Welcome to the July 18, 2014 edition of ACM TechNews, providing timely information for IT professionals three times a week.
Updated versions of the ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).
HEADLINES AT A GLANCE
MIT Scientists Develop Sensor-Operated Robotic Fingers
BBC News (07/17/14)
Massachusetts Institute of Technology researchers have developed a robotic extension to the human hand that could make common, everyday tasks easier. The extension, which is essentially two extra fingers, could be used to grasp, leaving the hand free to do other tasks. The extra fingers use sensors attached to the human hand to measure the position of the wearer's fingers, and an algorithm controls the output from the sensors to the robotic fingers. "If you have control and can communicate with them very well, you feel that they are just an extension of your body," says MIT professor Harry Asada. The robotic fingers are on either side of the hand, located on the outside of the thumb and pinky finger. "This is a completely intuitive and natural way to move your robotic fingers," Asada says. "You do not need to command the robot, but simply move your fingers naturally. Then the robotic fingers react and assist your fingers." He notes the device is still in the prototype stage, and the researchers hope to make it much smaller and foldable. "We could make this into a watch or a bracelet where the fingers pop up, and when the job is done, they come back into the watch," Asada says.
Some Universities Crack Code in Drawing Women to Computer Science
The New York Times (07/17/14) Claire Cain Miller
Just 18 percent of the computer science graduates from colleges and universities across the U.S. are women, down from 37 percent in 1985, but some schools are working to change that. Carnegie Mellon University (CMU), the University of Washington (UW), and Harvey Mudd College have pursued several different programs and policies that have helped them dramatically increase the proportion of women in their computer science programs. Both UW and CMU have partnered with companies and organizations to train high school computer science teachers and host camps and mentoring programs that help involve girls in computer science at an early age. Harvey Mudd has sought to make women's participation in the field more visible in marketing materials and campus tours, and CMU has established formal programs to support women in the computer science program, while also eliminating programming experience as part of the criteria for admission. Altering curriculum is a controversial choice, but UW says revamping its introductory course to include a focus on the creative and real-world applications of computer science has helped bring many women into the program who would not otherwise have studied it further. CMU decided not to change its curriculum, contending the belief that women place more importance on creativity and real-world applications is a myth.
Internet Society to Measure and Display Quality of Connections With Netradar App
Aalto University (07/15/14)
Researchers at the Internet Society and Aalto University have developed Netradar, a free service that provides neutral and accurate information, collected by end users, about the quality and diversity of Internet connections and mobile devices. The results are presented on maps of network connectivity and performance. "One of the tremendous benefits of Internet technology is the ability to connect diverse networks together," says the Internet Society's Phil Roberts. "With Netradar, we hope to expose and share some of that diversity and also to study over time how the Internet is changing." The Netradar app is available for most smartphone platforms. When users run the app they are presented data about the quality of their connectivity, which is concurrently uploaded to the Netradar database. The system does not collect any personal information, and it is unnecessary to have an account or login to use the system. The Netradar client has been installed more than 120,000 times and the database currently holds more than 3 million measurements. "This collaboration...will help researchers and Internet users study the evolution of the mobile Internet and form an even more accurate model of network quality worldwide," says Aalto University professor Jukka Manner.
No-Wait Data Centers
MIT News (07/17/14) Larry Hardesty
Massachusetts Institute of Technology (MIT) researchers have developed Fastpass, a network-management system that could reduce the average queue length of routers in a Facebook datacenter by 99.6 percent. Fastpass relies on a central server called an "arbiter" to decide which nodes in the network can send data to which others during a certain period of time. A node that wants to transmit data issues a request to the arbiter and receives a routing assignment. The researchers found that an arbiter with eight cores can keep up with a network transmitting 2.2 terabits of data per second. Fastpass relies on a method for dividing the task of assigning transmission times so it can be performed in parallel on separate cores. The system assigns each core its own time slot, and the core with the first slot scrolls through the complete list of pending transmission requests. Every time it comes across a pair of servers that have not received an assignment, it schedules them for its slot. "This paper is not intended to show that you can build this in the world's largest datacenters today," says MIT professor Hari Balakrishnan. "But the question as to whether a more scalable centralized system can be built, we think the answer is yes." The researchers will present their paper at the ACM Special Interest Group on Data Communication's annual conference in August.
Electric Vehicles Drive IoT Research at Four Universities
Government Computer News (07/15/14)
Internet2 announced that it will provide four U.S. universities with an all-electric micro vehicle, in cooperation with vehicle manufacturer UEV LLC, which will be used in a variety of research projects in Internet and network connectivity and reduce a university's carbon footprint. The university electric vehicles (UEVs) will feature Internet2's InCommon federated identity management solution for educational and research institutions, as well as its eduroam wireless networking technology. The technology will communicate operational data, including position, speed, and battery charge, to the schools' networks where it will be available to researchers. Users will be able to use their campus identification and login credentials and a mobile application to check if the UEV is available, reserve or activate it, and provide comments on the UEV's performance and utility. Internet2's Steve Wolff says the organization hopes the UEVs will help to drive broader research into the widespread adoption of the Internet of Things (IoT). The universities selected for the program are Colorado State University, the University of Pittsburgh, the University of Washington, and the University of Wisconsin. "We are excited about these four campuses that are regional leaders in sustainability using these electric vehicles to enhance research, adopt advanced technologies, and enhance campus carbon footprint reductions," Wolff says.
New Computing Cluster at Carnegie Mellon Traces Lineage to Los Alamos' Roadrunner Supercomputer
Carnegie Mellon News (PA) (07/14/14) Byron Spice
Carnegie Mellon University (CMU) researchers are using components of a decommissioned Los Alamos National Laboratory supercomputer to create Narwhal, a new computing cluster to be used in education and research related to large-scale computer systems. Narwhal is being built from 448 blade computers salvaged from the Cerrillos and Roadrunner high-performance computing systems, which were both taken offline last year. CMU professor Garth Gibson says the new computing cluster will provide a unique and powerful resource for students and researchers. "With Narwhal, we open a new front--assistance with large-scale computer systems software education," Gibson says. "Roadrunner and Cerrillos may be retired, but even a sliver of these machines' core capabilities is more capable than most educational computing resources." Researchers and students will use Narwhal to conduct experiments with parallel computing applications and infrastructure, controlling and instrumenting all software at a scale that is an order of magnitude larger than most university clusters. Roadrunner boasted a hybrid design in which general-purpose AMD Opteron processors were connected to Cell graphics processors from IBM. The AMD processors managed basic tasks while the IBM Cell processors executed computationally intense calculations, contributing to Roadrunner's overall speed. Narwhal will have 1,792 processor cores versus Roadrunner's 122,400 cores.
Math Can Make the Internet 5-10 Times Faster
Aalborg University (07/17/2014)
Researchers at Aalborg University, the Massachusetts Institute of Technology, and the California Institute of Technology say they have developed software that can make Internet communications via computer, mobile phone, or satellite significantly faster and more secure. The researchers used their new method to download a four-minute video five times faster than with existing state-of-the-art technology. "In experiments with our network coding of Internet traffic, equipment manufacturers experienced speeds that are five to 10 times faster than usual," says Aalborg University professor Frank Fitzek. "And this technology can be used in satellite communication, mobile communication, and regular Internet communication from computers." The new method involves network coding that uses specialized mathematics to store and send the signal in a new way that is different from the traditional packets. The advantage is that errors along the way do not require a packet to be resent. With the new method, the upstream and downstream data is used to reconstruct what is missing using a mathematical equation. "I think the technology will be integrated in most products because it has some crucial and necessary functions," Fitzek says.
A Robot With a Little Humanity
The New York Times (07/16/14) John Markoff
Massachusetts Institute of Technology professor and roboticist Cynthia Breazeal is known for her work in "social" robotics, including her creation of Kismet, a robotic head that used sensors and actuators to mimic human emotional states while interacting with humans. Breazeal has now turned her efforts toward the development of a more abstract social robot: the one-foot-tall, six-pound Jibo, which, instead of a robotic face, features a movable screen displaying an expressive orb. Jibo is meant to act as something of a family companion in a household context, able to send messages, take pictures, act as a personal messenger, and facilitate conversations between people in different locations. Jibo also will have its own personality, which will enable owners, Breazeal hopes, to view it as a friend. "It's really important for technology to be humanized," Breazeal says. "The next stage in computing, the next wave, is emotion." Breazeal, and the company designing Jibo, hope developers will use the robot as a platform to create new applications expanding its functionality to act as a tutor or coach, among other applications. Jibo will not be commercially available until the 2015 holiday season, but developer units will be available next fall.
Neural Networks That Function Like the Human Visual Cortex May Help Realize Faster, More Reliable Pattern Recognition
A*STAR Research (07/16/14)
A spiking neural network (SNN) employing suitable learning algorithms could potentially rival the speed and accuracy of the human visual cortex, according to researchers at the A*STAR Institute for Infocomm Research. SNNs are similar to the human visual cortex in that they encode visual information in the form of spikes by firing electrical pulses down their neurons. The A*STAR team's SNN has a feedforward architecture and consists of encoding, learning, and readout neurons. The learning neurons are capable of discriminating patterns in an unsupervised manner, but the researchers accelerated the process by incorporating supervised learning algorithms in the computation so the learning neurons could respond to changes faster. In testing involving images from the Mixed National Institute of Standards and Technology database, the SNN could recognize handwritten numbers ranging from 0 to 9 with a performance comparable to that of support-vector machines, the current benchmark for pattern-recognition methods. The precision of the SNN was about 94 percent for training images and about 79 percent for testing images. The A*STAR researchers say such an artificial neural network could have broad applications in biometrics, data mining, and image analysis.
UNSW Gets Robots Ready for Robo Soccer World Cup
Computerworld Australia (07/16/14) Hamish Barwick
The Brazilian city of Joao Pessoa will host a soccer tournament for robots in late July, and among those participating in the RoboCup Championships is a team from Australia's University of New South Wales (UNSW). The UNSW team, rUNSWift, will compete in the Standard Platform League, in which each team uses the same robots, according to UNSW's Maurice Pagnucco. He says all of the teams look to introduce some sort of innovation and rUNSWift has developed a sidestep for walking around opposing robots. "These robots have a hip joint and we have exploited the way this robot is constructed so that it can move as quickly as possible and introduce a few novel elements such as this sidestep," Pagnucco says. "[The robots'] walking is one of the most important things because the quicker they can walk to the ball, the better off we are." The robots are incapable of running, and Pagnucco says the machines' design dictates they have at least one foot on the ground. He also notes each autonomous robot has 125,000 lines of code that need to be maintained. The UNSW team also will employ a robot coach this year.
Microsoft Urges People to Use Weak Passwords Online
International Business Times (07/16/14) Anthony Cuthbertson
Microsoft and Carleton University researchers suggest in a new report that Internet users do not need to use complex passwords for all of their online accounts, contradicting established best practices for password security. Their study found password management strategies "ruling out weak passwords or password re-use is sub-optimal." The researchers note it is increasingly difficult for users to employ the best practice of using different, long, and complex passwords for every online account, going so far as to call it a "human impossibility." The researchers say common coping mechanisms such as writing passwords down, single sign-on services, email-based password resets, and password managers are all acceptable solutions. They also argue a better solution is for Internet users to separate their online accounts into high-value and low-value accounts, the former including email and banking accounts and the latter including less sensitive accounts such as those for chat forums. The researchers say high-value accounts should continue to be protected with complex passwords, but accounts in the second tier should use simple and commonly re-used passwords. "We note that while password re-use must be part of an optimum portfolio strategy, it is no panacea," the report says.
Fundamental Chemistry Findings Could Help Extend Moore's Law
Lawrence Berkeley National Laboratory (07/15/14) Kate Greene
Lawrence Berkeley National Laboratory researchers are trying to extend Moore's Law by developing a new type of resist that combines the material properties of two pre-existing kinds of resist, achieving the characteristics needed to make smaller features for microprocessors, including better light sensitivity and mechanical stability. "We discovered that mixing chemical groups, including cross linkers and a particular type of ester, could improve the resist's performance," says Berkeley Lab researcher Paul Ashby. The resist protects the material that makes up transistors and wires from being etched away and can enable the material to be selectively deposited. "The semiconductor industry wants to go to smaller and smaller features," Ashby notes. Extreme ultraviolet (EUV) light technology holds promise, but "you also need the resist materials that can pattern to the resolution that extreme ultraviolet can promise," Ashby says. The first kind of resist the researchers investigated is called crosslinking, which consists of molecules that form bonds when exposed to EUV light. The second type of resist is highly sensitive, but still lacks mechanical stability. The next step for the researchers is to further optimize the resist's chemical formula for the very small components required for future microprocessors.
Microsoft Challenge's Google's Artificial Brain with 'Project Adam'
Wired News (07/14/14) Daniela Hernandez
Google has been a leader in the field of deep-learning neural networks. The company's Google Brain system carries out the artificial intelligence (AI) calculations that underpin many of Google's services, such as voice recognition and Google Maps. It also is one of the few AI systems that has been able to handle a visual recognition benchmark test called ImageNet 22K, but Microsoft has equaled that feat with its new AI system, "Project Adam," which seeks to challenge Brain. Like Brain, Adam runs on an array of computer servers, in this case Microsoft's Azure cloud computing service. Like Brain, it uses deep-learning algorithms to mimic the neural structure of the brain, but it also makes use of a technique known as asynchrony, which is a method of splitting a system into parts that can largely run independently of each other and then come together to share their calculations. Asynchrony has been shown to work well on single devices, but tends to breakdown in large systems of servers. Adam builds on the HOGWILD! asynchronous technology developed by the University of Wisconsin that enables each processor in a machine to work more independently. Microsoft, which officially unveiled Adam earlier this week, says this technique enables neural networks to more quickly and accurately train themselves to understand inputs such as images.
Abstract News © Copyright 2014 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.
Non-Members: Unsubscribe
|
