[tt] singular simplicity

[Eugen Leitl]

http://www.spectrum.ieee.org/print/6273

Singular Simplicity

By Alfred Nordmann

This is part of IEEE Spectrum's SPECIAL REPORT: THE SINGULARITY

Take the idea of exponential technological growth, work it through to its
logical conclusion, and there you have the singularity. Its bold
incredibility pushes aside incredulity, as it challenges us to confront all
the things we thought could never come true—the creation of superintelligent,
conscious organisms, nanorobots that can swim in our bloodstreams and fix
what ails us, and direct communication from mind to mind. And the pièce de
résistance: a posthuman existence of disembodied uploaded minds, living on
indefinitely without fear, sickness, or want in a virtual paradise
ingeniously designed to delight, thrill, and stimulate.

This vision argues that machines will become conscious and then perfect
themselves, as described elsewhere in this issue. Yet for all its show of
tough-minded audacity, the argument is shot through with sloppy reasoning,
wishful thinking, and irresponsibility. Infatuated with statistics and
seduced by the power of extrapolation, ­singularitarians abduct the moral
imagination into a speculative no-man’s-land. To be sure, they are hardly the
first to spread fanciful technological prophecies, but among enthusiasts and
doomsayers alike their ­proposition enjoys an inexplicable popularity.
Perhaps the real question is how they have gotten away with it.

The trouble begins with the singularitarians’ assumption that technological
advances have accelerated. I’d argue that I have seen less technological
progress than my parents did, let alone my grandparents. Born in 1956, I can
testify primarily to the development of the information age, fueled by the
doubling of computing power every 18 to 24 months, as described by Moore’s
Law. The birth-control pill and other reproductive technologies have had an
equally profound impact, on the culture if not the economy, but they are not
developing at an accelerating speed. Beyond that, I saw men walk on the moon,
with little to come of it, and I am surrounded by bio- and nanotechnologies
that so far haven’t affected my life at all. Medical research has developed
treatments that make a difference in our lives, particularly at the end of
them. But despite daily announcements of one breakthrough or another,
morbidity and mortality from cancer and stroke continue practically unabated,
even in developed countries.

Now consider the life of someone who was born in the 1880s and died in the
1960s—my grandmother, for instance. She witnessed the introduction of
electric light and telephones, of auto­mobiles and airplanes, the atomic bomb
and nuclear power, vacuum electronics and semi­conductor electronics,
plastics and the computer, most vaccines and all anti­biotics. All of those
things mattered greatly in human terms, as can be seen in a single statistic:
child mortality in industrialized countries dropped by 80 percent in those
years.

So on what do intelligent people base the idea that technological progress is
moving faster than ever before? It’s simple: a chart of productivity from the
dawn of humanity to the present day. It shows a line that inclines very
gradually until around 1750, when it suddenly shoots almost straight up.

But that’s hardly surprising. Since around 1750 the world has witnessed the
spread of an economic system, by the name of capitalism, that is predicated
on economic growth. And how the economy has grown since then! But surely the
creation of new markets and the increasingly fine division of labor cannot be
equated with technological progress, as every consumer knows.

Age of Invention: Technological optimists maintain that the impact of
innovation on our lives is increasing, but the evidence goes the other way.
The author’s grand mother [see photo] lived from the 1880s through the 1960s
and witnessed the adoption of electricity, phonographs, telephones, radio,
television, airplanes, antibiotics, vacuum tubes, transistors, and the
automobile. In 1924 she became one of the first in her neighborhood to own a
car. The author contends that the inventions unveiled in his own lifetime
have made a far smaller difference.

Technological optimists maintain that the impact of innovation on our lives
is increasing, but the evidence goes the other way. The author’s grand mother
[see photo] lived from the 1880s through the 1960s and witnessed the adoption
of electricity, phonographs, telephones, radio, television, airplanes,
antibiotics, vacuum tubes, transistors, and the automobile. In 1924 she
became one of the first in her neighborhood to own a car. The author contends
that the inventions unveiled in his own lifetime have made a far smaller
difference.

Click here for a large version of this timeline [PDF format].

Even if we were to accept, for the sake of argument, that technological
innovation has truly accelerated, the line ­leading to the singularity would
still be nothing but the simple-minded ­extrapolation of an existing pattern.
Moore’s Law has been remarkably successful at describing and predicting the
development of semiconductors, in part because it has molded that
development, ever since the semiconductor manufacturing industry adopted it
as its road map and began spending vast sums on R&D to meet its requirements.
Yet researchers and developers in the semiconductor industry have never
denied that Moore’s Law will finally come up against physical limits—indeed,
many fear that the day of reckoning is nigh—whereas singularitarians happily
extrapolate the law indefinitely into the future. And just as the
semiconductor industry wonders nervously whether nanotechnology really can
give Moore’s Law another lease on life, singularitarians accept that this
will occur as a given and then appropriate the exponential growth curve of
Moore’s Law not only to all the nano- and biotechnologies but to the
cognitive sciences as well.

A typical example is the ­therapeutic development of brain-machine
interfaces. In 2002, people were able to transmit 2 bits per minute to a
computer. Four years later that figure had risen to 40 bits—that is, five
letters—per minute. If this rate of progress continues, the argument goes,
then by 2020 brain communication with computers will be as fast as speech.
This isn’t just the breathless cant of a true believer. The idea that an
enhanced communication of thoughts will exceed speech can also be found in
the 2002 report “Converging Technologies for Improving Human Performance,”
issued by the U.S. National Science Foundation and the Department of
Commerce. It says that such methods “could complement verbal communication,
sometimes replacing spoken language when speed is a priority or enhancing
speech when needed to exploit maximum mental capabilities.” Presumably, the
singularity will be reached soon afterward, when transmission rates exceed
the speed of thought itself, allowing the computer to transmit our thoughts
before we think them.

This fantastic vision works only by ignoring the critical limit, which is the
great concentration you have to muster to send the bits. It is a procedure
far more tedious than speech. To ease that requirement—to make a
brain-machine interface into a true mind-machine ­interface—we’d have to know
a lot more than we do about the relation between specific thoughts and
corresponding physical processes in the brain.

The seductive power of ­extrapolation has also been applied in ways less
spectacu­lar but no less foolish. The “lab on a chip” and other technologies
for biochemical analysis have significantly increased the number of
measurements—blood lipids, for instance—that can be obtained from a single
drop of blood. It’s a fine achievement, no doubt, but visionaries stretch the
imagination when they assume that a second Moore’s Law is about to produce
astounding success stories and a transformation of all medical diagnostics.

Yet that assumption, which extrapolates an extrapolation—Moore’s Law—to
another field, is precisely what lies behind the now commonly expressed fear
that increasing diagnostic powers are creating ethical problems in medicine.
Physicians, we are told, will routinely inform patients of impending diseases
for which they can offer no cure.

Yet in fact the path is very long from quicker blood analysis to
instantaneous detection of the near certainty of a dread disease in a
patient’s future. A lab on a chip may provide mountains of data, but without
great advances in many other fields—notably systems ­biology, ­pathology, and
physiology—no one will be able to do much with it. Doctors already have more
physiological information than they can profitably use.

Both examples of mindless extrapolation constitute wishful thinking. And in
both cases, public debate is diverted from the real moral issues and
quandaries that technology raises.

Rather than dreamaordinary claims made from within their own disciplines with
skepticism and even indignation. But they can find it very hard to maintain
such methodological vigilance in the hothouse atmosphere of a high-stakes
collaboration in which ­researchers want desperately to believe that their
own contributions can have wonderfully synergistic effects when combined with
those of experts in other fields. And so, modest researchers recruit one
another into immodest funding schemes.

The electronics engineer and the physiologist, the cognitive scientist and
the physicist, the economist and the manufacturing specialist—all must take
one another’s statements on trust. They must trust in the contributions from
other disciplines, trust in the power of visions to motivate the cooperation,
trust in techniques and instruments that remain somewhat opaque to their
users, trust in the ­trajectories of technical development.

Where trust has become a virtue even for scientists, there is little
incentive to challenge outrageous claims or to hold singularitarians
accountable. They describe the progressive realization of technical
possibility, after all, and their story has a pleasant ring to it. Indeed,
there is nothing wrong with the singular simplicity of the singularitarian
myth—unless you have something against sloppy reasoning, wishful thinking,
and an invitation to irresponsibility.

For more articles, videos, and special features, go to The Singularity
Special Report.  About the Author

ALFRED NORDMANN, author of “Singular Simplicity”, is a professor of
philosophy and the history of science at Darmstadt Technical University, in
Germany. His interests include the philosopher Ludwig Wittgenstein, the
physicist and philosopher of science Heinrich Hertz, and the birth of new
scientific disciplines, such as nanotechnology.



Sidebar 1

Semiconductors and the Singularity

By Bill Arnold, chief scientist, ASML

A common belief in singularity circles is that Moore’s Law will not only
continue indefinitely but will also apply to other areas of teents on
semiconductors. Sustained by enormous investments in R&D, the law still holds
true, 43 years later, much to the surprise of Moore himself.

In his interesting book The Singularity Is Near, inventor and entrepreneur
Ray Kurzweil accurately reviews the history of Moore’s Law and describes some
of the near-term challenges. Kurzweil then declares his belief that the
industry is on the verge of a new paradigm that again increases the
exponential growth rate of computational power, this time through the
introduction of three-dimensional ­integrated circuits.

He may be right. That idea is very much in line with the views of industry
leaders, who foresee not only stacked memories in the 2010-to-2013 time frame
but also hybrid devices that incorporate logic, sensors, and
microelectromechanical devices in the same 3-D package. An international
research effort known as More Than Moore is now looking for ways to build
extremely high-performance ICs for many different applications by combining
existing technologies on silicon or in the same package. Nanowires in 3‑D
chips will help keep Moore’s Law going by letting designers pack in
transistors vertically as well as horizontally, in much the same way that
skyscrapers in New York City pack in more people by housing them above and
below one another.

Kurzweil goes on to speculate that the advanced 3-D molecular-­computing
devices of the future will more likely be grown in a test tube than made in a
silicon fab. I find this forecast much less plausible. There is no such
process that shows any promise of coming anywhere near silicon’s
manufacturing capabilities and economies of scale. There are research efforts
that aim to build biological structures on silicon using photolithography to
define the lattice on which the structures would grow, like vines clinging to
a lattice. But these are in a very primitive state today.

Self-replication of superintelligent machines is another pillar of the
singularity hypothesis. That kind of reproduction would require the machines
to be able to fabricate superadvanced chips because those chips would form
the basis of the machines’ brains. It’s possible that a lithography scanner
itself could be the prototype of a self-replicating intelligent robot. I
should add here, though, that I have no idea whether truly human-level
artificial intelligence is possible, even if the device density and
parallelism or the number of computations per second achieved by chips comes
to rival those of the human brain.

Researchers will have to meet many, many challenges to keep lithography on
the astounding curve it’s been on for the past four decades. I have no doubt
that they will. Lithography will continue to be the core technology that
drives our future.