Welcome to the March 11, 2016 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
Beyond Silicon: The Search for New Semiconductors
The Conversation (03/10/16) Thomas Vandervelde
Researchers are looking at new semiconductor materials as an alternative to silicon, which is reaching its technical limits, writes Tufts University professor Thomas Vandervelde. He notes a key limitation is silicon's inefficiency in converting light into an electrical signal, and vice versa. Vandervelde says although this weakness was not very important in computer chips linked by metal wires, it has become a barrier to the migration to semiconductor use in light-related applications. He says finding materials that work well with light requires exploring a combination of semiconducting elements from the group III and group V columns on the periodic table, such as gallium arsenide. However, III-V elements are much rarer in comparison to silicon, as well as being far more expensive, and more brittle. Vandervelde notes III-VI elements such as combinations of zinc, cadmium, and mercury with tellurium are in use as semiconductors, but they also suffer from brittleness and are difficult to manufacture. In terms of future uses for new materials, Vandervelde envisions the electrical grid using gallium-nitride semiconductor electronics to convert power for high-voltage transmissions and back again. He also thinks silicon's use will be revitalized with the creation of more light-friendly silicon photonics.
Google's AlphaGo AI Beats Star South Korean Go Player Again
IDG News Service (03/10/16) John Ribeiro
The AlphaGo algorithm from Google DeepMind has won the second consecutive game of Go against human champion Lee Se-dol in a five-game tournament in South Korea. Lee told reporters AlphaGo led the game and made flawless moves in Thursday's round, whereas in the first game on Wednesday some of its maneuvers were problematic. The program only needs to beat Lee a third time to be crowned the tournament's victor, and its performance excites analysts and artificial intelligence (AI) researchers. According to Google DeepMind, AlphaGo seeks to maximize the likelihood of winning instead of optimizing margins, as humans often do. Event commentator Michael Redmond notes AlphaGo beat another champion Go player in an October match, but committed critical errors that made it look very much like an artificial program or computer. Consultant Patrick Moorhead sees Google's AlphaGo victory as helping to raise mass-market awareness of AI's cognitive computing capability, "portending to an increased AI offering to businesses."
Machine-Learning Algorithm Aims to Identify Terrorists Using the V Signs They Make
Technology Review (03/08/16)
Ahmad Hassanat from Mu'tah University in Jordan and colleagues have developed a way to distinguish people from the unique way they make the "victory" sign. The team photographed the V sign made by 50 men and women of various ages with their right hand several times against a black background using an 8-megapixel camera phone, and produced a database of 500 images. Hassanat and colleagues limited their analysis to determining the end points of the two fingers, the lowest point in the valley in between them, and two points in the palm of the hand. They also analyzed the shape of the hand using several statistical measures, and the two approaches created 16 features that can be used in identification. The team then used 66 percent of the images to train a machine-learning algorithm to recognize different V signs and used the remaining images to test its efficacy. Hassanat says the algorithm was able to distinguish people with an accuracy of more than 90 percent in some cases. He thinks that combined with other data, the approach could be used to identify mask-wearing terrorists making the victory sign.
The U.S. Government Launches a $100-Million 'Apollo Project of the Brain'
Scientific American (03/08/16) Jordana Cepelewicz
The U.S. Intelligence Advanced Research Projects Activity will invest $100 million in the Machine Intelligence from Cortical Networks (MICrONS) program, a project to reverse-engineer a section of the brain, study its computational mechanisms, and feed those insights into the improvement of machine-learning and artificial-intelligence algorithms. The goal is to enhance artificial neural networks so they get better at pattern recognition in cluttered environments and generalization. Three teams will employ distinctive methods to map out the neural pathways in a cubic millimeter of a rat's cortex as it engages in visual perception and learning tasks, and then determine how to usefully apply the information to machine-learning algorithms. Their resulting theories will be distilled into internal models they will test against the reverse-engineered brain data. MICrONS intends to use these models to make machines more automatic, especially in terms of simplifying and accelerating object recognition. MICrONS program manager Jacob Vogelstein believes extracting information from the brain at the computational level can bring the algorithms closer to brain-like performance. "We hope to achieve...better generalization, better capacity for abstraction, better use of sparse data," he says. The project's challenges include dealing with the massive dataset generated by brain measurements, and mining a vast array of images with segmentation using more refined computer-vision techniques.
Algorithm Reads Tweets to Figure Out Which Restaurants Make You Sick
IDG News Service (03/08/16) Katherine Noyes
University of Rochester researchers have developed nEmesis, an app that uses machine learning to help minimize the number of people affected by food poisoning. The software uses natural-language processing and artificial intelligence to identify food poisoning-related tweets, connect them to restaurants using geotagging, and identify likely hotspots. The researchers developed the app by analyzing nearly 4 million tweets generated by people in the New York City metropolitan area in late 2012 and early 2013. They then tested the app in Las Vegas through collaboration with the city's health department. For three months, nEmesis automatically scanned an average of 16,000 tweets from 3,600 users a day, and the researchers used the tweets to generate a list of the highest-priority restaurants for inspections. "Each morning we gave the city a list of places where we knew that something was wrong so they could do an inspection of those restaurants," says former University of Rochester researcher Adam Sadilek. The researchers estimate the app resulted in 9,000 fewer food-poisoning incidents and 557 fewer hospitalizations in Las Vegas during the course of the study.
Researchers Spoof Phone's Fingerprint Readers Using Inkjet Printers
eWeek (03/09/16) Todd R. Weiss
Michigan State University (MSU) researchers used off-the-shelf inkjet printers to demonstrate how fingerprint readers on popular smartphones can be manipulated into unlocking the devices using spoofed fingerprints made with printer inks. MSU's Kai Cao and Anil K. Jain sought to investigate the overlooked spoofing strategy, which is especially relevant because half of smartphones sold by 2019 are expected to have an embedded fingerprint sensor. "With the introduction of Apple Pay, Samsung Pay, and Android Pay, fingerprint recognition on mobile devices is leveraged for more than just device unlock; it can also be used for secure mobile payment and other transactions," the researchers note. Cao and Jain used an inkjet printer loaded with three silver conductive ink cartridges and a normal black ink cartridge, and scanned a fingerprint of a phone's authorized user at 300 dpi (dots per inch) or higher resolution. Afterward, the print was reversed or mirrored before being printed onto the glossy side of a piece of AgIC paper. "Once the printed [two-dimensional] fingerprints are ready, we can then use them for spoofing mobile phones," the researchers note. The spoofed print successfully unlocked Samsung Galaxy S6 and Huawei Honor 7 smartphones. Cao and Jain say their experiment "further confirms the urgent need for anti-spoofing techniques for fingerprint-recognition systems, especially for mobile devices."
System Loads Web Pages 34 Percent Faster by Fetching Files More Effectively
MIT News (03/09/16) Adam Conner-Simons
Researchers from the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL) and Harvard University have collaborated on Polaris, a system that speeds up Web page-load times 34 percent. Polaris determines how to overlap retrieving a page's objects to reduce page load time. The researchers will present a paper on their work at next week's USENIX Symposium on Networked Systems Design and Implementation (NSDI '16) in Santa Clara, CA. Polaris automatically tracks all interactions between objects, which can run into the thousands for a single page, then produces a "dependency graph" for the page. While other tech firms trying to expedite load times often focus on accelerating information transfer via data compression, the CSAIL team showed Polaris benefits from more consistent and substantial load-time improvements. "Recent work has shown that slow load times are more strongly related to network delays than available bandwidth," says MIT professor Hari Balakrishnan. "Rather than decreasing the number of transferred bytes, we think that reducing the effect of network delays will lead to the most significant speedups." Polaris is suitable for larger, more complex websites, which dovetails with the swelling of Web pages to contain thousands of objects. The system also works well with mobile networks, which usually have longer delays than wired networks.
Google's Robots Are Learning How to Pick Things Up
Popular Science (03/08/16) Dave Gershgorn
Google is training robots to pick up objects in one bin and place them in another, using continuous feedback to learn how to adaptively grasp objects of different shapes, sizes, and characteristics. The robots are constantly analyzing the object and their hand's relation to it, with two deep neural networks serving as the control mechanism. One network studies photos of the bin captured by a camera and predicts whether the robot's hand can correctly grasp the object, while the second network interprets how well the hand is grabbing so it can alert the first network to make adjustments. Google researchers note the robots do not have to be calibrated based on different camera placement; provided the camera has a clear view of the bin and arm, the neural network can adapt and continue learning to pick up objects. Google spent two months having the machines pick up objects 800,000 times. A surprising finding of the experiment was the robots learned different methods for picking up hard and soft objects. For hard objects, the grippers would only grasp their outer edges and squeeze, while for soft objects the robots would place one gripping finger in the middle and one around the edge, and then squeeze.
With Boxmate Malicious Programs Have No Place Left to Hide
Saarland University (03/09/16)
Saarland University professor Andreas Zeller and two graduate students have developed an approach to address a core problem in existing security systems. "The attack needs to have been observed at least once to be able to recognize it the next time--and then, you have to update everything again and again," Zeller notes. His Boxmate strategy systematically produces inputs to probe the program's normal behavior. "During this automatic testing, we log which critical data, and which critical resources, the program is accessing to perform these tasks, and the test generator ensures that all visible features actually are exercised," Zeller says. In the production stage, the program is put into a "sandbox" that monitors its operation and raises an alert whenever some data is being accessed that was not accessed in testing, and also catches and inhibits attacks if the program is compromised or acts in a previously unseen malicious manner. "Malicious programs no longer have a place to hide," Zeller says. He notes if a program wants to use certain kinds of data later, it will have to access it while undergoing testing by Boxmate, thus exposing its activity. "Any hidden functionality will be disabled by the sandbox, and this will make it hard for attackers," Zeller says.
Nations Ranked on Their Vulnerability to Cyberattacks
University of Maryland (03/09/16) Matthew Wright
Researchers at the University of Maryland (U-M) and the Virginia Polytechnic Institute and State University have co-authored a book ranking the vulnerability of 44 nations to cyberattacks. The U.S. was ranked 11th safest, while Scandinavian countries such as Denmark, Norway, and Finland were ranked the safest. China, India, Russia, Saudi Arabia, and South Korea ranked among the most vulnerable. "Our goal was to characterize how vulnerable different countries were, identify their current cybersecurity policies, and determine how those policies might need to change in response to this new information," says U-M professor V.S. Subrahamian, who led the research. The book, "The Global Cyber-Vulnerability Report," was based on a two-year study that analyzed more than 20 billion automatically generated reports, collected from 4 million machines each year worldwide. The rankings were partly based on the number of machines attacked in a given country and the number of times each machine was attacked. Trojans, followed by viruses and worms, posed the principal threats to machines in the U.S., but misleading software is much more prevalent in the U.S. compared with other nations that have similar gross domestic product, suggesting U.S. efforts to reduce cyberthreats should focus on education to recognize and avoid misleading software.
Hooray for Hollywood Robots: Movie Machines May Boost Robot Acceptance
Penn State News (03/09/16) Matt Swayne
Pennsylvania State University (PSU) researchers recently conducted a study in which 379 older adults who recalled more robots portrayed in films had lower anxiety toward robots than seniors who remembered fewer robot portrayals. Remembering robots from how they are portrayed in films may help ease some of the anxiety older adults have about using a robot, according to the researchers. Finding ways to ease anxiety about robot adoption also could help senior citizens accept robots as caregivers, the researchers add. "Robots could provide everything from simple reminders--when to take pills, for example--to fetching water and food for people with limited mobility," says PSU professor S. Shyam Sundar. In addition, the researchers found the trusting effect held up even when older adults recalled robots that were not friendly human-like helper robots. "So, it seems like the more media portrayals they can recall, the more likely their attitudes would be positive toward robots, rather than negative," says PSU researcher T. Franklin Waddell. The research also found people had a more positive reaction to robots that looked more human-like and those that evoked more sympathy. The researchers suggest robot designers incorporate features that remind older adults of robots in the media, and create robots with more human-like interfaces and models with features that increase sympathy.
21 Hot Programming Trends--and 21 Going Cold
InfoWorld (03/10/16) Peter Wayner
Certain programming trends have become more popular among modern coders, while others have cooled. Among popular trends is the advent of preprocessing programs that translate new code into older code with a set of libraries and application programming interfaces, while full language stacks are less desirable. Docker containers also are gaining popularity over hypervisors, because their usage and deployment are easier. Meanwhile, artificial intelligence and machine learning increasingly are favored over the term "big data," and big data solutions are less in demand among companies compared to basic solutions that can handle their large, but by no means vast, datasets. In addition, mobile Web apps are preferred over native mobile apps thanks to a faster HTML layer. Graphics processing units (GPUs) are gaining over central processing units, with computer scientists converting parallel apps to run many times faster on GPUs. There is now less vendor emphasis on cloud computing's simplicity than on its complexity, and infrastructure-as-a-service is overtaking platform-as-a-service, partly because coders do not like the latter's vendor lock-in. Moreover, today's programmers are becoming less willing to invest in four years of computing education when a more affordable and flexible option to learn new and relevant skills on an as-needed basis is available with online coursework.
Report: MOOC Instructors Need More Support
Campus Technology (03/04/16) Leila Meyer
A Pennsylvania State University (PSU) and University of Central Florida study of 14 current and former instructors of massive open online courses (MOOCs) found only four were interested in teaching such courses on a regular basis. Of the remaining 10 instructors, two said they did not want to teach another MOOC, four wanted to take a break, and another four expressed concern about the demands of teaching another course. MOOCs require instructors to write lessons and record video lectures, in addition to their regular teaching duties. One instructor said it took 400 hours of work to prepare a single course, while some instructors had to adjust from providing one-on-one guidance to teaching thousands of students. They also struggled to measure the success of the course, considering 90 percent of students leave MOOCs after two weeks. The researchers hope the study will help show MOOC instructors need more support. "By improving support for the instructors and their collaborators, we may also improve the MOOC experience for students and other stakeholders," says PSU doctoral candidate Saijing Zheng.
Abstract News © Copyright 2016 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: firstname.lastname@example.org
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.