Tom Georgoulias
Interview to Rudy Rucker
for Frontwheel Drive. Austin, Texas, 7/20/1999
Q: I’ve been reading your new nonfiction collection Seek!, and I’d
like to start with some computer science questions for you. You write
about simulated evolution to develop machines that are as intelligent
as their creators, yet today's AI research seems farther from reaching
the goal of intelligent machines than ever. What do you see as the
missing link necessary to bring AI research up to speed with your
visions of intelligent machines?
A: There’s a tendency to think that maybe if we can just throw enough
hardware at the AI problem, then evolution can take care of the rest.
Certainly that’s how God went about making us. We evolved inside a
planetary-sized round-the-clock simulation over maybe a billion years.
The catch is that there is such a great disparity between a desktop
computer and a billion-year planetary analog computation. Even with
the biggest imaginable kinds of increases in our computing power, our
machines will remain very tiny playpens.
So rather than relying on blind evolution to build our intelligent
programs we get into trying to tweak the process. That’s what
traditional AI is all about, trying to find little top-down tricks to
make a program behave more intelligently. But even in this kind of
context, there are scads of program parameters that you don’t really
know the best values for, and this is where simulated evolution can
help you.
Another point worth mentioning is that the stuff we are made of has
been evolving all along as well. New kinds of organic molecules
emerged, for instance. This is analogous to the fact that we are still
feeling around for the best kinds of computer architectures, operating
systems, and evolution frameworks. The evolution of robots is really
happening at a number of levels. And it’s not clear that we’ve really
found the best kind of system to try and evolve a mind on top of.
Neural nets, cellular automata, a soup of LISP strings --- we don’t
know. We just have to keep trying.
One final, encouraging, thought is that, as our machines become
networked into a planetary Web, the collective power of our machines
can experience some synergetic increases. Evolution takes a lot of
machine cycles, and when we can distribute this kind of search to lots
of users, we get a terrific speedup. The trick here is getting people
to run your simulation code. In my novel The Hacker and the Ants, the
evolution code was a kind of virus that took over the chips in
everyone’s TV sets. What if every time someone hit a particularly
juicy porno site, their machine became co-opted into working on
evolving intelligent software?
Q: In your writings about cellular automata (CA), you mention how
parallel processing hardware is best suited for running CA
simulations. There are many supercomputers designed with parallel
architectures, but for the most part engineers keep cranking out more
powerful computers based on the von Neumann architecture. What are the
final challenges left in designing parallel systems and how can they
be overcome so that CA programs can advance even further?
A: Well, I too wonder what ever happened to the dream of parallel
computing. About twenty years ago, the Connection Machine was supposed
to be the big new paradigm, but before long they bagged it and got
into making standard architecture workstations. I’ve never had a
chance to do anything with parallel hardware. I have of course written
a lot of CA code; usually the first thing you do is to set up a dual
buffer system so that you can simulate the parallel updates of the
arrays. And when you think about a CA rule itself, you are indeed
thinking in terms of a parallel computation. If CAs ever found a
really killer app, then the industry would be motivated to make
parallel hardware to run them. Not that there isn’t any such hardware
at all, Xilinx of San Jose, for instance, makes some a field
programmable gate arrays (FPGA) that are supposed to be good for
running CAs. I recently read that a man named Hugo de Garis at
Advanced Telecommunications Research (ATR) in Kyoto is trying to use
them to evolve an intelligent robot cat called Robokoneko. We’ll see
what happens. A lot of times projects like this run into the wall of
how much runtime it would take to actually evolve something truly
interesting. The search spaces are just so superexponentially big. In
any case, I’ve never tried using a FPGA myself. There’s kind of a
limit to how many new operating systems and hardware configurations
you’re willing to learn in one lifetime, and I’m getting awfully close
to maxxed out.
Q: The use of computers and programs like Mathematica have rapidly
advanced the field of mathematics over the last 20 years, bringing
topics such as complexity, chaos, and CA to the front lines. What are
some of the newer areas of research in math that have sparked your
interest?
A. My favorites are chaos, fractals, cellular automata, artificial
life, and higher dimensions. Anything gnarly. I love that computer
science has made mathematics into something like an experimental
science. I was never all that good at proving things, but I love doing
computer experiments. Makes me feel competent. These days I’m wasting
most of my time writing a book with the working title Live Windows:
Games and Graphics with Visual C++ and MFC. I’d sort of like to just
call it How to Write Cool Windows Programs, but Bill Gates has sort of
uncooled the word “cool,” hasn’t he? In fact whenever I write a novel,
I do a search on the text when I’m done to make sure I didn’t slip up
and use the world “cool” in it anywhere. But here I am putting down
Gates, and I’m writing a book using the Microsoft Foundation Classes
(MFC)? Well, you gotta live in the real world. I want to see gnarly
math things on my screen, and hopefully on lots of other people’s
screens, and the best way to get the things out there is with MFC.
After awhile you even get to like it. Kind of a Patty Hearst/Stockholm
Syndrome thing, where prisoners get to be fond of their jailers. What
the hey, MFC is where it’s at. I just hope to God it doesn’t fucking
disappear before I finish my book.
Q. Now tell me something about your other new book, Saucer Wisdom. Is
it a novel?
A: Saucer Wisdom is a cross between a transreal novel and a popular
science book of speculations about the future. It’s my personal
contribution to Millennium madness.
Saucer Wisdom arose from three interests of mine. First of all, I have
a lot of ideas about the course of future technology, and wanted to
write a book about that. Secondly, I’m very dissatisfied with people’s
current ways of thinking about UFOs, and I thought it would be
worthwhile to write a novel which treats them in a more interesting
and amusing fashion. Thirdly, I like to write somewhat
autobiographical books that give transreal representations of various
periods of my life.
So Saucer Wisdom features a main character named "Rudy Rucker." Rudy
is approached by a man named Frank Shook who’s been frequently
abducted by flying saucers. But rather than giving Frank Shook medical
exams and lecturing him on world peace, the aliens have been showing
Frank all sorts of things about our future. Frank gets Rudy to help
work his notes up into a book, a book named Saucer Wisdom.
The main areas of future technology described in the book are
Communication, Biotechnology, Femtotechnology, and Transhumanity. The
material is presented in terms of stories about things that Frank and
the aliens looked at. And I enhanced the text with fifty-seven line
drawings (supposedly by Frank Shook.)
You might well wonder what "femtotechnology" is. This will be the
science of transforming one kind of matter into another; for instance,
of making air into gold or chicken soup. One of my motivations in
writing any kind of science book is always to develop new things to
use in my science fiction, so you can expect to see femtotechnology
turning up in my forthcoming SF novel Realware (Avon, 2000).
Frank’s stories are a grab-bag of sketches and vignettes of little
episodes from our future. And overarching these tales is the story of
Frank and Rudy’s interactions, which are none too serene. At one point
Frank breaks into Rudy’s house and disappears for two years. Frank and
Rudy have their final meeting at . . . where else but the same Devil’s
Tower made famous by Close Encounters of the Third Kind. And after
this meeting, Rudy has a dream in which he finally gains true Saucer
Wisdom. Check it out.