IEEE Spectrum June, 2008 - 4

For more special features related to our singularity report,
go to http://www.spectrum.ieee.org/jun08/singularityspecialreport

spectral lines
Un-assuming
The Singularity

Boston
Dynamics'
DARPA-funded
robotic dog,
BigDog.
photo: Boston Dynamics

iven the current state of computer science
and robotics, it's hard to understand how
"the singularity" meme has become lodged
in the serious discourse of the technosphere. This
is the idea that, as a consequence of exponentially
accelerating technological innovation and continuously self-improving artificial intelligence, computer
power will outstrip human brainpower, leading to the
end of human culture as we know it. Not a century
from now, mind you, but somewhere
between 2030 and 2045, depending on
whom you talk to.
The concept was framed in its most
tech-savvy form by computer scientist
and science-fiction writer Vernor Vinge
in 1983 in Omni magazine. It has since
morphed into a complicated "theory" that
for some, notably prolific inventor Ray
Kurzweil, includes a posthuman afterlife
in which we abandon our biological selves
and are uploaded into digital and possibly
robotic vessels, there to spend eternity as
cybernetic Methuselahs. It is also thought
by its followers to be inevitable, not merely one of many
possible future scenarios.
The singularity represents an untestable set of assumptions about our near future. So why are so many
willing to take it seriously? That's what we set out to discover in our special report, "The Rapture of the Geeks,"
in this issue. Given that it's the 25th anniversary of Vinge's
seminal work, it seemed like a good time to call upon the
science and technology experts-including Vinge-to get
a sense of the merits and the demerits of the singularity
case. We were particularly interested to learn what, if any,
technology supports the extraordinary claims made by
the singularity's proponents.
What we found is that there's a lot of hyperbole distracting us from the real work under way in nanotechnology,
brain implants, and machine learning. Researchers are,
with some success, making machines more intelligent
and responsive to solving real-world problems. The explosion of disciplines involved in these pursuits gives you
some sense of their complexity. Robotics departments
have now added developmental, epigenetic, or evolutionary
to their names; control and systems are becoming more

www.sPectRum.ieee.oRg

and more intelligent; AI is coursing through the blood
of embodied cognitive science.
But we're still a very long way from understanding how consciousness arises in the human brain, let
alone figuring out how to re-create it in a machine.
We're even a long way from the much simpler goal of
creating autonomous, self-organizing, and perhaps
even self-replicating machines.
Simple locomotion-like walking-has only recently been conquered by roboticists. And there's
still a lot of work to be done to integrate walking with
other functions, like seeing and hearing. An example
that's been causing a sensation on YouTube is Boston
Dynamics' DARPA-funded robotic dog duo, BigDog
and LittleDog. They are the braindogs of IEEE member
and company founder Marc Raibert, who figured out
in the 1980s that robotic running could be controlled
by a few decoupled control laws.
BigDog-which actually looks more like a large
headless spider that's been chopped in half-is designed to help soldiers in the field carry heavy equipment and supplies across rocky terrain. It can run, walk,
and climb and is able to right itself after stepping into
a hole or tripping over a branch, but it can't figure out
how to avoid these obstacles-at least not yet.
LittleDog, on the other hand, is learning to "see" its
environment before taking a walk. Software and sensors help the robot evaluate the surface it's about to
step on so it can decide where to go next. The interesting thing, as our journalism intern Sally Adee noted in
her blog post about LittleDog, is just how much processing time it takes the robopup to make "walk here,
not there" decisions. It's certainly not ready for Tokyo's
busy crosswalks.
This is hard scientific and technical work, and it
involves the deep understanding and use of dozens of
biological and electromechanical systems. The ability to
remain upright while avoiding potholes is important, but
it's just an infinitesimal step along the road to anything
resembling human consciousness.
Wireless communication, ubiquitous computing,
nanotechnology, distributed sensing, and embedded systems are going to converge and deliver wonders. Electronic prosthetic devices and biopharmaceuticals will
help us correct or expand our physical and mental capabilities. Ultimately, we may even learn enough about consciousness to re-create it in a machine and create artificial
vessels for our own minds. But with all we have to do over
the next 30 to 40 years, we don't expect to be hitting the
"Upload to digital heaven" button anytime soon.
Do you? Let us know.
-Susan Hassler
June 2008 * Ieee Spectrum * nA

http://www.spectrum.ieee.org/jun08/singularityspecialreport

Table of Contents for the Digital Edition of IEEE Spectrum June, 2008