Nanotechnology appears to hold the key to the Holy Grail of a genuinely lightspeed optical
network that would be up to a hundred time faster than today�s fastest
telecommunications networks.

As if a hundred-fold increase in bandwidth were not enough, we may soon see a
10,000-fold increase in data storage capacity, by storing data in individual atoms.
Spintronics is one of the emerging technologies already contributing to
increased data storage capacity in today�s storage media. Together with the
emerging technology of teleportation, it seems set to contribute also to the
development of quantum
computing
.

Cluster computing is the poor scientist�s supercomputing, but it relies on
hooking up dozens of machines and is power hungry. A California startup has
built a cluster inside a single PC.

Other computing/communications news:

 

  • The silver lining to spam may be in stimulating AI research to produce more
    intelligent machines. 

     

  • We reported in January
    on the IBM-Mayo project to mine patient data and take evidence-based medicine to
    the next level. The Wall Street Journal provides a progress report

     

  • A sprinkling of smart dust may
    be all it takes to run the planet. 
Lightspeed Internet

University of Toronto and Carleton University researchers are using a
nanotech polymer material to build an optical switch that would replace the
electronic switches and routers in optical communication circuits, whose overall
speed will then increase by �up to 100 times the speed of today�s fastest
networks.�

Achieving the �Holy Grail� of an all-optical network will not be easy or
soon, however. Much research has been abandoned because of obstacles and costs,
and some question the need for higher-speed networks when already available 40
Gbps networks �aren�t getting off the ground,� as one commentator said.

Separately, optical-networking company Infinera has developed a hybrid
photonic/electronic integrated circuit on a single chip, that can transmit data
at up to 100 Gbps.

Reference: Gartner, John (2004). �Internet
Heading to Light Speed
.� Wired News, August 17.

Atomic Storage

Scientists in Switzerland and Sweden have succeeded in re-engineering a
single atom, using a low-temperature scanning tunneling microscope and a voltage
pulse to place an electron on an individual gold atom, turning it into a
positively charged ion. They were then able to remove the electron, returning
the atom to its neutral charge state.

The breakthrough could lead to a new way to control the chemical, optical,
and magnetic properties of materials, and to devices that work at the atomic
scale, such as a nonvolatile memory cell that would store information in a
single atom. A atomic-scale memory would increase the amount of data that can be
stored in a given area by 10,000 times, according to the researchers.

Reference: Unknown (2004). �Single Gold Atoms
Altered.� Technology Research News (citing Science), August 12.

Another Step to Quantum Computing

�Some researchers,� says a news clip in MIT�s Technology Research
News
, maintain that �quantum computers . . . will never be built . . .
because controlling individual particles like atoms, electrons and photons is
extraordinarily challenging.� However, American and Dutch scientists have
developed electronic methods of detecting the spins of individual electrons, and
that is a first step toward controlling them. And in another step toward quantum
computing, Austrian physicists have teleported photons over an 800m-long optical
fiber in �real-world� conditions (actually, through a sewer pipe under the
Danube; previous successful efforts have been under strict lab protocols.)
Unlike the teleportation of �Star Trek,� the original photon stays where it is,
but is properties are transferred to a photon at the other end, essentially
recreating a copy of the original instantaneously. This form of information
transfer could be used in quantum computing.

Reference: Unknown (2004). �Chips
Measure Electron Spin
.� Technology Research News, August 20.

Reference: Rincon, Paul (2004). �Teleportation
goes long distance
.� BBC News, August 18.

PC Power

Orion Multisystems, a California start-up company, is offering an under
$10,000, 12-processor desktop �cluster in a box� computer that uses about the
same amount of power as a standard desktop PC but delivers ten times the
performance. It is also offering a 96 microprocessor under-desk model, priced
below $100,000, that can still be plugged into a standard electrical outlet.

The low power consumption is achieved by using Trasnmeta�s new, ultra low
power, Efficeon microprocessors.

The systems are creating quite a buzz in Silicon Valley, with some detractors
and some enthusiasts for the technical approach and the market need.

Reference: Markoff, John (2004). �A PC That Packs
Real Power, and All Just for Me
.� New York Times, August 30.

Spam Forces AI to Next Level

Despite the growing sophistication of spam filters, �It turns out that the
only way to be sure whether something is spam is to look at it like a human,
with all our knowledge of context, language, meaning and intent. In short, you
must be truly intelligent to do the job,� writes Rupert Goodwins for ZDNet
Australia. Presto! �AI has a real job to do: saving the world [from spam].�

University of Melbourne scientists have �stapled a Support Vector Machine to
an e-mail firewall to get a claimed rate of 90 e-mails a second with one error
every 25,000 messages.� What is a �Support Vector Machine�? It is a �fearsome
mathematical construct that has only just escaped from the lab,� and it is �one
of the most powerful methods of handling real-world data within a computer that
has yet been developed.� �If the evil of spam leads to a renaissance of
well-funded research into fundamental knowledge systems�nothing else will do�it
could be the final kick we need to create truly intelligent machines,� he says.

Reference: Goodwins, Rupert (2004). �How spam may feed the thinking machine.� ZDNet Australia, August
25.

Applied AI in Medicine

The Mayo Clinic and IBM have agreed to build a system to mine information
from Mayo�s 4.4 million patient records and enable a doctor to ask, for example,
how patients with the same gender, age, and medical history as the patient they
are currently treating have responded to particular therapies. The system will
use �pattern-recognition and data-mining tools commonly used by direct-mail
companies to determine which consumers get certain offers and by banks to spot
credit-card scam victims within their customer base,� writes William Bulkeley in
the Wall Street Journal.

The new agreement builds on a 2001 agreement under which IBM converted the
patient records�including lab tests and diagnostic images�to computer format.
Going forward, genetic and even proteomic information will be included in the
patient record.

Almost uniquely among American hospitals, Mayo has manually archived and
mined patient records for almost a century, employing human �abstractors� to
comb through the paper records looking for patterns of diseases and treatments.
The new system will be automated and near-instantaneous.

Some experts doubt that the system can do all that is claimed, but IBM�s CEO
sees it as the beginning of a business opportunity �in the hundreds of billions
of dollars.� Artificial intelligence techniques will be built into the system to
�spot and scrub� information that might identify an individual patient �before
humans see it,� writes Bulkeley.

A Mayo researcher hopes the system will help discover answers to such
questions as why lung-cancer drug Iressa shrinks tumors by 50 percent or
more�but in only about 10 percent of patients, and why a drug that cures most
victims of a type of childhood leukemia also kills 2 percent of patients.

Reference: Bulkeley, William M. (2004). �Mayo, IBM
Join To Mine Medical Data
.� Wall Street Journal, August 4.

Control Networks

The cacophony of chatter among wirelessly networked sensors, machines, and
computers is growing into a roar, reports Barnaby Feder in the New York
Times
. Already established in air traffic control systems, control networks
that manage and maintain systems with little or no human involvement are being
applied to lighting systems, construction defect detection systems, agricultural
irrigation and harvesting systems, bridge maintenance, oil tanker maintenance,
and . . . well, �The range of potential market applications is a function of how
many beers you�ve had,� as one source put it to Feder.

As control network components reach the size of dust particles, they will
become even more pervasive as they become �strewn by the thousands on fields and
forests. . . . to monitor forests for fire, warn soldiers of dangerous
substances on the battlefield and alert border guards to activity in remote
areas.� �Strewn� is a key word: Smart dust-based control networks can be
established for orders-of-magnitude less cost, time, and expertise than today�s
networks, especially wired networks. A communication method called �mesh
networking� facilitates this by allowing components to have minimal
communication power, since messages are simply relayed through neighboring
components perhaps only feet or even inches away, until they reach a central
computer.

So far, �most mesh systems have proved far more error-prone in the field than
in research settings,� and signal interference and security, as well as power,
are of major current concern. However, �There�s a whole ecosystem of hardware,
software and service guys springing up� to resolve the issues and set standards,
said one of Feder�s sources.

Feder supplies the following examples of control nets under development:

  • The largest processor of sugar beets in the US is developing networks to
    monitor conditions in beet storage, to detect pockets of heat buildup before the
    rising temperatures start to break down the sugar in the beets. 

     

  • The US Defense Advanced Research Projects Agency is developing a network of
    mobile antitank land mines that automatically shift position to fill gaps after
    other mines have detonated. 

    Reference: Feder, Barnaby J. (2004). �Wireless
    Sensor Networks Spread to New Territory
    .� New York Times, July 26.

 

Leave a Reply

Your email address will not be published. Required fields are marked *