Are We Enlightened Guardians, Or Are We Apes Designing Humans?
Thanks in part to molecular manufacturing, accelerated developments in AI and brain reverse-engineering could lead to the emergence of superintelligence in just 18 years. Are we ready for the implications -- like possible annihilation of Homo sapiens? And will we seem to superintelligence what our ape-like ancestors seem to us: primitive?
Originally published in Nanotechnology Perceptions: A Review
of Ultraprecision Engineering and Nanotechnology, Volume 2, No.
2, May 8, 2006. Reprinted with permission on KurzweilAI.net, May
22, 2006.
Most students of artificial intelligence are familiar with this
forecast made by Vernor Vinge in 19931: "Within
thirty years, we will have the technological means to create superhuman
intelligence. Shortly after, the human era will be ended."
That was thirteen years ago. Many proponents of super-intelligence
say we are on track for that deadline, due to the rate of computing
and software advances. Skeptics argue this is nonsense and that
we're still decades away from it.
But fewer and fewer argue that it won't happen by the end of this
century. This is because history has shown the acceleration of technology
to be exponential, as explained in well-known works by inventors
such as Ray Kurzweil and Hans Moravec, some of which are elucidated
in this volume of essays.
A classic example of technology acceleration is the mapping of
the human genome, which achieved most of its progress in the late
stages of a multi-year project that critics wrongly predicted would
take decades. The rate of mapping at the end of the project was
exponential compared to the beginning, due to rapid automation that
has since transformed the biotechnology industry.
The same may be true of molecular manufacturing (MM) as self-taught
machines learn via algorithms to do things faster, better, and cheaper.
I won't describe the technology of MM here because that is well
covered in other essays by more competent experts.
MM is important to super-intelligence because it will revolutionize
the processes required to understand our own intelligence, such
as neural mapping via neural probes that non-destructively map the
brain. It also will accelerate three-dimensional computing, where
the space between computing units is reduced and efficiency multiplied
in the same way that our own brains have done it. Once this happens,
the ability to mimic the human brain will accelerate, and self-aware
intelligence may follow quickly.
This type of acceleration suggests that Vinge's countdown to the
beginning of the end of the human era must be taken seriously.
The pathways by which super-human intelligence could evolve have
been well explained by others and include: computer-based artificial
intelligence, bioelectronic AI that develops super-intelligence
on its own, or human intelligence that is accelerated or merged
with AI. Such intelligence might be an enhancement of Homo sapiens,
i.e. part of us, or completely separate from us, or both.
Many experts argue that each of these forms of super-intelligence
will enhance humans, not replace them, and although they might seem
alien to unenhanced humans, they will still be an extension of us
because we are the ones who designed them.
The thought behind this is that we will go on as a species.
Critics, however, point to a fly in that ointment. If the acceleration
of computing and software continues apace, then super-intelligence,
once it emerges, could outpace Homo sapiens, with or without piggybacking
on human intelligence.
This would see the emergence of a new species, perhaps similar
in some ways, but in other ways fundamentally different from Homo
sapiens in terms of intelligence, genetics, and immunology.
If that happens, the gap between Homo sapiens and super-intelligence
could quickly become as wide as the gap between apes and Homo sapiens.
Optimists say this won't happen, because everybody will get an
upgrade simultaneously when super-intelligence breaks out.
Pessimists say that just a few humans or computers will acquire
such intelligence first, and then use it to subjugate the rest of
us Homo sapiens.
For clues as to who might be right, let's look at outstanding historical
examples of how we've used technology and our own immunology in
relation to less technologically adept societies, and in relation
to other species.
When technologically superior Europeans arrived in North and South
America, the indigenous populations didn't have much time to contemplate
such implications because in a just few years, most who came in
contact with Europeans were dead from disease. Many who died never
laid eyes on a European, as death spread so quickly ahead of the
conquerors through unknowing victims.
Europeans at first had no idea that their own immunity to disease
would give them such an advantage, but when they realized it, they
did everything to use it as a weapon. They did the same with technologies
that they consciously invented and knew were superior.
The rapid death of these ancient civilizations, numbering in the
tens of millions of persons across two continents, is not etched
into the consciousness of contemporary society because those cultures
left few written records and had scant time to document their own
demise. Most of what they put to pictures or symbols was destroyed
by religious zealots or wealth-seeking exploiters.
And so, these civilizations passed quietly into history, leaving
only remnants.
By inference, enhanced intelligence easily could take choices about
our future out of our hands, and may also be immune to hazards such
as mutating viruses that pose dire threats to human society.
Annihilation of Homo sapiens could occur in one of many ways:
- The "oops" factor: accidental annihilation at the
hands of a very smart klutz, e.g. by something that is unwittingly
immune to things that kill us, or that is smart in one way, but
inept in others. Predecessors to super-intelligence may only be
smarter than us in some ways, and therein lies a danger. An autistic
intelligence could do us in by accident. Just look at current
technology, where computers are more capable than humans in some
ways but hopeless in others.
- Annihilation in the crossfire of a war-like competition between
competing forms of super-intelligence, some of which might include
upgraded Homo sapiens. One of the early, deadlier competitions
could be for resources as various forms of super-intelligence
gobble up space that we occupy, or remake our ecology into an
environment more suitable to their needs.
- Deliberate annihilation or assimilation because we are deemed
inferior.
If Vernor Vinge is right, we have 18 years before we will face
such realities. Centuries ago, the fate of Indian civilizations
in North and South America was decided in a similar time span. So,
the time to address such risks is now.
This is especially true because paradigms shift more quickly now;
therefore, when the event occurs we'll have less time, perhaps five
years or even just one, to consider our options.
What might we use as protection against these multi-factorial threats?
Sun Microsystems' cofounder Bill Joy's April 2000 treatise, "Why
the future doesn't need us,"2 summarized one field
of thought, arguing the case for relinquishment-- eschewing certain
technologies due to their inherent risks.
Since that time, most technology proponents have been arguing why
relinquishment is impractical. They contend that the march of technology
is relentless and we might as well go along for the ride, but with
safeguards built in to make sure things don't get too crazy.
Nonetheless, just how we build safeguards into something smarter
than us, including an upgraded version of ourselves, has as yet
gone unanswered. To see where the solutions might lie, let's again
look at the historical perspective.
If we evaluate the arguments between technology optimists and relinquishment
pessimists in relation to the history of the natural world, it becomes
apparent that we are stuck between a rock and a hard place.
The ‘rock’ in this case could be an asteroid or comet.
If we were to relinquish our powerful new technologies, chances
are good that an asteroid would eventually collide with Earth, as
has occurred before, thus throwing human civilization back to the
dark ages or worse.
For those who scoff at this as an astronomical long shot, be reminded
that Comet Shoemaker-Levy 9 punched Earth-sized holes in Jupiter
less than a decade after the space tools necessary to witness such
events were launched, and just when most experts were forecasting
such occurrences to be once-in-a-million-year events that we would
likely never see.
Or perhaps we would be thrown back by other catastrophic events
that have occurred historically, such as naturally induced climate
changes triggered by super-volcanos, collapse of the magnetosphere,
or an all-encompassing super-nova.
Due to those natural risks, I argue in my book, Our Molecular
Future, that we may have no choice but to proceed with technologies
that could just as easily destroy us as protect us.
Unfortunately, as explained in the same book, an equally bad "hard
place" sits opposite the onrushing "rock" that threatens
us. The hard place is our social ineptness.
In the 21st century, despite tremendous progress, we still do amazingly
stupid things. We prepare poorly for known threats including hurricanes
and tsunamis. We go to war over outdated energy sources such as
oil, and some of us increasingly overfeed ourselves while hundreds
of millions of people ironically starve. We often value conspicuous
consumption over saving impoverished human lives, as low income
victims of AIDS or malaria know too well.
Techno-optimists use compelling evidence to argue that we are vanquishing
these shortcomings and that new technologies will overcome them
completely. But one historical trend bodes against this: emergence
of advanced technologies has been overwhelmingly bad for many of
the less intelligent species on Earth.
To cite a familiar refrain: We are massacring millions of wild
animals and destroying their habitat. We keep billions more domestic
farm animals under inhumane, painful, plague-breeding conditions
in increasingly vast numbers.
The depth and breadth of this suffering is so vast that we often
ignore it, perhaps because it is too terrible to contemplate. When
it gets too bothersome, we dismiss it as animal rights extremism.
Some of us rationalize it by arguing that nature has always extinguished
species, so we are only fulfilling that natural role.
But at its core lies a searing truth: our behavior as guardians
of less intelligent species, which we know feel pain and suffering,
has been and continues to be atrocious.
If this is our attitude toward less intelligent species, why would
the attitude of superior intelligence toward us be different? It
would be foolish to assume that a more advanced intelligence than
our own, whether advanced in all or in only some ways, will behave
benevolently toward us once it sees how we treat other species.
We therefore must consider that a real near-term risk to our civilization
is that we invent something which looks at our ways of treating
less intelligent species and decides we're not worth keeping, or
if we are worth keeping, we should be placed in zoos in small numbers
where we can't do more harm. Resulting questions:
- How do we instill into super-intelligence 'ethical' behavior
that we ourselves poorly exhibit?
- How do we make sure that super-intelligence rejects certain
unsavory practices as we banned slavery?
- Can we reach into the future to prevent a super-intelligence
from changing its mind about those ethics?
These questions have been debated, but no broad-based consensus
has emerged. Instead, as the discussions run increasingly in circles,
they suggest that we as a species might be comparable to 'apes designing
humans'.
The ape-like ancestors of Homo sapiens had no idea they were contributing
DNA to a more intelligent species. Nor could they hope to comprehend
it. Likewise, can we Homo sapiens expect to comprehend what we are
contributing to a super-intelligent species that follows us?
As long as we continue to exercise callous neglect as guardians
of species less intelligent than ourselves, it could be argued that
we are much like our pre-human ancestors: incapable of consciously
influencing what comes after us.
The guardianship issue leads to another question: How well are
we balancing technology advantages against risks?
In the mere 60 years since our most powerful weapons—nuclear
bombs—were invented, we've kept them mostly under wraps and
congratulated ourselves for that, but we have also seen them proliferate
from at first just one country to at least ten, with some of those
balanced on the edge of chaos.
Likewise, in the nanoscale technology world that precedes molecular
manufacturing, we've begun assessing risks posed to human health
by engineered nanoparticles, but those particles are already being
put into our environment and into us.
In other words, we are still closing the proverbial barn doors
after the animals have escaped. This limited level of foresight
is light years away from being able to assess how to control the
onrushing risks of molecular manufacturing or of enhanced intelligence.
Many accomplished experts have pointed out that the same empowerment
of individuals by technologies such as the Internet and biotech
could make unprecedented weapons available to small disaffected
groups.
Technology optimists argue that this has occurred often in history:
new technologies bring new pros and cons, and after we make some
awful mistakes with them, things get sorted out.
However, in this case the acceleration rate by its nature puts
these technologies in a class of their own, because the evidence
suggests they are running ahead of our capacities to contain or
balance them. Moreover, the number of violently disaffected groups
in our society who could use them is substantial.
To control this, do we need a "pre-crime" capacity as envisaged
in the film Minority Report, where Big Brother methods are
applied to anticipate crime and strike it down preemptively?
The pros and cons of preemptive strikes have been well elucidated
recently. The idea of giving up our freedom in order to preserve
our freedom from attack by disaffected groups is being heavily debated
right now, without much agreement.
However, one thing seems to have been under-emphasized in these
security debates:
Until we do the blatantly positive things such as eliminate widespread
diseases, feed the starving, house the homeless, disenfranchise
dictators, stop torture, stop inhumane treatment of less intelligent
species, and other do-good things that are treated today like platitudes,
we will not get rid of violently disaffected groups.
By doing things that are blatantly humane, (despite the efforts
of despots and their extremist anti-terrorist counterparts to belittle
them as wimpy) we might accomplish two things at once: greatly reduce
the numbers of violently disaffected groups, and present ourselves
to super-intelligence as being enlightened guardians.
Otherwise, if we continue along the present path, we may someday
seem to superintelligence what our ape-like ancestors seem to us:
primitive.
In deciding what to do about Homo sapiens, a superior form of intelligence
might first evaluate our record as guardians, such as how we treat
species less intelligent than ourselves, and how we treat members
of our same species that are less technologically adept or just
less fortunate.
Why might super-intelligences look at this first? Because just
as we are guardians of those less intelligent or fortunate than
us, so super-intelligences will be the guardians of us and of other
less intelligent species. Super-intelligences will have to decide
what to do with us, and with them.
If Vinge is accurate in his forecast, we don't have much time to
set these things straight before someone or something superior to
us makes a harsh evaluation.
Being nice to dumb animals or poor people is by no means the only
way of assuring survival of our species in the face of something
more intelligent than us. Using technology to massively upgrade
human intelligence is also a prerequisite. But that, on its own,
may not be sufficient.
Compassion by those who possess overwhelming advantages over others
is one of the special characteristics that Homo sapiens (along with
a few other mammals) brings to this cold universe. It is what separates
us from an asteroid or super-nova that doesn't care whether it wipes
us out.
Further, compassionate behavior is something most of us could agree
on, and while it is often misinterpreted by some as a weakness,
it is also what makes us human, and what most of us would want to
contribute to future species.
If that is so, then let's take the risk of being compassionate
and put it into practice by launching overarching works that demonstrate
the best of what we are.
For example, use molecular manufacturing and its predecessor nanotechnologies
to eliminate the disease of aging, instead of treating the symptoms.
That is what I personally have decided to focus on, but there are
many other good examples out there, including synthesized meat that
eliminates inhumane treatment of billions of animals, and cheap
photovoltaic electricity that could slash our dependence on oil—and
end wars over it.
Such works are not hard to identify. We just have to give them
priority. Perhaps then we will seem less like our unwitting ancestors
and more like enlightened guardians.
1. The Coming Technological Singularity: How to Survive in the
Post-Human Era http://www-rohan.sdsu.edu/faculty/vinge/misc/singularity.html
2. http://www.wired.com/wired/archive/8.04/joy.html
© 2006 Douglas Mulhall. Reprinted with permission.
| 
