INTERVIEW
"We've got things in the computer capable of evolving, reproducing,
metabolizing,
things having complex patterns in space and in time. Are they not
alive?" For a pioneer in
the emerging scientific field of artificial life, that
is the big question
CHRISTOPHER
LANGTON
His first encounter with an artificial life form took place while he was at
Massachusetts
General Hospital. His job involved wheeling dead bodies between
the morgue and autopsy
room. One night about three A.M.-this was after he'd
seen Night of the Living Dead-he and
a co-worker were ferrying the latest corpse
through a dank underground hallway lit by a
single bulb. The body was covered
with a sheet, all very Frankenstein-like, when suddenly
the corpse started
moving. It started to rise! The body . . . sat up! "And it made this
roar!"
says Chris Langton. "I turned to the guy next to me and he was gone. The
double
doors at the end of the hallway were going flap, flap, flap. . . ." No
explanation ever did
turn up. "People in the morgue liked to play jokes," he
says. "You'd go to wheel a body,
open up a drawer, and the body'd sit up. This
one could have been planted, too."
Langton's
second encounter was a little less spooky. Working in Mass General's
Psychiatric Research
lab as a systems programmer, he was trying to make one
computer simulate the operational
structure and functions of a second computer.
These simulations, he realized, involved
reducing a given machine's operations
to a finite set of rules and instructions, a bunch of
abstract logical
relationships. Was there anything, he wondered, whose workings you
couldn't
simulate in this fashion? What was life itself, after all, but a lot of
essentially
dead materials organized so that, somehow, living processes emerged?
If you correctly
simulated the underlying pattern or structure of a living
thing, Langton thought, wouldn't
that simulation itself in some sense be alive?
His third encounter with artificial life
also took place in a hospital, only
this time he was a patient. In 1975, before entering
the University of Arizona,
he crashed his hang glider, breaking 35 bones, including both
legs. As he
regained consciousness, information patterns marched through his head,
exploding
like fireworks across his visual field. "It was as if they were self-existing
entities completely taking over the hardware of my brain," he recalls. He spent
the next
five months recuperating and thinking about what he'd seen.
Years later Chris Langton
established the field of artificial life as a distinct
scientific discipline. In 1987,
while a research fellow at the Los Alamos
National Laboratory, he organized the first
conference on artificial life. More
than 100 scientists brought an entire menagerie of
artificial animals. In
addition to a smattering of robots, there were computer-based
genes, ferns,
flowers, worms, and bugs; there were schooling fish, flocking birds, and
buzzing
bioinformatic bumblebees; there was a warren of artificial foxes and rabbits in
their
own artificial ecology. The workshop was capped with an artificial life
4-H show:
prizewinning entry, the ferns.
Two years later, even before receiving his doctorate,
Langton helped organize a
second workshop. Today artificial life embraces the study of
complex adaptive
systems in all their myriad forms: from prebiotic chemical evolution, to
biological evolution, evolution of languages and cultural systems, to evolution
of global
economies. While some researchers try to get synthetic life going in
chemical media,
Langton prefers to work with computers. His latest project,
which he's collaborating on
with Kristian Lindgren (NORITA, Copenhagen) and
others, attempts to computer-simulate a
complete biological cell,
You can be as skeptical as you want about artificial life until
you see that
simulation. There in living color is a pulsing, undulating cell membrane,
exactly
what you might see watching cell division through a microscope: The cell
wall puckers,
pinches down on itself, and a second cell splits off. This is not
a movie. The wriggling,
dividing structure is a pattern generated by a program
much the way DNA codons generate
biological organisms. Somewhere in the bowels
of the Connection Machine, a massively
parallel supercomputer that sits a few
doors away from Langton's Los Alamos office, exists
the electronic analog of a
living cell,
Says Langton: "We're going to try to capture a
cell's behavior while it
incorporates stuff from the outside-taking in metabolites and
turning them into
cell constituents-turns genetic code into entities that move around
inside the
cell and make it do things like divide, produce offspring with variants, and
die."
Omni: Can you define "artificial life"?
Langton: My first sense of it came around 1979 when
I was trying to describe
what I wanted to do for my Ph.D. It was then a way to study
biological phenomena
by building computer models of them, rather than by studying the real
biological
organisms themselves. It attempted to re-create in some other medium the
processes
important to life, and to study those processes in other mediums or in
the abstract. Today
I define artificial life as the study of artificially
constructed systems that exhibit
behaviors typically thought to be
characteristic of real life.
Omni: Your focus is not the
materials of life, but its basic structure?
Langton: The hardware of life is not really
what life is all about. Biological
things are wet and squishy, so we've come to associate
life with wet and squishy
stuff. But that's because we've seen life only in those
materials, In fact, you
can often separate the material from the behavior it exhibits and
envision other
materials that could exhibit that same behavior. What's important are the
functional relationships between parts. I see no reason why you can't lift
those
relationships from the natural world and emulate them on a computer. You'd
then have a
realization of life in another kind of hardware.
Omni: What's the advantage of studying
living processes in media such as
computers?
Langton:Biologists would love to be able
somehow to rewind the evolutionary tape
back to certain initial conditions and run it
again. You can do some of that
with Drosophila [fruit flies] and E coli [bacteria], but
there's a lot you can't
do. Simulations enable you to restart experiments from the exact
initial
conditions, changing just a single parameter and then seeing that parameter's
effect
on the resulting history. You could start off with the same exact
situations but with
different seeds in the random number generator, for example,
to see the whole envelope of
resulting histories.
Once you get genomes that pass on the information, mutations,
recombinations,
then you ask, What happens? The same thing? Or something different? Do you
see
punctuated equilibrium, long periods of stasis followed by brief periods of
rapid
change? Explosions of diversity followed by the filtering out of
individual lines? Do you
see extinctions? As people have gotten better and
better at implementing these things, you
see all that stuff.
Omni:So extinctions occur naturally, without the intervention of
comets?
Langton: Right. Kristian Lindgren's little evolutionary models indicate it's
plain
as the nose on your face that you get extinctions. Clearly, the earth has
been bombarded
by big things having a huge effect, ruining the days of the local
population. But most of
the extinction record's structure is probably due to
the natural dynamics of population
evolution instead of externally imposed
perturbations. This seems a natural feature of
most evolving systems
experimented with on computers. If we see extinctions in these
simulations,
it's natural to go on to ask, How much of the extinction record can we explain
by natural evolutionary dynamics without invoking external catastrophes?
Omni: Is anyone in
good mad-scientist fashion trying to create a living thing
out of nonbiological
components-in a petri dish, for example?
Langton: People at MIT recently constructed a
chemical system in which molecules
replicate by template synthesis, the way DNA replicates,
only they weren't using
DNA. This wasn't a computer simulation but was done in real
"beakerware." They
used certain kinds of adenosine triphosphate.
Gerald Joyce at Scripps
Clinic Research Institute is trying to build what he
calls an "RNA world" to address a
fundamental problem about the origin of life:
Current life depends upon a tightly coupled
interaction between proteins,
enzymes, RNA, and DNA. The tight coupling is that the DNA
codes for protein
synthesis, whose products themselves decode the DNA and mediate its
replication,
To get proteins you need DNA, but to get DNA you need proteins. How could
this
intricate interdependency have gotten started? The recent discovery by [Nobel
laureate]
Tom Cech that RNA molecules can function as enzymes points at one way
out of the dilemma.
Joyce is trying to construct a completely closed RNA world
in which information-storing RNA
molecules code for RNA enzymes, which in turn
decode the information-storing RNA and
mediate its replication. All without
proteins. Another approach is to get the whole thing
going in a strictly
protein world, Some colleagues here at Los Alamos and at the Santa Fe
Institute
are working on that.
Omni: Would you regard these human-made chemical structures
as alive?
Langton: There's no generally accepted definition of life. That's part of what
we're trying to get at. The more of the phenomenology of life you're able to
capture-in a
computer or test tube-the more you're pushing into this gray area
where it is hard to
decide if they're alive. You know: Well, gee, they look
kind of like life, Maybe they're
not completely alive, but the only thing we
have to compare them to is what evolved on this
planet, this one example. We
really need a class to see what's universal across that class
and what's
accidental in particular instances or members of that class. Any definition of
life we might make based solely on our own experience of life on Earth will be
too narrow.
Omni: But with a computer simulation you don't have a physical entity in front
of you.
Langton:
That's not such a big obstacle. It all depends on one's definition:
Does your definition
have any reference to physical properties or not? Scratch
any biologist and he'll give you
a list of things living entities ought to do:
reproduce, metabolize, be a pattern in time
and space, have complex
organization, be capable of reproducing offspring that are slightly
different
and belong to an adapting and evolving lineage. In The Growth of Biological
Thought,
evolutionary biologist Ernst Mayr provides a classic list of properties
that living things
ought to have. But none are really tied to physical
properties, It's all behaviors.
Probably the biologist will say there's
something else life has to do. He'll add something
new to the list, make a
qualification of how the entity has to go about these things. But
we can make
progress, even if we new generate something on the that biologists admit is
alive,
by forcing them to be more careful about what they mean when they say
"life.
Omni: In
principle, then, is it possible to have life inside a computer?
Langton: There's a strong
and a weak claim about computer simulations of life.
The claims are analogous to similar
claims for artificial intelligence. The
weak claim is that these are only computer models,
tools to help you study real
phenomena, The strong claim is that these processes can be
more than
simulations, that real intelligence and life could be embedded in the artificial
material. The term artificial refers to the material, not to the life.
I believe the
strong claim: To me material is irrelevant. Many different ways
exist to realize any
particular set of functions. Multiple realizability! This
is the functionalist school of
philosophy, either about intelligence or life.
Some people argue: "Life can't be
independent of material. Look at enzymes. So
many of their properties depend on the
chemical interactions and properties of
the atoms involved in specific chains." Yeah, sure,
but there are plenty of
other ways to get complicated structure-function relationships.
Clearly, you
have to realize these functions in some sort of hardware, but the specific
hardware
is often irrelevant to the function itself. Because other materials
also may be viable,
computers could provide a sufficient material basis for
life.
Omni: What arguments do those
who deny the strong claim use?
Langton: A standard argument is that if you simulate a
hurricane or thunderstorm
on a computer, nobody gets wet. But they're missing an important
point here.
Simulations of something like wetness differ from those of something like life
or intelligence in that wetness has very physical attributes. Wetness and
liquidity are
defined by physical properties, whereas life is not necessarily so
defined.
Omni: Still, any
example of a living entity is a physical thing.
Langton: Look, a computer is a physical
thing, too. But a computer can exhibit a
lot of behavioral properties, whereas as yet it
can't exhibit a lot of physical
properties. So certain physical attributes like viscosity
will be hard to get
on computers. I'm so much of a computationalist that I believe you can
have
real wetness in a computer. But you have to drop the specific physical
attributes from
your definition of wetness and define it solely in terms of
behavioral properties. Tom
Toffoli at MIT refers to certain classes of
computers as "programmable matter." Some of my
thesis research showed that this
programmable matter can exhibit solid, liquid, and gaseous
phases of behavior,
just like real materials. So hardware can act wet. It all depends on
definition and how you interpret observations. For me the altered definitions
are much more
powerful and useful than the older, more restricted ones.
Omni: Has anyone ever come out
and said, "This artificial life stuff is
nonsense. You should do something else"?
Langton:
Well, yeah. On one of these radio call-in shows a guy who purported to
be a scientist
said, "There's no such thing. This is just a bunch of scientists
trying to promote their
careers. You can't have life inside a computer. It's
not really life, but a sort of
pattern of energy and magnetic molecules on a
disk in a memory somewhere."
And I said,
"Throughout your life you constantly change the cells in your body,
but there's some
pattern in space and time that persists. Your actual physical
media are pretty transient."
He had some counter to that. He just couldn't buy
the whole thing.
Omni: Is everyone in the
field trying to create life?
Langton: No. Lots of people are not studying life, not trying
to make something
alive when they're looking at an evolving population. They're studying
evolution as a process. The term artificial life covers a lot of things that
living things
do.
Artificial life also covers the analogy between biological evolution and
evolution of
language and culture. In graduate school I had this epiphany-that
words, sentences, and
paragraphs were like genes. Language is like social DNA.
The mechanisms by which language
is transmitted, the basic mutations and
recombinations of words and concepts, have analogs
in biology. More broadly,
social intercourse, whereby cultural information gets passed on
from one
generation to the next, does for cultural information what sexual intercourse
does
for biology. Social intercourse recombines cultural information packets,
putting them in
new contexts in slightly different ways.
Omni: People in artificial intelligence have made
grandiose claims about
producing human intelligence in ten years, and they've failed. Is
there a
lesson here for artificial life?
Langton: The problems in artificial intelligence
proved harder than people
initially thought. But the problem of life may be more solvable
than the
intelligence problem. We know a hell of a lot more about how cells work than
how
the brain works. We know next to nothing about how the brain works. I'm
not claiming
we'll have life within a computer within ten years. But this is
mainly because we may not
have a good definition of life in ten years, not
because we'll be unable to do a fairly
good job simulating the process of life
on computers in ten years.
Omni: What are the
possible dangers of creating artificial life? Could these
things get out of the box and
start eating up the biosphere, unleashing
unspeakable horrors on humankind?
Langton: Some of
these horrors are already being unleashed, and not by people
working in artificial life.
Computer viruses, for example, are one of the
things existing out there closest to
artificial life. In several instances, one
computer virus has overridden another,
generating a virus nobody really wrote.
This was a combination of two viruses, both
viable, that spread around targeting
the same sector of your disk.
Computer network
technology is close to the point where you've got a big
distributed system with powerful
information processors at every node, with no
central controller. These big nonlinear
dynamic systems with spatial
distribution have already thrown out examples of emergent
phenomena where it's
hard to figure out what's going on. Problems the Bell System had with
some of
their switching networks, where the whole system went down for several hours,
may
have been due to nonlinear interactions between switching stations. When
they loaded in
some new software, they had an emergent state come up that sort
of locked out the whole
system.
This is not a virus but an emergent property of the interactions of these
programs
talking to each other. The more you have autonomous decision makers
that take in local
data and make decisions affecting what other agents are
doing, the more that medium is ripe
for the emergence of complex, high-level
phenomena. You can get all kinds of funny
behaviors that crystallize out at the
whole-network level that were completely unintended
and unanticipated by the
program designers.
The same is true for stock-trading programs,
buying and selling programs: The
system is ripe for chaos. Each computer interacts via the
market, and other
computers are looking at the same database, making local decisions ha
affect the
database, too, but in a distributed rather than centralize This nonlinear
dynamic
system can in principle give rise to the spontaneous emergence of
something with a lifelike
dynamic.
So stuff is going to start happening out there. The only way we're going to be
able to understand and control it, and not be swept under the rug by it in ways
we don't
understand, is to study it in these local, small-scale models.
Omni: Recombinant DNA
research was constrained in the early days by guidelines
designed to minimize possible
dangers. Should the artificial life community do
something like that?
Langton: The virus
panel during our second workshop discussed the ethics and
potential risks of working on
these things. People working with computer
viruses, self-reproducing programs, partial
programs, program fragments,
shouldn't turn them loose on the network. At the panel,
Eugene Spafford [Purdue
University] said people who create computer viruses and turn them
loose on the
network are the moral equivalents of those who'd dump a toxic biological virus
into public drinking water. Some people don't yet realize it's a bad thing to
do. These
high-school hackers would never break into a hospital or take some
AIDS virus and dump it
into a reservoir. But they don't see that what they're
getting their jollies with right
now is in the same category. It will in
principle have the same effect down the line.
Anyway, I'm putting into the
proceedings of the second workshop a list of things to avoid
doing if you're
working on computers.
Omni: But some of the very same virus panelists
defended the rights of computer
viruses to exist. What do you make of that?
Langton: I just
can't dismiss those claims out of hand. But I'm also not going
to run right out and
protect the rights of computer viruses. We all murder life
every day, all the time: We cut
the grass, swat flies, poison ants. I don't
poison ants so much anymore, now that I have a
better appreciation for ant
colonies, but I'm trapping mice in my house right now-they're
eating me out of
house and home.
The closer life is to us, the more rights we give it. It's
a very
anthropocentric, chauvinistic view. Only if it's like us does it have a right
to
life, otherwise we get to decide individually whether it lives or dies. Seems
to me most
people will consider computer-based life as pretty far away from us.
But it's worth
addressing philosophically: What are the moral rights of a
process versus a kind of
material? If we had a simulated human being in a
computer that otherwise behaved and acted
like you or me, would it have a right
to electricity? Could we pull the plug?
Omni: So
then, aren't you artificial life guys playing God?
Langton: [Long pause] Well, yeah, in
a way. I have to admit it. In fact,
someone once said to me, "Congratulations, I'm keeping
track of gods, and you
have joined the club. You're an official god in the club of gods
because you
have created a universe-one that exhibits interesting behaviors." But what does
that mean, "playing God"? You can in some sense call artificial life
"experimental
theology." If you create certain sorts of universes, there's no
way in hell-if I can use
the phrase-you could know in advance what's going to
come out of that universe. Dave
Ackley [Bellcore] found this out after he'd
gotten a set of some pretty sophisticated
critters to evolve in his evolutionary
model. When he figured out what he thought the
fittest ones were doing, he
decided to engineer something even better. And when he stuck
his new,
"improved" genotypes in there, they immediately just got eaten up by the other
ones.
What he hadn't taken into account were the ecological interactions among
the creatures.
It's difficult to overestimate the interrelationships of things that evolve in
each other's
presence. Subtle dependencies you weren't aware of are always
there, This is why when you
perturb any part of an ecological situation, it's
difficult to predict the ultimate effect.
Omni: Do you ever worry that you're interfering with the natural order of
things? Gaining
forbidden knowledge?
Langton: What? This notion of artificial, in the sense of made by
humans instead
of nature, is a funny concept. Why do we degrade things we make by calling
them
artificial, as opposed to natural? We're part of nature, and what we do is part
of
nature. But we're not blind watchmakers, we're seeing watchmakers. Nature
is not to be
held responsible because there is no conscious entity capable of
foreseeing consequences.
We, however, are responsible for consciously, actively
taking care to be sure we understand
the consequences of building these things.
Omni: Are you skeptical that evolution is the
only mechanism to account for the
complexity of humans and biological organisms in general?
Langton: I wouldn't say I'm skeptical. There's much to evolution we haven't
understood yet.
Evolution is such a powerful, simple theory, it's just got to
be right. But one thing
we're learning from nonlinear dynamics is that evolution
did not have to discover,
painstakingly, all the components of some complex
organismic structure of behavior.
Aggregates of things interacting in nonlinear
ways make for a situation pregnant with
emergent dynamic possibilities.
Nature's just going to be tripping over these
possibilities right and left.
Like a kid in a candy store, nature probably has had a
surfeit of possibilities
to choose from rather than a difficult time working this, that,
and the other
things out. Evolution works with whole aggregates, large populations.
Having
variants on the plan-even identical things-in cooperation with each other
generates
zillions of different patterns of activity to select from.
Omni: When successful, will
artificial life supplant natural life?
Langton: That depends on whether we decide to
release living, evolving,
autonomous machines with rights to existence into the biosphere.
Right now to
some extent we're populating the biosphere with all kinds of "unnatural"
things:
computers, robots, robot elevators, and trains. But trying to speculate about
the
future of artificial life is like trying to speculate about evolution.
We'll design some
initial things ourselves, but if artificial life really gets
going, it's only logical to
turn over the design process to evolution itself.
Genuinely autonomous artificial life
forms should have the capacity to evolve.
And with that, they could give rise to
intelligent, rational beings. They could
give rise to us, implemented in a different
hardware, or to anything else! I
fully expect that they would.
Future life will probably
involve symbiotic relationships between autonomous
machinery, autonomous people, autonomous
plants existing together in
self-contained capsules. Analogs of the original protocells,
these habitats
will reproduce themselves as they spread through space.
Omni: The
human-machine entity will be analogous to an individual cell?
Langton: Yes. That's how it
happened in the past. Collections of molecules
formed cells; a collection of cells formed
more complicated cells; a collection
of these more complicated cells formed multicellular
organisms. When evolution
takes a really big step, it's this jump from a collection of
individuals at one
level forming a single individual at the next level.
Omni: Let me ask
again: What is the meaning of "artificial life"?
Langton: The larger meaning is that we can
no longer point to ourselves and say,
"We are alive, and those things aren't." Artificial
life doesn't bring life down
to the washing machine, the printing press, and car level. It
doesn't degrade
life; it upgrades machinery to our level. I now have a greater
appreciation for
the potential of machinery. I think machines can achieve the same state
of this
qualitative thing we call life. We can no longer consider ourselves special.
Life
is a property anything can have if it's organized correctly.
Omni: Why is there only one
type of life on Earth: carbon-based life?
Langton: I'm not convinced it is the only kind of
life. IBM, in some sense, is
alive. Sociocultural institutions, in away, constitute
organisms in and of
themselves. The conditions for life to emerge may be coming up all the
time,
all over the planet, but they just get eaten up by carbon-based life forms.
That's
the big advantage of being first on the scene: You get to wipe out the
things that come
after you.
NAME:
Christopher Langton
AGE:
Forty-two
PLACES OF WORK:
Complex Systems Group
Theoretical Division, Los Alamos
National Laboratory; Santa Fe Institute
LANGUAGES FLUENT
IN:
C, LISP, BASIC, Pascal,
FORTRAN, 6502 assembly language, PDP-7 assembly language, etc.
FAVORITE ARTIFICIAL LIFE
FORM I:
Ed McMahon
FAVORITE ARTIFICIAL LIFE FORMS II:
Langton's cell
simulation on the Connection Machine;
Rod Brooks's insects at the MIT Artificial Insect lab
RECENTLY READ:
Mary Shelley's Frankenstein (for the third time):
"I want to understand
Frankenstein, to understand life from the
perspective of the so-called monster."