ECE Professor Introduces New 'Desktop
Supercomputing'
University of Maryland (06/24/07)
University of Maryland professor Uzi Vishkin, along with his colleagues at
the A. James Clark School of Engineering, have developed a prototype of
what may be the next generation of personal computers. The prototype is
based on the idea of parallel processing operating on a single chip and is
capable of computing speeds 100 times faster than current desktops. The
prototype uses a circuit board about the size of a license plate and has 64
parallel processors. A parallel computer organization the researchers
developed allows the processors to work together and make programming easy
for software developers. Large-scale parallel processing has been used in
supercomputers for years, but any desktop application of parallel
processing has been difficult because of major programming challenges. In
early June, Vishkin and his Ph.D. student Xingzhi Wen published a paper on
the parallel processing technology for the ACM Symposium on Parallelism in
Algorithms and Architectures, and showcased it at the ACM International
Conference on Supercomputing. "The single-chip supercomputing prototype
built by Prof. Uzi Vishkin's group uses rich algorithmic theory to address
the practical problem of building an easy-to-program multicore computer,"
says Massachusetts Institute of Technology computer science and engineering
professor Charles E. Leiserson. "Vishkin's chip unites the theory of
yesterday with the reality of today." Vishkin believes his technology will
revitalize the computer industry as it could someday include 1,000
processors on a chip the size of a fingernail. "The manufacturers have
done an excellent job over the years of increasing a single processor's
clock speed through clever miniaturization strategies and new materials,"
Vishkin says. "But they have now reached the limits of this approach. It
is time for a practical alternative that will allow a new wave of
innovation and growth--and that's what we have created with our parallel
computing technology."
Click Here to View Full Article
to the top
Billionaire Thinks in Trillions for His Computer
Designs
New York Times (06/26/07) P. C1; Markoff, John
At a high-performance computing conference is Dresden, Germany, Sun
Microsystems co-founder Andreas Bechtolsheim plans to introduce a
supercomputer called the Sun Constellation System that will compete for the
title of the world's fastest computer when installation is finished later
this year. Bechtolsheim remains just as dedicated to computer design today
as when he was a graduate student at Stanford University. "It is hard to
believe that 30 years later I am still working on the same problem,"
Bechtolsheim says. Bechtolsheim has been lauded for his ability to create
elegant and simple computer engineering designs. "A lot of these high-end
systems are superego machines," Bechtolsheim says, referring to an industry
practice of competing to make computers that can perform a single type of
mathematical calculation the fastest, but then struggle when given problems
that require moving significant amount of data between processors.
Bechtolsheim believes he found a solution to that problem by modifying an
industry standard data switch, which allowed any of the 13,000-plus AMD
Barcelona microprocessors to communicate with each other more than 10 times
faster than existing switches. Bechtolsheim described the technological
advancement at an annual retreat of the world's leading supercomputer
designers. "It's a pretty interesting architecture," says University of
Tennessee computer scientist Jack Dongarra, who tracks the world's fastest
computers. Bechtolsheim's newest machine will be tested against IBM's
redesigned version of its BlueGene supercomputer, named BlueGene/P, which
will be installed next year and is expected to break the petaflop computing
barrier, or the ability to execute a thousand trillion mathematical
operations per second.
Click Here to View Full Article
to the top
Computer Scientists Pull a Tom Sawyer to Finish Grunt
Work
Wall Street Journal (06/27/07) P. B1; Gomes, Lee
Computer scientists are using Internet users, including children, to
finish boring and repetitive grunt work by turning the work into games.
The practice, known as "human computation," was developed because humans
still perform certain tasks better than computers. Human computation was
first developed by Carnegie Mellon University professor Luis con Ahn, who
developed the ESP Game. To play the ESP Game, a player is matched with
another unknown user. The two players view an image and must write in a
key word to describe the image. If the two users' entries match, points
are awarded. While it appears to be a simple matching game, in reality,
previously unlabeled pictures collected on the Internet are being assigned
keywords that can be used to retrieve or find the image during a search.
Other researchers are starting to adopt the idea of soliciting Web users to
help with their projects. University of California, San Diego, graduate
student Douglas Turnbull developed a similar game that asks players to
match descriptions of musical pieces so the text descriptions can be used
in a program to help with musical recommendations. Human computing games
have some limitations, however. The games tend to produce very common and
simplistic responses because players are trying to earn points so they
submit what they believe other players will. Players are also of little or
no help when trying to identify technical or scientific images.
Nevertheless, the use of human computing is expanding as Google has
suggested that human responses may help weed out low-quality Web sites from
search results, and Amazon offers a site called "Mechanical Turk" that pays
people tiny amounts for performing simple tasks. Von Ahn says human
computation is only a temporary solution, and that eventually image
recognition and similar computing problems will be solved.
Click Here to View Full Article
to the top
Transparent Transistors to Bring Future Displays,
'E-Paper'
Purdue University News (06/26/07) Venere, Emil
Purdue University and Northwestern University researchers have developed
transparent transistor circuits using nanotechnology. The nanowires were
made of zinc oxide or indium oxide, and unlike traditional metal computer
chips, the thin-film transistors can be made for less and at lower
temperatures, making them well-suited plastic films that melt during
high-temperature processing. The transparent transistors could be used to
created devices such as "e-paper," flexible color screens on consumer
electronics, smart cards, information displays on eye glasses, and
"heads-up" displays on auto windshields. Northwestern University chemistry
professor Vladimir N. Ipatieff and professor of materials science and
engineering Tobin J. Marks say that while other researchers had previously
created nanowire transistors, the metal electrodes in the transistors were
not transparent, which made the structure opaque. "Our study demonstrates
that nanowire electronics can be fully transparent, as well as flexible,
while maintaining high performance levels," Marks says. "This opens the
door to entirely new technologies for high-performance transparent flexible
displays." One possible technology is e-paper, which is designed to mimic
ordinary paper and ink. The difference between traditional displays and
e-paper is that traditional displays backlight to illuminate pixels, while
e-paper reflects light like ordinary paper and can hold text and images
indefinitely without using electricity, but allowing the image to be
changed later. E-paper could be used as a low-cost, energy efficient way
to display information and video. The transparent transistors could also
be combined with another emerging technology called organic light-emitting
diodes to create more vibrant and power efficient televisions.
Click Here to View Full Article
to the top
Which Supercomputers Rule?
CNet (06/27/07) Ogg, Erica
The latest Top500 List of Supercomputers is marked by more turnover than
the last report, but IBM's BlueGene/L at Lawrence Livermore National
Laboratory heads the list again, reaching speeds of 280.6 teraflops.
Cray's Jaguar system at Oak Ridge National Laboratory jumped from No. 10 to
No. 2 at 101.7 teraflops, and the only other supercomputer to top 100
teraflops was Cray's Red Storm at Sandia National Laboratory at 101.4
teraflops. Newcomers to the list include IBM's New York Blue at Stony
Brook University at No. 5, IBM's similar Blue Gene system at Rensselaer
Polytechnic Institute at No. 7, and Dell's Abe PowerEdge 1955 server at the
University of Illinois' National Center for Supercomputing Applications at
No. 8. IBM has six of the top 10 supercomputers, but Hewlett-Packard has
203 of the top 500, or 40 percent. However, HP has a total teraflop sum of
1,202, which IBM nearly doubles with 2,060. The 500 top supercomputers
offer a total performance of 4.92 petaflops, or 1,000 teraflops, which is
up from 3.54 petaflops from the previous list. The latest report will be
released Wednesday at the International Supercomputing Conference in
Dresden, Germany.
Click Here to View Full Article
to the top
Researchers Hope to Help Computers Think Like
Humans
Telegraph-Journal (New Brunswick) (06/25/07) P. B1; Shipley, David
University of New Brunswick computer science professor Michaela Ulieru,
the Canadian Research Chair in adaptive information infrastructures, is
part of an international effort to incorporate the human ability to think
into computers and networks. Assisting Ulieru is Dietmar Dietrich, head of
the Institute of Computer Technology at Vienna University of Technology.
Dietrich uses neuro-psychoanalysis, the use of biology and psychology, to
develop human intelligence in models and machines. "The results of
artificial intelligence or cognitive computing are not enough," Dietrich
says. "We have to find a paradigm shift in this area and that's why I'm
looking at the neuro-psychoanalytic model." Dietrich believes that
engineers looking to create intelligent machines should not try to
re-invent the wheel, but should build on the work being done by
neuro-psychoanalytic scientists. Ulieru says a scientific understanding of
how the human brain works will allow it to be emulated, which can be used
to create intelligent machines that help improve human safety and security.
For example, when a power outage is imminent, instead of shutting down the
system, an intelligent network could self organize and evolve to a high
level of resilience, Ulieru says. Dietrich and Ulieru, and several others,
are organizing a conference called "Emulating the Mind," which will be held
in Vienna in July.
Click Here to View Full Article
to the top
Students Pitch Software Ideas to Gates
IDG News Service (06/26/07) Gohring, Nancy
Microsoft's Bill Gates and Craig Mundie on Tuesday asked questions and
tested software innovations developed by students from Egypt, Japan,
France, Poland, Korea, the United States, and the United Kingdom. The
students are some of the leading competitors in Microsoft's Imagine Cup, a
worldwide software design competition for students. This year over 100,000
students from 100 countries participated in the challenge to build a
product designed to improve education for everyone. The Egyptian team
designed software that would allow teachers to input questions for a test
and automatically generate a test designed for students with special needs.
For example, a math test would be converted to pictures of houses instead
of numbers for dyslexic students. The U.K. project would allow kids to
learn software programming. A user looking at an image of a fish bowl
could move one of the fish and simultaneously see on the side of the screen
the code that makes the fish move. For another project students developed
social networking services that let students with similar interests meet
online, while the U.S. team created a service that allows students to help
each other learn foreign languages, improving their rankings on the site
each time they help another student. The final competition will be held in
Korea in August.
Click Here to View Full Article
to the top
Baby Steps Robots Are a Long Way From Dancing, But U-M
Professor is Taking Them Closer
Ann Arbor News (MI) (06/25/07) Hamon, Amanda
Hollywood often shows robots capable of walking, running, and even
dancing, but in reality bipedal robots still have difficulty operating
smoothly. However, University of Michigan electrical engineering and
computer science professor Jessy Grizzle is working to improve robotic
dexterity. Grizzle is working on a project that he hopes will provide an
invaluable contribution to robotics, and to education as a whole. The
unnamed robot is a collaborative effort between the University of Michigan
and Carnegie Mellon University. "The idea is to make it energy efficient,
and to study the dynamics of running," Grizzle says. If everything works
as designed, the robot will be the first bipedal machine with
spring-infused joints and feedback control, which would allow the robot to
recover if pushed off course. After being tested, the robot, which is
funded by a $450,000 National Science Foundation grant, will be used as a
teaching aid in Michigan. The project was started in 2004, and when
completed the robot will stand 5 feet, 9 inches and weigh 150 pounds,
though there is still significant work still to be done. "The challenge
for me is writing the algorithms that allow the robot to move smoothly and
to run," Grizzle says. Grizzle says the technology could lead to the
creation of robots capable of going where humans are unable. A better
understanding of bipedal motion could also lead to improved prosthetic limb
technology and new approaches to stroke and accident victim rehabilitation
methods.
Click Here to View Full Article
to the top
GPLv3 Upgrade Set for Friday
InfoWorld (06/26/07) Krill, Paul
The Free Software Foundation is likely to release the new update to the
GNU GPL (General Public License) on Friday. GPL version 3 for open-source
software offers copyright technology with uniformity for everywhere in the
world, and ensures that users will be able to modify software installed on
personal computers or household devices. The controversial update protects
contributors to free software from being sued for patent infringement, and
allows users to copy Apache-licensed code into GPL projects. Still, there
are some concerns whether developers will migrate to GPLv3. Only a small
number of projects are likely to use the license at this time, says Black
Duck Software CEO Doug Levin, adding that the number could change. "GPL
version 3 does clean up a lot of things about GPL that were sort of
implicit in previous and become much more explicit here," says Mark
Spencer, chairman and CTO of Digium. "There are some complications around
the patent terms that place some additional requirements upon the
developers and distributors that we need to fully understand in more
detail."
Click Here to View Full Article
to the top
Puzzles Will Save the World
Boston Globe (06/24/07) Karafin, Amy
Martin Demaine, an artist-in-residence at MIT's Computer Science and
Artificial Intelligence Laboratory, believes that puzzles and alternative
ways of solving problems have already made cars safer, candies easier to
unwrap, and may one day help cure diseases. Demaine sees computer science
as a type of riddle. "For me, the excitement is to solve a puzzle that's
never been solved before," Demaine says. "As a researcher, you don't solve
things that have been done; you ask questions, you create puzzles."
Demaine and his colleagues, which includes his son Erik who is also on the
faculty at the MIT lab, have used puzzles and games to make important
contributions to engineering, computer science, and mathematics. An
example is the "fold and cut" problem, which asks if there is a way to fold
a piece of paper so that making a single cut can create a hole of any given
shape. The Demaines created a solution that has since been used in the
manufacturing of car air bags. The MIT researchers are also working on
folding nanostructures, which could be described as DNA origami. Martin
Demaine says that if protein-folding can be understood, it would be
possible to design proteins to cure diseases. The purpose of the Demaines'
research is not to create applications, but to show there are more
efficient ways to build things and that research can be fun. "Puzzles will
save the world," Demaine says.
Click Here to View Full Article
to the top
Robotic Cars Could Take Pressure Off Nation's
Highways
Los Angeles Times (06/23/07) Greenberg, Joel
Stanford University associate professor of computer science and electrical
engineering Sebastian Thrun believes that widespread use of robotic cars
will lead to accident-free, unclogged highways. "There is no other way out
of the current disaster that happens on U.S. highways," Thrun says. "There
are so many aspects of society you could change if you just make cars drive
themselves." Thrun, who leads Stanford Racing Team and their efforts to
create a fully autonomous vehicle for the Defense Advanced Research
Projects Agency's 2007 Urban Challenge, says a robotic automotive vehicle
would "combine the convenience of a train with the convenience of a car."
About 50 teams will compete in the DARPA Urban Challenge, a 60-mile test of
city driving whose purpose DARPA says is to promote "the development of
robotic-vehicle technology" on the battlefield. However, civilian
researchers see limitless applications. "This is the point in time where
cars are really ready to become robotic," says Mike Montemerlo, a senior
research at the Stanford Artificial Intelligence Lab. "We're excited about
the potential this might have for reducing the number of fatalities on the
road." Montemerlo and Thrun believe that eventually high-occupancy vehicle
lanes, and even entire highways, will be filled bumper-to-bumper with
fast-moving robotic cars carrying commuters focused on other tasks. Thrun
says the U.S. inefficiently uses its extensive road system. "What hasn't
been done is to make cars drive closer together in a safe way," Thrun says.
"It is absolutely feasible."
Click Here to View Full Article
to the top
The IT Girls
Austin American-Statesman (TX) (06/23/07) P. F1; Gallaga, Omar L.
Project IT Girl is a three-year program for young women in the Austin,
Texas, area interested in computer science and tech-related fields.
Participants meet every Wednesday and one Saturday every month starting
their sophomore year of high school and continue to meet through the summer
following their senior year. The project, funded by the National Science
Foundation and administered by the nonprofit Girlstart, was developed to
boost the percentage of women in tech-related professions. According to
the National Science Foundation, only 9 percent of engineers, 29 percent of
computer scientists, and 29 percent of computer programmers in the United
States are women. During a two-week summer academy, the 61 young women
participating in the program researched and shot public service
announcements about problems such as AIDS in Africa, racial strife, and
drug abuse. The research, data management, and filming process is intended
to teach participants how to work on large-scale projects that require
teamwork, problem solving, and goal setting, simulating how engineers and
scientists work. Participants who complete the program receive a paid
internship and scholarship from the Texas Advanced Computing Center (TACC)
in North Austin, Texas. TACC associate director Kelly Gaither says she is
not at all surprised by the IT Girls' enthusiasm and skill. "I think
they're capable of tremendous things," Gaither says. "They're getting past
the stereotype that things like this are male-oriented activities."
Click Here to View Full Article
to the top
ICANN to Tackle Transparency
IDG News Service (06/25/07) Perez, Juan Carlos
ICANN's week-long meeting began on Monday to discuss critical topics such
as the organization's efforts to be more transparent and accountable. The
international public meeting was the second of three ICANN has scheduled
for this year and included discussions on significantly expanding available
Internet Protocol addresses and the process of accrediting registrars. The
meeting also includes the first General Assembly of the Latin American and
Caribbean Regional At Large Organization, which was formed in March.
Critics frequently complain that ICANN is not transparent enough in its
decision-making process and needs to be more accountable. ICANN
commissioned an independent study focusing on its transparency and
accountability from London's One World Trust, which reported in March that
ICANN is very transparent, but that it can improve in certain areas,
including providing better explanations of how it uses input from
stakeholders. At the meeting, ICANN will release a set of principles and
frameworks for accountability. ICANN CEO Paul Twomey says ICANN wants to
make it easier for people to find information on its Web site, and be
quicker in posting information online about its meetings and initiatives.
However, analysts say true transparency and accountability will likely
remain open for discussion and criticism until ICANN removes all ties with
the U.S. government. The meeting also will examine the Internationalized
Domain Names initiative, which hopes to redesign the domain name system to
support domain names in a variety of languages and alphabets that cannot be
represented in the ASCII character set. "You'd think this would be
simple," says ICANN Chairman Vint Cerf. "It has turned out to be really
hard, technically."
Click Here to View Full Article
to the top
Five Ideas that Will Reinvent Modern Computing
PC Magazine (06/20/07) Metz, Cade; Bsales, Jamie
A quintet of projects at top high-tech research labs promises to radically
enhance computing. One such project is HP Labs' Pluribus, described by its
creators as "cluster computing for projectors." Pluribus software
integrates multiple projected images into a cineplex-quality image that
boasts a high degree of redundancy, and which could be particularly
applicable to 3D games. Moving closer and closer to the dream of a working
quantum computer is the goal of Bell Labs' "topological quantum computing"
project, which seeks to facilitate the execution of quantum computations
while avoiding decoherence by "forming knots in the space-time path,"
according to researcher Steven Simon. Meanwhile, Microsoft Research is
developing Soap, a wireless optical mouse that allows PC navigation via
hand gestures. "Basically, it's a mouse and a mouse pad in the same
device," explains Soap creator Patrick Baudisch. "But instead of moving
your mouse over your mouse pad, you move your mouse pad over your mouse."
Palo Alto Research Center scientist Van Jacobson intends to shift the
center of the network from the server to the data, removing many of the
inefficiencies of classic point-to-point networking, through the
Content-Centric Networking project. Researchers at IBM's Almaden Research
Center are attempting to make artificial intelligence a reality by
constructing a brain out of hardware and software, rather than following
the more traditional route of mimicking the human cortex. The Cognitive
Computing project's first step is to build a "massively parallel cortical
simulator" that replicates a mouse's brain, and the project leader believes
simulators of increasing complexity will be possible through the dual
advancement of computing power and neuroscience.
Click Here to View Full Article
to the top
Virtual World Sharpens Mind-Control
New Scientist (06/26/07) Knight, Will
A collaborative project involving the Graz University of Technology in
Austria and the University College London (UCL) has developed a simulated
virtual world that can be explored through thoughts and may provide new
rehabilitation possibilities for disabled patients. The Graz University of
Technology specializes in measuring brain signals with electrodes or
implants while UCL works on creating immersive virtual worlds. The two
schools' projects were united by a European consortium called PRESENCCIA.
The system uses electrodes attached to a person's scalp and
electroencephalograms (EEGs) to monitor electrical activity in the brain.
The system can be trained to recognize neuronal activity patterns when the
person is thinking about walking or moving their arms. The thoughts can
then be used to move forward or turn. The user views the virtual world as
video footage projected on a wall, utilizing a pair of shuttered glasses to
create the illusion of a 3D environment. After testing the system on a
test subject, the researchers asked a man paralyzed almost completely from
the neck down to think about walking up to the virtual characters and wait
for each character to say hello. The paralyzed subject was able to
successfully control the system 90 percent of the time. UCL researcher
Doron Friedman says the patient loved the feeling of thinking about walking
and seeing his environment change. Friedman says that virtual reality is
becoming a popular physical and psychological rehabilitation tool, and the
new system could provide novel possibilities. Eventually, such technology
may be used to allow disabled people to operate devices using their
mind.
Click Here to View Full Article
to the top
Flexible and Fearless, Seeking Rescue Work
New York Times (06/25/07) P. A12; Blumenthal, Ralph
Texas A&M University's Texas Engineering Extension Service operates a
52-acre "Disaster City" where fire fighters and other emergency responders
from across the globe can participate in training exercises. The site was
recently the scene of a robotics exercise sponsored by the Department of
Homeland Security's Science and Technology Directorate and the Commerce
Department's National Institute of Standards and Technology. Several
varieties of rescue robots participated in the exercise, which included
obstacle courses based on mock set-ups of the Oklahoma City bombing, 1993
World Trade Center bombing, and Mexico City earthquake. The robots
included a 30-foot, snake-like optic robot that slinks through crevasses
and holes while providing images of its discoveries. That robot, produced
by university researchers in Japan, is attached to the operator's body,
unlike most robots, which are operated via consoles or laptops. One Texas
A&M official predicted that robots will soon become a regular part of
rescue work.
Click Here to View Full Article
to the top
Can the Internet Be Saved?
Chronicle of Higher Education (06/29/07) Vol. 53, No. 43, P. A25;
Fischman, Josh
The Internet is bowing under constant pressure from spam, malware, mobile
devices, a lack of security, and spotty connections, and the National
Science Foundation officially launched the Global Environment for Network
Innovations (GENI) project to reinvent the Net in May. The biggest problem
with the Internet is security, and at the root of this problem is the lack
of authenticated identity and the erroneous assumption that every network
insider is to be trusted; Princeton professor Larry L. Peterson thinks one
solution is to construct a network that can contain attacks launched by
end-users' machines. To effect more reliable data transmission,
researchers are investigating the potential of using more programmable
routers that communicate with each other, facilitating a more global
perspective and allowing operators to split the network into virtual
"slices" so management is easier. As for the problem of increasing numbers
of mobile devices and the strain this places on the network, researchers
are experimenting with ad-hoc networking on facilities that include
Internet2 and the National LambdaRail network. The GENI project office is
run by BBN Technologies engineer Chip Elliott, who thinks the effort calls
for two strategies. "First, if you don't like conventional Internet
protocols, try something completely different," he explains. "Second, do
it on a large enough scale, with enough users, so that your results
actually mean something." Among those involved in the GENI planning
process are researchers from MIT, UC Berkeley, USC, and Princeton
University, as well as Intel and other industry players.
Click Here to View Full Article
to the top