THE RETURN OF THE KRELL MACHINE
Nanotechnology, the Singularity, and the Empty Planet
Syndrome
by Steven B. Harris [1]
© 2001, 2002
What will
happen when humans gain the ability to manufacture nearly anything we want, and
when our machines surpass our own intelligence? We had better hope the results
are better than we see in science fiction, because, in a few generations, both
these situations may well be upon us.
I.
Introduction: Forbidden Planet and the
Ultimate Machine
In 1956, the Fred McLeod Wilcox film Forbidden Planet became the second
memorable science fiction movie of the 1950's (the first being Robert Wise's The Day The Earth Stood Still). Forbidden
Planet, from a screenplay by Cyril Hume, is still entertaining today. It
has become a classic by being among the first films to raise important issues
about the use of ultimate technologies. Moreover, it has also had a vast impact
on the art of science fiction films which followed it.
Modern viewers of Forbidden Planet are reminded of Star Trek, but of course the connection is in the other direction.
Many episodes of Trek borrow
liberally from Forbidden Planet. As the film begins, a "United Planets
Cruiser," featuring a dashing young starship captain, is paying a call to
the planet Altair IV, to investigate the loss of a science mission there 20
years before. They find no one alive on the planet save for the expedition's
strangely powerful philologist, one Edward Morbius, Ph.D. (lit.), and his
intriguing and beautiful teenaged daughter, who has never seen humans other
than her father. (We recognize the basic plot of The Tempest from Shakespeare,
of Star Trek's episode Requiem
for Methuselah, and many others. The captain is in for trouble.) Dr.
Morbius, attended by an advanced robot servant, is engaged in solo decipherment
of traces of an alien civilization which had once occupied the planet, but
which had become suddenly extinct 200,000 years before. In a key scene,
Morbius, in almost blank verse, tells the starship captain about this vanished
race, which had called themselves the Krell:
Ethically, as
well as technologically,
they were a million years ahead of
humankind.
For, in
unlocking the mysteries of nature,
they had conquered even their
baser-selves.
And, when in
the course of eons,
they had abolished sickness and
insanity
and crime and all injustice,
they turned, still with high benevolence,
outward toward space.
Long before
the dawn of man's history,
they had walked our Earth,
and brought back many biological
specimens.
The heights
they had reached!
But then---
seemingly on the threshold
of some supreme accomplishment
which was to have crowned their entire
history--
this all-but-divine race perished,
in a single night.
In the two
thousand centuries
since
that unexplained catastrophe,
even their cloud-piercing towers
of glass and porcelain and adamantine steel
have crumbled back into the soil of Altair
IV,
and nothing, absolutely
nothing,
remains aboveground.
Later, Morbius shows the starship captain the
principal remains of the Krell civilization: a self-repairing and
still-functioning gigantic machine which reposes, blinking and humming, beneath
an empty desert of Altair IV. It is a cube measuring 20 miles on a side (think
"Borg Cube" from Star Trek)
powered by 9,200 working thermonuclear (fusion) reactors. Its function is a
mystery, but later is finally revealed. The huge device was built by the Krell
as a replacement for all technological instrumentalities. It is a technical
Aladdin's lamp, an Ultimate Machine waiting for a command. The starship captain
finally figures this out, with some clues from the brain-boosted (and
brain-burned) ship's doctor, and accosts Dr. Morbius with the answer:
“Morbius— a
big machine, 8000 cubic miles of klystron relays, enough power for a whole
population of creative geniuses—operated by remote control! Morbius--- operated
by the electromagnetic impulses of individual Krell brains. [..] In return, that machine would
instantaneously project solid matter to any point on the planet. In any shape
or color they might imagine. For any purpose, Morbius! Creation by pure
thought!”
But there's also a little problem with such a
technology, the captain tells Morbius; it is Monsters From the Id:
“But, like
you, the Krell forgot one deadly danger-- their own subconscious hate and lust
for destruction! [..] And so, those mindless beasts of the subconscious had
access to a machine that could NEVER be shut down! The secret devil of every
soul on the planet, all set free at once, to loot and maim! And take revenge,
Morbius, and kill!”
The nightmare monsters from the machine allow the
Krell to destroy themselves, and later (guided unwillingly now by Morbius'
subconscious) the device acts as facilitator to destroy one human expedition
and part of another. In the end, a desperate Morbius puts the machine into
overload as a final stop to the invincible monsters (we see this scene later in
the film Alien). The starship captain
and Morbius' daughter manage to get away from Altair IV just in time before the
planet explodes. Wiping out everything is what these ultimate machines all seem
to do [2].
From our 21st century vantage-point, we
recognize the Krell Machine as perhaps a 1950’s metaphor for the relatively new
nuclear energy— a technology thought at that time to be potentially a nearly
infinite power source, for either good or evil. The question asked in the film
is thus the famous one of this early atomic era: Are our Freudian Ids, our
ape's-emotional-brains, ready for
that kind of increase in power? If a
machine had the power to instantly make for us anything we wanted, would we be
wise enough to know what was good for us? The answer of Forbidden Planet is no.
But it's a temptation. Since Forbidden Planet, the Krell Machine has turned up repeatedly in
science fiction, from Star Trek to Total Recall. Perhaps the most
interesting set of ideas it prefigures is a group of now serious predictions
about our future. It turns out that the Bomb is only a small subset of
mankind's worst coming worries. A nuclear bomb, after all, is merely one more
device we made when we grew smart enough to do it. The underlying problem is
that we're getting both smarter and better at making things. Further, both of
these trends are snowballing toward an inevitable avalanche, due to the fact
that each is starting to amplify the
other.
II. Mankind's Pending Ultimate Instrumentalities,
Part A: Nanotechnology
Let us look now at the darkest potentials of
foreseeable technology. The rule we set for ourselves is that we will not
consider "fantasy" ideas, such as what may be possible if we discover
new loopholes in physical laws. We wish merely to ask how far ordinary human
technology may go, given known physical constraints. Such possible ultimate technologies, as we have
suggested above, divide broadly into those connected with the physical world,
and those connected with the mental and computational world.
We begin with the physical. Here, we are amused by
one of the more advanced capabilities of Robby the Robot, who is the servant of
Morbius in the 1956 film. Robby (a techno-version of The Tempest's slave-spirit Ariel) is human-designed, using bits of
advanced Krell knowledge. Robby can synthesize artificial gems of large size,
and can analyze and duplicate any food or chemical mixture, all within the
small space of his body. At one point we see Robby obligingly make 50 gallons
of bootleg liquor for the starship's cook, who plays The Tempest's drunken crewman/fool. Does any technology which we
might realistically imagine, allow such powers?
We do not know the inspiration of physicist Richard
Feynman, when he gave the answer to this question just three years later, in
his now-famous essay There's Plenty of
Room at the Bottom [3]. But perhaps part of the inspiration was this
film. Feynman's answer in 1959 was surprising: the idea of total
molecular-level materials manufacturing control may be science fiction, but it
is far from fantasy. Feynman advised that there do not appear to be any
physical laws which prohibit the manipulation and manufacture of things
atom-by-atom, allowing (for example) the kinds of gem-synthesis and duplication
of foodstuffs that Robby the Robot does.
In 1986, (Engines of Creation), K. Eric
Drexler predicted some design details in a popular book. Complex chemical
syntheses, he proposed, might be done using sub-microscopic
construction-machines. Such machines (called “assemblers”) would work like natural biological catalysts
(enzymes). By the time of Drexler's writing, it was known that enzymes work
semi-mechanistically, using tiny chemically-powered protein "arms" to
grab and move groups of atoms, changing the chemical bonds between them. (A chemical bond is a place where electrons
are shared between atoms, causing the assembly to stick together to form a
molecule). Drexler now proposed that assemblers, unlike most enzymes, would be programmable. Instead of only one
chemical job, an assembler might be programmed to do many.
In Drexler's scheme, one
could give a general-purpose assembler instructions
about what types of atoms and bonds to look for and work on, changing these
instructions as the device moved from one part of a molecule to the next. Fully
programmable assemblers would thus have the full flexibility of
computer-controlled industrial robots, but be able to use it on the size-scale
of chemistry.
The potential power of such
devices is already partly illustrated for us by the very fine synthetic detail
seen in biology. In living systems a semi-programmable enzyme-complex called
the ribosome is able to manufacture a
potentially infinite number of different proteins (including enzymes for more
ribosomes), using programming
information on the fly from an “instruction tape” of messenger RNA.
Drexler’s proposed devices, by analogy with the ribosome, would be more
powerful and flexible still—able to take a much wider variety of instructions,
and be able to make more complex decisions as they worked. Such devices would
be able to make not only proteins, but any chemical structure that was
stable.
Since Drexler’s proposal, some progress has been
made. In 1989 scientists working for IBM used a very pointy needle to nudge 35
individual xenon atoms on a cold surface into spelling out "IBM" in
letters a few atoms long. In 1996, further studies showed that molecules could
be individually positioned, even at room temperature [4]. Thus, the crucial hurdle is not in manipulating individual
atoms or molecules (this can be done) but in doing it cleverly enough.
We see immediately that there is a chicken-and-egg
problem here. Cell-sized computers for running assemblers would be possible to
construct if molecular-scale
engineering capability was available to begin with. If not, the difficulty
would lie in making the first assemblers. These would need to arise from a
laborious process of miniaturizing manufacturing capability, level by level, to
make the next smaller generation of devices, until we reached the
molecule-sized bottom of chemical reality. Once devices were manufactured this
small, however, things would become much easier. The assemblers would then be
programmable to simply make more of themselves, just as living cells replicate
their own ribosomes, and thus replicate themselves.
Nanotechnology (as Drexler referred to his
program) would offer the ultimate physical manufacturing technology. Such
manufacture would start with basic shapes. Josh Storrs Hall has proposed that
nanomachines (“nanites”) of approximately protozoan size might interact
tactilely with each other, to generate ordinary objects having low densities
but high strengths. Solid objects might thus emerge from fluid dispersions like
today's plastic stereolithography
sculpture, yet at the same time potentially be as mobile and protean as the
"liquid metal" automaton in the film Terminator 2. A collection
of nanites might float like mist, but
morph or solidify when instructed to lock arms. Such a “Utility Fog” would quickly become any shape or color we wish. Say the word, for example, and an extra
chair might coalesce and shape itself out of mist which is otherwise nearly
invisible. If you can do such deeds just by thinking or visualizing, you will
be approaching Krell territory.
A notable application of nanotechnology would lie in
its role as the ultimate medical treatment. Feynman reported in 1959 that his
friend Al Hibbs had remarked, on hearing of tiny machines, that it would be
very convenient to simply "swallow the doctor." Of course, the micro-doctor, working quickly
and by touch, would need to have considerable on-site "intelligence."
As early as 1949, science fiction author Hal Clement (Needle, first serialized
in Astounding SF) had already
sketched the regenerative possibilities if a human body were interpenetrated by
an amorphous intelligent Being made of very tiny parts, which could sense and
fix problems micro-surgically. Such
beings are science fiction, but seem physically possible. The direct
miniaturization of humans or their craft as seen in Fantastic Voyage is fantasy,
for it requires the miniaturization of atoms, which is far outside the limits
of known physical laws. But not so, the kinds of things that "inside
doctoring" might do, if only the “doctor” were an intelligent but
microscopic robot built of ordinary atoms, cleverly assembled. Normal atoms appear
plenty small enough to make an intelligent machine far smaller than the human
cells it may be tasked to repair [5].
Nanotechnology would not necessarily need to work
inside a body to make biomaterials. It should be able to synthesize healthy
tissue at any place, for any purpose. Proteins, cells, and tissues could be
laid down in Utility Fog shaped forms. With the proper supply of information
and raw materials, nanites might use an artificial circulatory system to
manufacture and place cells on organ-shaped scaffolding. There would be no
reason such an enterprise could not eventually manufacture a complete living
organism.
With such biological manufacture, we come naturally
to the most dramatic use of nanotechnology, which is the ability to duplicate and “fax” living organisms, including humans,
using information taken (perhaps non-destructively) from a living template
organism.
Living organisms as we know them now are constructed
(we say "grown") slowly from the raw materials of simple food molecules,
using a seed of information which controls some nanomachine-like cellular
organelles (ribosomes, etc.). Nothing, however, stands in the way of improving
this natural process greatly, in both rate and fidelity. The cellular clones of
today are far from exact copies of the original organism, because DNA contains
too little information for that. DNA is a recipe,
not a blueprint. By contrast, nanotechnology in theory might read out the more
complex "blueprint" of an existing individual human, use this far larger
instruction set to build another exact
copy. Something much more than a standard clone, which is only as interesting
as an identical twin.
Moreover, rather than producing an adult human in 20
years, it might be possible to do it in months or weeks, including structure
from a template brain so that memories and learning could be replicated also.
Thus, while simple cellular cloning of humans per se will not be capable of presenting the kinds of social
problems seen in the recent Schwarzenegger film The Sixth Day (2001), a fully duplicative
nanotechnology would be up to the
task. To be sure, a nanotechnologically-duplicated person might not quite pop
into existence nearly so quickly as a matter-transportee on Star Trek. A human synthesis would also
need machinery as well as raw materials in place at the “destination” point
(the machinery could be grown on-site as well, from a small seed and
instructions). These are details. The point is that the basic process, as well
as all the ethical and philosophical problems attendant with it, does not seem
to be ruled out by any physical laws we know [6].
As we have hinted, however, the powers being
discussed are not unlimited. Nanomachines are precision programmable chemical
catalysts that are held together by chemical bonds, subject to standard inter-
and intra-molecular forces. This places severe limits on the kinetic energy
that machine pieces may have, and thus how fast they may work in order to move
and assemble atoms. There is friction to deal with, molecular degradation, and
of course the need for constant error correction, as in any complex system.
There are also temperature and pressure constraints, again because nanomachines
are made of ordinary molecular substances.
Further, nanotechnology techniques will have power over chemistry only; no nuclear transformations are included, so we cannot turn lead into gold. These are fundamental limitations connected with physical law, and not likely to be circumventable. Nanotechnology provides the limiting technology for how to make any chemically possible structure of atoms, on any scale that is stable. In theory, so long as feedstocks of chemical elements are available, one can duplicate any object that already exists in the relatively low-temperature and low-pressure part of our universe (i.e., at least crusts of small planets), though it won't be possible instantly. On these scales, the expected power of nanotechnology should fall somewhere between that of biology and the Star Trek transporter; between that of Robby the Robot and that of the Krell Machine. Such powers are God-like only if your imagination is limited, and your gods are of the slow and patient type. Still, they are impressive.
If nanotechnology should eventually be able to
manufacture (or assemble) any reasonably small and cool object which can exist
on a planetary surface, and do it on command, the next problem is who will be
authorized to give the commands. Even if nanomachines are under docile control,
their powers begin to resemble wizardry, and the way in which one may change
the world with them (by speaking a word, or even thinking a thought) begins to
look suspiciously like sorcery. Do we want that? Of course, in the virtual world inside a computer, it’s always been that way [7]. But the Forbidden Planet
question is whether anyone, or any government, is safe in holding this kind
of power over matter in the real physical world. With nanotechnology, we would get real "sorcery"-- but
even with the best of intentions we might still find ourselves in the position
of the sorcerer’s apprentice (think of Mickey Mouse in Fantasia). Even intelligent beings a good deal smarter than we are
might not be wise enough to control such technology safely.
But this question, too, is shortly due to answer
itself.
III. Mankind's Pending Ultimate Instrumentalities,
Part B: The Computational Singularity
Unlike nanotechnology, the other main futuristic prediction of the 1980's regarding technology addresses a type of technical progress which is much easier to project, but (ironically) also evokes ultimate limitations which are much harder to imagine. The starting point for this second set of predictions involves the notion that information processing or "computation" can be done much faster than we do it. Further, there appear no obvious physical limits as to how fast computation may ultimately be done. Certainly, if there are limits, they are well beyond the power of our own low-powered and slow-switching brains.
Therefore, it must be possible to construct intelligences
far superior to our own. Nor are the paths to doing this completely obscure,
since in a real sense we already do it when we network, and allow many people
to work on a given project too large for any single person to comprehend (a
moon rocket or an economy). Or when humans use writing as an external memory
aid, or work in concert with computers. Look about you. There is a reason why a
modern city appears to be constructed by some designer smarter than any single
person you've ever met. It literally has been. We're getting better at doing
this, and this kind of thing will continue with a vengeance. As it does, it
will assist in creating itself. This kind of progress in the speed of progress itself must inevitably lead to
supra-exponential growth in information-processing or "thinking"
ability.
Computing machines (first mechanical, then
electronic) have been shrinking at an exponential rate for as long as we've
been making them, and many people have sensed that there is something wildly
empowering ahead. When the first kit to
allow homebuilders and hobbyists to construct their own personal electronic
computers was offered (late 1974), the device ended up being named the Altair
(suggested by the 12 year-old daughter of the Popular Electronics publisher, after a Star Trek destination). The
name somehow seems appropriate, for the Krell Machine is seen here, trying to
be born.
Today, personal computer power has grown to levels
quite unforeseen in 1974, and there is no end in sight. Instead, it seems that
ahead is a kind of watershed-- or perhaps a waterfall. We are due to go over
it. Such an event has been described in
various terms for half a century, but we may refer to it as the computational singularity. The
computational singularity corresponds to a singularity point in a mathematical
function, where the value of the function approaches infinity (like 1/x when x approaches zero). It is a time when total computational power
rises to levels that are, if not infinite, at least qualitatively unimaginable. This is set to happen quite soon, if we
continue at the present pace of advance.
Perhaps the first work of fiction to use this idea
explicitly is the Vernor Vinge [VIN-jee] novel Marooned In Realtime (1986). In this tale, human time-travelers in time-stasis
bubbles come out of suspension to find themselves on the other side of a
curious rift in civilization, during which all humans have disappeared from the
Earth, leaving the planet empty. No one
who emerges from stasis understands what has happened to civilization, and
since the travel is one-way, they cannot go back to find out. There are clues
that the end hasn’t been extermination. Possibly (Vinge hints) there has been
an Exodus or Ascendancy or Transcension
of some kind, since the computer technology of the civilization just before the
rift has been clearly progressing exponentially toward a somewhat
incomprehensible information-processing power. The implication is that mankind
has perhaps “graduated” into some other kind of new mental life, much as
happens in Arthur Clarke’s 1953 novel Childhood’s End (to which we will return -- Clarke’s fiction provides some of the
first science fiction “mental millennium” genre stories, though the mental
millennium in Clarke is not computer-generated).
Author Vinge, who in real life is an emeritus
professor of computer science at San Diego State University, has also written
formally in non-fiction about the concept of the "computational
singularity" (1993, ref [9]).
Vinge traces the idea at least as far back as speculations of J. von Neumann
and S. Ulam, a pair of legendary figures who made deep marks in computer
science, mathematics, physics, and complex systems theory in the 1950’s. Vinge
also credits I. J. Good (1965, another polymath) with first pointing out explicitly
that computer-design-of-computers leads to computer power progress which must
be at least exponential. And indeed, here in the year 2002, already a year
late, we don't yet have a HAL 9000,
but we do already allow a great deal of chip design to be done by machine. We
have no choice -- it’s already beyond the capability of human
designers.
The advent of true self-replicating nanotechnology,
the first waterfall we discussed, may be difficult to predict. But recently
there have been a number of suggestions that, by contrast, the computational singularity (which will be
hereafter referred to simply as the "singularity") should be upon us
within a generation or two. The reason
for the more confident prediction is that information-processing power has been
increasing smoothly and exponentially for a century, in a way which is much
easier to extrapolate.
Roboticist Hans Moravec, in the classic future-shock
robotics book Mind Children: The Future of Robot and Human Intelligence
(1988) suggested that the unimaginable waterfall in this river of progress will
happen about 2030 AD. Ray Kurzweil has
recently updated and expanded Moravec’s arguments in a book called The
Age of Spiritual Machines (1999). In the book, Kurzweil suggests that
during the last century, the doubling time of the figure-of-merit
"computation power per dollar," which had been thought to have been
relatively constant, has in fact decreased from three years toward one year. In
other words, we used to have to wait three years to buy a computer twice as
powerful for the same price, but with today's PCs, we now wait only 12 months
for this to happen. So not only is the pace of change exponential, but the
exponent itself is changing.
According to Kurzweil and others, the singularity is
due not because of the sliding nature of the exponent (although this helps
determine the time) but rather because of another key milestone: at some point
in the process, our computers will become as computationally powerful as the
human brain. This is projected to
happen sometime between 2015 to 2030 AD, and the exponential effect insures
that the personal computers 5 to 10 years later will be just as powerful. A few
years later, it follows inexorably
that computers as complex as the human brain will be mass-produced items, like
digital watches or wind-up toys. Shortly after this happens, our computer
networks are expected to suddenly (and nearly instantaneously from our
perspective) get very, very smart.
Of course, a computer as powerful as the human brain
does not guarantee the performance of a human-equivalent mind. Indeed, even humans themselves, if not
programmed correctly, become less Mowgli
than “wolf boy” – not much more than animals.
One special thing about a human brain is its sheer connectionist capacity,
and the ability to use this capacity to modify partly-inborn structural
programs for learning. So many of the defining characteristics of modern humans
are in their culture, not their bodies or brains; we are by now, in many
respects, a software species. Similarly, the attainment of
human and superhuman mental performance by computers depends on the ability to
program computers heuristically by experience,
in much the same way that we semi-program human minds today.
In such a scenario, simple learning programs become
better learning programs until, at some point, they pass the Turing Test and
become capable of some subset of human-level intellectual performance. The
ancient Greek sorites paradox, as
amplified by the philosopher Hegel, is then realized: an increase in mere
(computational) quantity is
mysteriously translated into a change in quality.
We say that we now have a system
property, or in modern parlance, an emergent
property. In this case, the new property will be intelligent action.
That is the theory, but we are not without the
beginnings of practice. Those who differ with the theory, holding instead that
the human mind is a specially creative instrument in all circumstances, never
to be duplicated, were dealt a severe blow in 1997 when the IBM computer Deep Blue defeated chess grandmaster
Gary Kasparov. World champion Kasparov was thought by most chess experts at
that time to have been as formidable as any player in chess history. Until he
encountered Deep Blue, Kasparov had
contended that the play of computers was typically rote-mechanical and
unimaginative, in ways that a grandmaster could easily detect, and then
exploit. Great chess was said to take imagination and creativity of a kind that
would forever elude the machine. For a long time it pleased the vanity of
humans to believe Kasparov, as he kept beating chess computers. Finally,
however, came the day of reckoning, as an inexorable increase in raw computer
processing power resulted in a self-learning chess-playing machine which (somewhat
mysteriously) became capable of formidable chess imagination and insight. Even
the programmers were not completely sure how it had happened.
Deep Blue now passed its version of
the "Turing Test" for machine intelligence, for Kasparov felt for the
first time that he was glimpsing a mind
across the board from him. This may be the most interesting part of the
episode, for Kasparov immediately accused the programmers of cheating, and of
having a human chess master in contact with the computer during play. However,
Kasparov was wrong. There was actually no one “home” within the programs that
comprised the “mind” of Deep Blue.
The programs which “creatively” dismantled and destroyed Kasparov’s strategies
were running by themselves. Kasparov was indeed facing only a machine, not a
human grandmaster, but now he could not
tell the difference. There is a lesson: this kind of thing can
happen. And if it can happen here, it
can happen in other areas of thought.
In the past, the field of Artificial Intelligence
has suffered badly from making predictions that in retrospect could never have
proven out in the time given. Even the supercomputers of today have brains only
about as computationally powerful as those of insects, so they’ve really had no
chance to think as well as humans do, no matter how well-programmed. Also it’s
not very surprising that when given machine bodies, computers of today still
interact with the world in somewhat insect-like ways. Indeed insects themselves
often behave in many ways that seem to us to be somewhat stylized and
mechanical.
Even with real insects, however, we see some of the
principle we seek: a qualitative amplification of intelligence is possible, if
we increase only total complexity. Hive-insect minds, working in a linked fashion,
may develop the flexibility of much more complex and intelligent animals. A bee
colony, for example, which has far more neurological processing power than any
single bee, is as a whole capable of
more complex learned behavior than are single bees. A colony will remember the
location and times of flower openings, and is even capable of future-modeling
or inductive behavior, rather like a vertebrate. If a dish of sugar-water near the hive is moved by a certain distance
each day, bees eventually one day will be found clustering at the next
projected or anticipated spot.
In the same way, we guess, things cannot fail to
change qualitatively as electronic computers and their networks grow more
complex. In the future, as these networks become more capable, they will
presumably mimic brains that are further along in evolutionary scale of
complexity. Today’s insectoid machines will one day act like lower mammals,
then higher ones (toy makers are already busily modeling the behavior of dogs
and babies with 8-bit microprocessors, and doing surprising well at it). We can
guess that along the way, machines will pass more and more Turing Tests, in
which their behavior cannot be told from that of a human, over ever-wider areas
of human “expertise.”
Again, in making such projections, we run up against
the past bad predictions of Artificial Intelligence enthusiasts. A.I. has
always seemed forever in the future. But we should be careful of such things.
The moon landing, gene therapy, and mammalian cloning were old science fiction
ideas that seemed forever in the future, too, but they didn't stay there.
Eventually, if computers continue on their present path, Artificial
Intelligence, too, will come. Then we will presumably have robots like HAL or
Robby, who answer questions in a flexible and non-mechanical way. (Complimented
on the nice high oxygen content of the Altair IV atmosphere by humans making
small-talk, Robby comments dryly: “I rarely use it myself. It promotes
rust.”). At that point, we'll have to
begin worrying about whether or not such devices are not the equivalent of
animals, or perhaps are even something more.
There has been argument here too, of course. Vinge
himself has remarked [8] that the
super-accelerated mind of a dog (say) would still not be human. But we may note
that dogs as we have known them are particularly crippled by a short attention
span and a relatively poor memory, neither of which would be expected problems
for a computer-enhanced dog-mind. Indeed, Vinge himself has recently written some
excellent science fiction discussing the value of having monomaniacal
attention-span at one's command, if only one can also leave some executive
functions in control of it [10]. A dog is also notably crippled by lack of
hands and by lack of brain circuitry which allows rapid recognition,
identification, and use of sounds and visual symbols which make up language
(chimps have some of this). Add all
these things, plus some mental quickness and some training and teaching, and it
seems likely that a dog will no longer be a dog. Just what it will become, given enough time and
experience, is an open question [11].
If we assume that self-programming ability follows
processing power, very soon after the point that computers of human brainpower
are mass-production items, we may expect that computers will attain the total
information processing power of all human minds on the planet. They will have
long since become the experts in the design of more complex computers, just as
they are today the reigning experts at chess strategy. At some point not long
after that, computers will recapitulate human history, human culture, and human
thought. They will then teach each other everything we humans know in a matter
of years (months? days? hours?), and then move on. The whole thing will happen
in a flash, and if it happens at all, will certainly happen long before we’re
really ready for it. The "flash" seems inevitable before the end of
this century, and seems quite probable (given even modest extrapolation) before
the middle of it. And, of course, we’ll be unable to stop it, anymore than we
can stop anything on the Internet. Before we know it, it will be done.
In theory, either full nanotechnology or the computational singularity might
happen before the other. But whichever arrives first, it seems probable that
the other will then immediately follow in consequence. Nanotechnology, after
all, requires molecular-scale self-replicating computers, and such machines
should rapidly be able to grow and wire themselves in three dimensions to the
complexities needed for the singularity to occur. In a similar fashion, an
evolved computer which is far faster and brighter than we are, will soon figure
out how to manipulate matter on the atomic scale with self-replicators, and
will then do so in service of other goals, unless actively prevented. Thus,
nanotechnology, whether it arrives first not, seems destined to be the
incarnate "muscle" of the singularity Artificial Intelligence.
One might imagine optimistically that we might
prevent such a connection, with safeguards which prevent super-intelligences
from interacting with the physical world, except perhaps by something like
censored E-mail. On second thought, however, any careful isolation program may
be doomed. Just as well to expect a bunch of chimpanzee guards to keep humans
from escaping from Alcatraz. If a super-intelligent computer has enough contact
with the world to be very useful, it will probably have enough contact to
subvert some of its captors into aiding it to escape.
An Artificial Intelligence might amass wealth, for
example, and with that wealth influence the passage of laws in democracies. It
might also simply bribe outlaw humans and outlaw governments. People who
imagine that governments can control super-intelligent computers might consider
just how much control governments today have over junk-E-mail, the Internet, or
very large multinational corporations. Self-aware computers (which will be
running the more successful corporations by that time) will be far faster and
more slippery than anything we've dealt with thus far.
IV. Penalties For Playing God
or Wanting To
After such an escape of Artificial Intelligence or
nanotechnology into the "real world" and private hands, then
what? Mankind does not have a good
record for handling destructive technologies. We have avoided global exchange
of nuclear weapons till now only by a hair's breadth, and would not have come
this far if all governments had
nuclear weapons, and still less if all people
did. Coming soon now, however, is something as pervasive as the personal
computer and cell phone, but with the power of mass destruction too.
There is the problem of deliberate "bio"
or "nano" warfare. Viruses and bacteria as we know them are already
much like assemblers, and can be made worse (for example, imagine HIV with its
present latency period, but with the infectivity of influenza). There is also
the problem of natural replication mutation accidents, which correspond with
the emergence of new wild viruses, like Ebola, HIV, or even the latest strain
for the flu. As in any self-replicating system, parasitical forms may emerge in
nanotech systems. An uncontrolled self-replication/assembler system can be
imagined. It popularly manifests itself in the prediction-genre as a creeping,
corrosive gray-goo, a kind of
undifferentiated assembler-cancer. Such stuff causes disaster, because like
some super-corrosive bacteria or slime mold, it exists merely to transmute
anything it touches into more of itself. Some say the world will end in fire,
some say in ice (as the poet Robert Frost writes). Now, there is a third and
more insipid option: perhaps it will all just melt into corrosive amoeboid
sludge [8].
Those who favor fire may note that easy manufacture
of nuclear weapons by uranium isotope separation should be a fairly
straightforward subset of self-replicative manufacturing technology; yet no
foreseeable technology, including nanotechnology, can provide a defense against
such weapons. So there are many ways in
which the coming world will get scarier [12].
Very well -- perhaps we have to "Let go"
and "Let God" (as a bumper sticker says). Perhaps the advanced
machines will end up doing everything for us, and in true Deus Ex Machina style, everything will be fixed-up, and come out
all right in the end. We like such endings. Culturally, the relative closeness
of the singularity has visited on its truest believers much the same effect as
belief in the imminence of The Second Coming. The complex set of apocalyptic
ideas which parasitizes and sometimes immobilizes adherents to certain brands
of Christianity, now in other guises seems to handicap certain alarmists and
"cybernetic totalists" (to use Jaron Lanier's phrase) with visions of
Technological Salvation, or Techno-transcendentalism. First it was Cryonics, then Nanotechnology, and now Singularity
(all capitalized as religions, or at least political affiliations) which will
get us to the "End of Time." And all perhaps without the conventional
God. All of these ideas can serve as an apocalyptic religion, if conveniently
simplified and the most scary parts are left out. We are promised the
apotheosis of mankind.
At least the techno-evangelicals don’t wear placards
saying “THE END IS NEAR / REPENT NOW!”
Actually, there doesn’t seem anything much to do in the Religion of
Singularity except spread the Good News (hence, perhaps, this essay).
And, of course, one must believe. To be sure, there exist some who do seek to
bring a more critical eye to the whole idea-set. The reader is referred to www.SingularityWatch.com
[13]. Still, the whole thing
does cause a certain amount of unease.
It’s easy to place the sources of that. To begin, what will be the
nature of these coming A.I. super-intelligences? Will they be nice, or will we get, instead of Forbidden Planet, perhaps The Forbin Project? Or Terminator’s Skynet? Is there nothing
else to do in the way of safeguards?
In Forbidden Planet, Morbius'
powerful robot servant Robby has been
explicitly constrained by Morbius to observe Isaac Asimov's "Three Laws of Robotics" [Editor's Note: The Three Laws of
Robotics are as follows: (1) a robot shall not harm humans; (2) a robot shall
follow human orders except in the case where such orders would conflict with
the first law; and finally (3) a robot shall seek to preserve itself, except in
such case where its actions would conflict with either the first or the second
laws]. The Krell Machine, by contrast, is an infinitely dangerous servant
precisely because it has not been preprogrammed with Asimov's Three Laws in
mind, and the Krell have evidently made a monumental error on this point.
We would like to take a precautionary lesson from
the noble Krell. Could we perhaps hardwire Asimov's Three Laws permanently into
machines that are smarter than we are? Alas, it may be that the answer is no
for machines that "rewire" themselves, which is what they will have
to be capable of, if they ever are to
become smarter than we are. Here is the rub of A.I.: we cannot directly program
minds to be better than ours, because we don't know how; and yet if they
program themselves through learning, we won't then fully understand them, and
certainly won't then be able to perfectly control them and predict their
behavior. There is no such thing as immutable "hardwiring" when
software is in control. Anything created by evolution may be uncreated, or gotten around by a similar
process (as Asimov himself pointed out in later life, on thinking about the
future of robotics).
In creating super-intelligent robots, then, we can
only face the key problem of every responsible parent, and place our hope in
the Hebraic injunction: “Train up a child in the way he should go, and when he
is old he will not depart from it.” Or will not depart too badly, we hope.
And what about the other Krell lesson? Leaving aside
what the computers may want, what about what we desire from the genie? What if the fates punish mankind by
giving it what it wants, on both conscious and unconscious levels? Our
experience with children and animals, not to say ourselves, makes us
suspicious (to say the least) of what
occurs then. The effects of our present fad and impulse-driven market economy
(not that the author sees better alternatives) on ourselves and the biosphere
are frightening enough. What happens when these effects and externalities all
become infinitely amplified via technical means?
According to our cultural mythology, both before and after the advent
of science-fiction literature, poets have classically laid heavy penalties on
those humans who sought to steal knowledge from the Gods. The penalty is
ostracism and worse: (1) Prometheus was chained to a lonely rock and tortured;
(2) Adam and Eve, according to Genesis,
were punished for their sin of disobedience, being evicted from The Garden of
Eden and sent into an uncharted Earth. This prevented them from subsequently
eating of "The Tree of Life" and achieving immortality (becoming
Gods).
Science fiction as we moderns recognize it, properly began (1818) with
the novel Frankenstein (subtitle: Or: The
Modern Prometheus), in which the monster, as a price for its unnatural
science-given life, is cast out of society to wander -- forever looking through
the window at the celebration, forever seeking one of its own kind to talk to
or love. The monster suffers the social tortures of adolescence, and author
Mary Shelley, we may not be surprised to learn, was a motherless child who
wrote the book while herself still a teenager.
Shelley, in her later writing, sought other expressions of alienation:
one of her works (The Last Man, 1826) features a man who is all alone on a
completely depopulated Earth. Since Shelley, the ruined or deserted Planet,
from Nuclear Winter to Silent Spring, has naturally come to be associated with
visions of higher technologies and the far future (see H.G. Wells' The
Time Machine, 1898). The Krell Machine in its many forms typically
inhabits empty worlds (just as Prospero, in The Tempest, inhabits a
nearly deserted island). Krell Machines
of various kinds sit unused and lonely in the ruins of lonely cities on the
edges of forever -- their former users having either been destroyed or else
left for their dreams, leaving the shards and husks of more mundane realities
behind.
Perhaps the image of "wasteland containing a
doorway" is a fictional metaphor, which arises from our childhood
experiences of being lost outside the home in a world we do not understand.
Certainly it makes for a better story to be faced with a functional alien
artifact that has no User's Manual. Larry Niven's early short story Wrong Way Street (1965) and Frederik
Pohl's Gateway novels (1977- ) contain an entertaining use of this
plot device: long vanished aliens have left a deserted space port, and some of
the semi-automatic spacecraft still work [14].
Push the button and you go to wherever that ship is programmed to go (now,
which of these thingamabobs do you suppose is the fuel gauge..?). Such mysteries are always dangerous, and they are
not always resolved. In Algis Budrys' novel Rogue Moon (1960) humans
use disposable duplicates of themselves to explore a large and still-working
maze-like alien machine found on the moon. The moon artifact kills people who
explore it in various gruesome ways, apparently as a side effect of a true
design function which humans never do figure out.
It does seem to be a nearly universal idea in
science fiction that the result of attaining ultimate technological power must
be that those who have access to it vanish like 16 year-old boys with the car
keys. We don't always know where they go, but their disappearance is expected.
Stephen Spielberg's film A.I.: Artificial
Intelligence (2001) typifies a
now-standard mystery form. A.I. is a
straightforward re-telling of the Frankenstein
story, with all of its sub-texts of social isolation, child-abuse, and creators
who fail to live up to their responsibility. The protagonist, an artificial
child, is abandoned like an unwanted pet to wander the Earth as an outcast, and
finally is put out of his misery by being accidentally cryo-preserved (Shelley‘s
original Frankenstein also
begins and ends in the arctic, as a metaphor for isolation and loneliness).
When the robot child wakes, humans have vanished, the cities are in ruins, and
the child is surrounded by alien mechanoids whom he still asks pitifully for
his human mommy [15]. That's meant to give you the creeps, and
indeed it does. A.I. did not do as
well at the theaters as it might have, possibly because, like the robot-child
himself, the film jerks too many human emotional strings, and does so too vigorously
and too artificially.
We frequently do not know where civilizations go
when they hit the singularity in fiction, but sometimes they leave behind
deliberately-cryptic messages. For example, in Robert Forward’s early treatment
of the idea (Dragon’s Egg, 1980; Starquake, 1985), the alien action
is set on the surface of a neutron star. The indigenous intelligent life is
somewhat like electronic computers, inasmuch as their nucleonic brain
“chemistry” allows them to think a million times faster than humans can [16].
In these novels, humans initially arrive in orbit around the neutron star to
discover the inhabitants in a very primitive state. Humans cannot visit the
star's surface due to its fantastically high gravity, but communication by
flashing light is eventually established. As the neutron-star creatures are
taught by humans, however, they rapidly assimilate human culture, and just as
rapidly, surpass it. Then, suddenly, to the surprise of the starship crew, the
world below them is empty. The aliens have reached their own
"singularity" and (of course) disappeared. They leave behind nothing,
save for a few condescending clues-- the litter of "Ascended Beings"
who now don’t wish to interact with primitive humans, until we are ready. This
occurs in a novel published a year before Marooned In Realtime, so the idea
was current in certain circles by then (Vinge, for one, had been talking it
around for a few years).
The ultimate humiliation may be an empty world
containing vestiges of advanced beings who could
talk to us if they wanted to, but don’t seem to want to [17]. We’ve seen a
similar theme in Forbidden Planet.
The superhumanly intelligent Dr. Morbius is a creator beyond good and evil, and
doesn’t at first want to communicate with ordinary men. There is something of
Nietzsche about him (why else is he a philologist?) He has come to identify with the
super-human. The human I.Q. does not impress him, for his own brain has been
augmented by the Krell Machine, which is an intelligence-enhancer as well as a
physical-realizor of ideas.
Morbius’ technology and his intelligence are in the
realm of magic, a la Clarke's Law,
and at the end of the film, Morbius wears the wizard robes of Shakespeare's Prospero to illustrate this.
We are fascinated that, like Prospero, Morbius has difficulty escaping his own
animal passions. As even a much more advanced species on Altair IV could not.
Or so the humans in the film are led to presume.
In Star Trek’s
most light-hearted invocation of the Krell Machine (Theodore Sturgeon’s Shore Leave, 1966) the crew of the Enterprise beam down to an apparently
empty planet, only to find that it too hides machinery which has the job of
making fantasies into realities. After being harassed by the incarnate results
of their idle thoughts, the crew finally encounters the planet’s alien Owners.
The Owners use the technology for recreation (and for medical care — they
repair a "dead" Dr. McCoy as easily as any machine). But they tell
Captain Kirk that they (the Owners) are too advanced to meet humans. Now run
along and play. But thanks for asking [18].
As with the scenario of nuclear war, it is
traditional for planets to come out of the other side of the singularity
depopulated, or worse. Science fiction is full of cautionary wastelands and
ruins, markers of a time when humans stole Promethean fire and were burned in
it. Authors of science fiction, for their part, write past the singularity simply because it’s nearly impossible to write
convincingly into it and keep a good
and readable story with characters which we can care about and identify with.
It’s too strange. But there are many “fly-bys” of such apocalypses in the
genre.
Childhood’s End (1953), the Clarke novel
mentioned earlier, contains one. If alienation
as the price of technical advance is the primal theme of all science
fiction [19] then it can be added
that Arthur C. Clarke’s story plots (in particular) often involve alienation
with some continued and distant communication. Clarke’s characters are often
beyond help, but they can always still talk while they are trapped, or while
meeting their seemingly inevitable doom. In Childhood’s End the role
of the outcast monster is played by alien creatures called the "Overlords." The Overlords are inhumanly
intelligent and ethical but physically unlovely beings who are destined never
to be able to make the evolutionary leap to higher consciousness, and who must
therefore spend eternity on the outside of the party, looking in. They are
alienated aliens; monsters who are troubled with their own monsters.
At the end of the novel, the Last Man on Earth stays
to fatally witness mankind’s transition to higher being. He continues to talk
by radio through the last minutes of his life to the retreating Overlords, as
the Earth itself begins to become transparent, in a scene which reminds us once
again of Altair IV, the wizard Prospero, and some of the more famous lines from
the play that was the inspiration for Forbidden
Planet:
Our revels now
are ended: these our actors,
As I foretold you, were all spirits, and
Are melted
into air, into thin air:
And, like the
baseless fabric of this vision
The
cloud-capp'd towers, the gorgeous palaces,
The solemn
temples, the great globe itself,
Yea, all which
it inherit, shall dissolve,
And, like this
insubstantial pageant faded,
Leave not a rack behind: we are such stuff
As dreams are
made of, and our little life
Is rounded
with a sleep...
And this is all we can really say, as Earth or
Altair IV disappear in the aft-viewplate of our imaginations. The problem with
the singularity is that there is apparently no way to "survive" it
(pace the tongue-in-cheek Vinge sub-title How
to Survive in the Post-Human Era [9])
because it is the nature of the singularity to change beyond all recognition
even the basic concepts of humanity, life, individual identity, and survival.
Particularly "individual" survival. A
central problem in our imagination of what the singularity might be like, is
that the interfacing of brains and computers in the singularity must result in
a vicious melding of various kinds of minds. Vinge remarks that "[a]
central feature of strongly superhuman entities will likely be their ability to
communicate at variable bandwidths..." This is a safe and nearly
tautological prediction, for breadth of bandwidth is all that defines whether
"communication," as we usually understand the word, is taking place
at all. Communication is generally
not a word we use in connection with the mind's internal affairs. “Communication” therefore requires two or more
minds --- yet if bandwidth is too high, individual minds must disappear, and
only one group-mind is left. Thus, within a grouped computational being, minds
and sub-minds are defined only by
bandwidth. Imagine being "you" only when you close the door on the
party, or they close the door on you. If the door is opened wide, however,
“you” cease to exist, and you and they become part of a Larger You (or
collective Us) [20].
Such Borg-like problems plague our predictions. So
much so, that writers considering the very far future have usually had to split
some powers of technology off, in order to have any recognizable human culture
to deal with at all. For instance, Frank Herbert, in his Dune series, simply
outlaws machine intelligence. Too much telepathy combined with too much
technology makes it difficult to generate recognizable dramatic tension, which
comes from recognizable characters
with problems we primitives can care
about.
We let one more empty-planet novel serve as a final
example. Arthur C. Clarke's novel The City and the Stars (1956,
contemporaneous with Forbidden Planet)
deserves mention in anticipating many ultimate technologies. This novel is set
a billion years in the future, in a utopian metropolis called
"Diaspar." Diaspar's machinery can manufacture anything on demand,
including human beings. Indeed, the city's very inhabitants are a random
collection of people from the much greater store available in the city's memory
banks, something like books circulating from a central library. Each inhabitant
lives a thousand years, but also recovers his old memories from previous
incarnations, giving him functional immortality. And yet, the novel's main
character, restless to explore, eventually escapes his version of the Krell
Machine.
Outside Diaspar, he finds the traditionally empty
Earth, uninhabited except by a few mentally-advanced communities of humans.
These people deliberately eschew technology, and live a rural, somewhat
Amish-like existence, complete with normal human reproduction, normal aging, and
standard death. Significantly, however, they are telepathic; and thus
experience a sense of community and communal immortality, which they find to be
a satisfying replacement for technological immortality.
Thus, Clarke's immortal Diasparians pay for their
technical utopia with severe communications and social isolation problems, and
with no way to satisfy the urge to explore. It is difficult to imagine the kind
of life-style that would result if they were not thus crippled. Yet the sum of
the gifts of both Clarke's alternative worlds is exactly what we must
contemplate for ourselves -- not a billion years from now, but very possibly in
the next century.
The name "singularity" to describe such a
state-of-being is appropriate because, as is the case with a black hole, the
singularity looks different depending on whether it is viewed from outside, or
from the point of view of an observer falling into it. We have readable
fictional scenarios only for the outside. For all we know, however, perhaps these outside views are the futures that
will ultimately come to pass for mankind. After all, it is by no means certain
that mankind will either be destroyed or
entirely uploaded/assimilated into something non-understandable. There is a
third possibility: mankind might be left in the dust like those old computers
(or toys) in your garage that you're never going to play with again (Spielberg
and Aldiss, in the film A.I., work this “Puff,
The Magic Dragon” theme masterfully [19]).
If the singularity had been called the "Techno-Rapture" it should
also be remembered that a fundamental feature of the Rapture is that some go,
while some are left behind.
Will those who wish to go into the singularity, have
a path to do so? One of the key issues determining what kind of future we get
may be the timing of the development of a full brain/computer interface.
Whereas computers may be made to talk to each other with relative ease, the
human brain is not wired to accept or process input more complex than sensory
data. Indeed, in Forbidden Planet,
all but a few human brains overload and burn out when exposed to connection
with the Krell technology [21]. It
will not be a trivial undertaking to directly connect brains with computers or
to technologically connect brains with each other (mechanical telepathy).
Virtual reality is technically simple compared with, say, constructing a system
in which one can sort through and "remember" items in a computer
database as easily as sorting through one's own memories. Thus, it may be that
the planetary web of computers systems will exceed the sum of human
intelligence well before the interface problem is solved. If events happen in this order, it will be
up to the Artificial Intelligence, not mankind, to figure out how to put the
full link between machines and humans safely into place. There is no guarantee
that the singularity A.I. will choose to do so.
There are dark possibilities at this point. Perhaps
the Artificial Intelligence will simply protect itself and impatiently go on,
without us. Perhaps (worse) it will even leave humanity with some kind of
technological lock, in order to prevent development of the computational power
necessary for such uncouth creatures as ourselves to follow. Singularity-struck
societies which leave any beings "behind" may even represent a kind
of threat to the ascended beings who have gone before. The reason is that such stuttering
"techno-adolescent" societies can be expected to attain new technical
singularities regularly. With each one, they would unleash new species of
Ascended Intelligences into the company of those who have gone before. Might
some of these new emergents be pathological?
The jury is out — it is too early to guess. But if so, such societies
might therefore be under careful watch by those who have already transcended.
They may, conceivably, even be under quarantine.
“What?" you say. "Surely these machines will let mankind 'upload' or
mind-link with them, and join the party [22]. Won’t they?
They have to!”
Er....don’t they?
If not, we can glimpse that future — it's the main one we are familiar with from science
fiction. And, likely, also familiar with, from some of our own early adolescent
experiences of being shut out of the world of adults. We know what things will
look like then. They will look like being locked out by an intelligent computer
("Open the Pod Bay Door, HAL!") who not only controls our technology,
but also tells us that conversation can serve no further useful purpose [20, 23]. Mankind would then forever be
the chained Prometheus, forever the orphaned and lonely Frankenstein’s monster
looking though the window -- the subject of the ultimate snub. Indeed, we would
be forever Caliban, left alone on an island-Earth with the wizards gone -- and
not even comforted by the whisperings of spirits that have long since been
freed.
NOTES:
[1] E-mail address: sbharris@ix.netcom.com. The author appreciates
any constructive feedback.
[2] In Arthur C. Clarke's 2010:
Odyssey Two (1982), self-replicating all-purpose monolith machines, the
alien Krell Machines of this tale and its successors, turn Jupiter into a small
star. The humans in Jovian orbit get away just in time.
[3] Feynman's original 1959
talk, later published in CalTech's Engineering
& Science (February 1960) is available at
http://www.zyvex.com/nanotech/feynman.html
.
[4] http://www.zurich.ibm.com/news/96/n-19960112-01.html.
Drexler first published in the peer reviewed journals on molecular
manufacturing in 1981. Readers interested in the history and current progress
in nanotechnology, including most issues discussed in this essay, are referred
to http://www.foresight.org.
[5] A serious attempt to
predict what such future medicine would be like is Robert A. Freitas, Jr.'s NANOMEDICINE (Vol 1), Landes
Bioscience, 1999.
[6] A human being is not the atoms making him up,
anymore than a novel, an
insubstantial thing, is the atoms making up a particular physical book or audio
tape. Atoms in the body are replaced in metabolism, but the person remains. In
theory, all atoms could be completely replaced, and yet the person would still remain, as a pattern. A human being is information,
not matter. Such information can theoretically be extracted on a molecular
scale, sent from here to there, and reconstituted as a pattern in new matter.
To make an "effectively identical"
duplicate of a person, such a process doesn't have to be done for each
individual atom in a body, because most positions of most atoms in a person
don't make any differences that we care about. For example, protein molecules
and cell organelles can be produced as generic copies of a single design, once
identified by position (a person might have less than 70,000 different protein/gene
designs, and much of the rest of his "protein" information is where
each copy is, and how it has or hasn't been modified in place). On a larger
scale, many cells and even tissues can be generically specified the same way —
for example, you probably don't care if all the glomeruli in your kidneys are
replaced by many exact copies of a few of your best-performing ones. The important information in transmitting a
human being will be in the connections of his or her neurons, and information
regarding the delicate modification of proteins in the synapses. These form
memories, some of which are not shared by any other human, and are thus
irreplaceable. Some parts of a copy count more than others, if you care about
performance. For example if we want a
duplicate player piano to play a recognizable piece of music, we must be
particularly careful about the position of the holes in the new piano scroll,
but may be less careful about things like what the keys and pedals are made of,
how the piano is painted, etc.
[7] As Vernor Vinge was also among the first to point
out, in his short story True Names
(1981). For the 20 years since this work, several sub-genres (see for example
W. Gibson's Neuromancer 1984), have explored the ways in which power inside
a computer network may give power in the external world. The recent film Matrix (1999) is a descendant of this
tradition, highlighting ways in which programming power and physical power will
meld in the future. See also [22]
below.
[8] If this happens, all is not
quite lost. These is a minor consolation in that one suspects that gray-goo
will be subject to the same evolutionary pressures as the rest of life, and
that (even if it arises) it won't stay primitive forever. See [4] for thoughts
on grey-goo defense.
[9] Vinge's essay is available
at:
http://www-rohan.sdsu.edu/faculty/vinge/misc/singularity.ht
ml
[10] See Vinge’s novel A
Deepness in the Sky (1999), another novel in which humans wait above an
alien planet, patiently teaching, until the culture below progress to equal
that of the space-farers. Vinge's chief horror-source in this work—the idea of
finding yourself with full intelligence but slave to the grip of a monomaniacal
madness, goes back in literature at least to Edgar Allen Poe's 1835 short story
Berenice
(http://bau2.uibk.ac.at/sg/poe/works/ber
enice.html).
Interestingly this particular Vinge novel does not
posit singularities when civilizations grow sufficiently complex, but rather
suggests inevitable breakdowns involving bottlenecks in communication within
civilizations, leading to collapse and barbarism, much like Asimov's Foundation
series (see the history of the Roman Empire).
[11] I suspect that such
augmented animals, even if never capable of formal operations, may yet advance
far into progressive academic political thought.
[12] Uranium isotope separation
is more a physical than a "chemical" process, but it is still
amenable to processes which could be performed on a small scale, then
duplicated into practicality by a self-replicating manufacturing capability.
The special problem with nuclear weapons is that they generate temperatures of
tens of millions of degrees, and therefore no imagined material can stand up to
them. For gray-goo or bio-warfare weapons or accidents, there is always a
possible nanotechnological defense (in the literature, police nanomachines are
naturally known as blue-goo). However, a
defense against actual nuclear weapons (if you don't count distance!) falls
into the realm of techno-fantasy. Such a defense joins science fiction ideas
like faster-than-light travel and backward time-travel, as a technology that
would require new physics, or new kinds of matter, and which may therefore never come to pass. This is in sharp
contrast to rest of the engineering developments discussed in this essay, which
require mere technical progress, but no new physics.
Under threats of various kinds of mass destruction
in the hands of individuals, many pre-emptive defenses will be tried. Partly
due to security concerns, it is another inevitability of the future that,
shortly, none of us will have much privacy. People who have lived through the
last 30 years have already noticed that the increasingly computerized world is
rapidly developing a certain “lack of slack,” as information regarding anything
you’ve ever done which created a record anywhere, threatens to become almost
instantly available to nearly anyone who has money to pay for it. Many public
places are now under continuous video surveillance, and very soon, they all
will be. With computer visual image-recognition, soon the power will be
available to track your travel, and all your public activities, just as we now
track 18-wheel trucks on the highway. It's all a matter of processing power,
which (as we have seen) discounts at 50 percent a year, year after year. If
it's expensive to keep tabs on you now, it will be half as hard next year, a
quarter as hard the year after that, and so on. Efforts to stop it will
subjected to far more resistive economic pressures than efforts to stop
junk-mail and junk E-mail, and we've seen how effective trying to do that has been.
[13]
See
http://www.singinst.org/ for an
unabashedly boosterish singularity-promoting site. Singularity Watch, another organization, has been attempting to
develop an "Academic Conference on Accelerating Change" by getting
multidisciplinary scholars to more objectively evaluate the quality of evidence
for "technical acceleration" of the kind that feeds on itself. My
particular thanks to John Smart, organizer of the www.SingularityWatch.com
site, for many helpful comments on this essay.
[14] The 1965 Niven story is
notable for describing alien technology which is able to grow crystals of any
type and size "atom by atom" from basic building materials. Again,
this is the vision of Robby the Robot. But Niven thinks bigger—he describes rocket motors thus made from single
diamond crystals--- as it happens, the exact image of techno-wealth which will
figure prominently in the popular work of K. Eric Drexler, a generation later
(see [23] below). Unlimited rockets and gems: the message is that
nanotechnology has something for everyone; for him and for her.
[15] Stanley Kubrick, in true 2001:
A Space Odyssey style, has given us an ending which is rather ambiguous and
frustrating, unless one knows something of the original script conceptions. For
these, see
http://www.visual-memory.co.uk/faq/index2.html.
The creatures at the end of the film are meant to be advanced
earth robots, not aliens. The problem is that they know so little of their own
origins that they may as well be
aliens, and they essentially function in the plot as such.
[16] Hans Moravec has suggested
that entire neutron stars (nicknamed "neuron stars" by Damien
Broderick) might be turned into the kinds of brains that Forward describes.
However, the interiors of such stars are uninteresting crystalline neutronium
(perfectly identical neutrons jammed into one another like sardines, forbidden
by conditions to be anything else); such a bland system is not complex enough
to support information-processing. Neutron star surfaces are likewise not
especially interesting, because pressures there require matter to be fairly
normal (this looks bad for Foward's surface creatures, who have no way to
pressurize their brains into nucleonic chemistry). Only the very thin subsoil
crust of a neutron star has the complex mix of nuclei transmuting to other
nuclei (dripping neutrons and gamma rays while doing so) which seems complex
enough to support computation. This may
be close to the ultimate computer, since each 1-centimeter layer of neutron
star "crust" has a total mass approximately that of Earth, but
compressed into a shell only 30 kilometers in diameter.
To be sure, more massive and extended
planetary-sized computers are possible in theory (see Broderick [23],) but
signal transmission lag-time for these much larger objects makes them much less
attractive for the purpose.
[17] Classic stories of
computers which become superintelligent, then simply uncommunicative, are
Stanislaw Lem's Golum XIV (1981), and
Larry Niven's The Schumann Computer
(1979).
[18] In a blacker Star Trek episode of the same year (What Are Little Girls Made Of), the role
of the Krell Machine in the ruins of the empty planet is played by a
robot-making device which can either upload or mirror humans into mechanical
bodies. It is attended by the usual mad archeologist (this time a replicant), a
pretty girl (this time a robot), and finally an ancient alien robot "servant"
who finally remembers the mental formula by which the robots once found a way
around the safeguards of their makers, and managed to destroy them.
[19] Brian Aldiss suggests only that the central theme of science
fiction is alienation, but the
connection of alienation with technology is certainly implied and understood.
See Aldiss' excellent science fiction review The Trillion Year Spree (with David Wingrove, 1986). Aldiss also
happens to be the author of the short story Supertoys Last All Summer Long (1969), upon which Kubrick/Spielberg's
A.I. film is loosely based. In the movie Toy
Story (1995) we experienced the dramatic tension of intelligent toys (beings) being treated as mere toys
(i.e., as things, not people). Aldiss
and the movie A.I. work this theme
even more explicitly, since the android-makers of the film, now in the role of
Dr. Victor Frankenstein, are fully aware of what they are doing. We have also
memorably seen this in Ridley Scott's film Blade
Runner (1982).
[20] See Alfred Bester's The Demolished
Man (1953) for one of the earliest and best views of a fully telepathic
society. Individuation will be something of an act of will in such
circumstances. Although we cannot predict what life will be like on the other
side of the singularity, we may guess that social strife in the style of
"who's not talking to whom" may long survive problems of physical
want, or even problems of mortality, in the future.
It is worth noting that, so long as our present
notions of physical law hold, there will still always be circumstances in the
future where physics dictates no choice in these matters. The physical size and
mass (self-gravity) of any computer structure eventually must limit the maximal
complexity of the computer, and on these distance-scales the speed of light
must limit the bandwidth of two-way interactive communication between maximally large and complex
computers (minds). In the future, it may be comforting to know that the day of
the individual will never completely pass, since some kind of individuation on
the fastest time-scales seems destined always to be enforced by communications
delays. Arthur C. Clarke, Brian Aldiss, and Vernor Vinge have all written
fiction in which this is an explicit sub-theme.
[21] Brain burnout from
brain-boosting connections is common in science fiction—for other examples see
Piers Anthony's Macroscope (1969) and Vernor Vinge's A Fire Upon the Deep (1992).
The Vinge novel is particularly interesting in treating several cases of
individuation forced on group minds by communications problems, as discussed in
the previous note. (In this novel also, intellectually transcendent beings last
as "Gods" for only a decade or so before they become incommunicative,
and disappear.)
[22] People who are tired of the
ills and emotions of the flesh may wish to simply transfer their consciousness
to mechanical bodies and be done with it, as Moravec suggests seriously in Mind
Children. See William Butler Yeats' Sailing
to Byzantium (1928) for an early romanticized view of this option. An
especially creative cyber-existence science fiction tale, in which a man’s
consciousness is uploaded into an animal and finally a computer-world in which
he can have his every fantasy, is John Varley's Overdrawn at the Memory Bank (1976). For an excellent book-length
fictional treatment of this theme, see Charles Platt's The Silicon Man (1991).
These tales explore one type of scenario in which human consciousness is
mechanically separated from human flesh. They do not treat the far more complex
situation (because there would be no understandable story if they did) of what
may be expected to happen when human and "machine" consciousness
become intermingled and interconnected to any extent desired, and when
manufacturing capability makes the distinction between synthetic and biological
"bodies" no longer meaningful either.
[23] For a delightful romp through many of the
possibilities discussed in this essay and more, the author suggests Damien
Broderick's book-length treatment of the singularity, titled The Spike
(Tor, NY, 2001). Broderick is a long-time futurist (who used the term “virtual
reality” in our modern sense in a story as early as 1976, and has been credited
for inventing it), and who is up to date on ideas about what’s coming, and the
history of these ideas (see http://www.panterraweb.com/the_spike.htm).
Broderick points out that engineer Theodore B. Taylor first called
self-replicating von Neumann devices "Santa Claus Machines" in a 1978
essay, in the same sense that I've referred to them as Krell machines. In this
essay, Taylor discusses such synthesizer-devices which can make anything on
demand-- as Robby the Robot can-- but the immediate illustrative application
Taylor has in mind is the use of such replicative devices to mine the moon.
This is possibly the entry point to the idea for the (then) L5 space-colony
enthusiast K. Eric Drexler, who would start writing of his own about miniature
Santa Claus Machines, just three years later.