanotechnology, how do i invest in nanotechnology, nanotechnology companies, nanotech companies, nanotech products, nanotech stocks, nanotechnology stocks, nanotechnology investment, nanotech investment, nanotechnology investing, nanotech investing, nanotechnology reports and white papers.
|
|