Association for Computing Machinery
Welcome to the August 31, 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).


Who Would Win the Coding Olympics?
The Washington Post (08/30/16) Karen Turner

U.S. programmers landed in 28th place among their international peers in a HackerRank compilation of the results of 1.4 million coding challenges by approximately 300,000 developers. China topped the list of the most accomplished coders, followed by Russia, Poland, Switzerland, and Hungary. The ranking found China's top coding category was algorithms, while Russia's was data structures. A key factor in these nations' coding success is likely the introduction of math and computer education at a much earlier age than occurs in the U.S., says HackerRank CEO Vivek Ravisankar. "In my opinion, the U.S.'s position here mirrors a lot of the other world ranking reports, such as STEM (science, technology, engineering, and math) education performance, or even other international coding competitions," he says. Last year's Pew Research Center analysis of STEM test scores found U.S. students were middle-of-the-pack underperformers compared to those in Singapore and South Korea. Moreover, this year's International Olympiad in Informatics was led by Chinese, Russian, and Eastern European contestants, while the highest-scoring U.S. coder came in 15th place. The Chinese and Russians also scored victories at the ACM International Collegiate Programming Contest, and at Google Code Jam.
View Full Article - May Require Free Registration | Return to Headlines | Share Facebook  LinkedIn  Twitter 

South Africa's New Supercomputer Delivers World-Class Competitiveness
Scientific Computing (08/29/16) Randall Cronk

South Africa's Center for High-Performance Computing (CHPC) in Cape Town supports a supercomputer named Lengau, designed to make the country a global player in the international HPC research community. Lengau, which means "cheetah" in Setswana, became fully operational in May and was ranked 121 on the June Top500 list of the world's fastest supercomputers. A total of 1,013 servers support Lengau's 5-petabyte storage capacity, and the machine boasts more than 24,000 cores and a maximum interconnect speed of 56 Gbps to offer about a petaflop of computing speed. CHPC director Happy Sithole says the research areas the center covers include "chemistry, bioinformatics, astronomy, computational mechanics, engineering applications or systems, and the earth sciences including climate change." Sithole also says South Africa's HPC ambitions extend beyond scientific advancement to include industry enhancement in domains such as virtual prototyping. He notes CHPC also is the only HPC research facility in the African continent, which makes its position unique in terms of partnership opportunities. "I think high-performance computing is vital for competitiveness in developed countries," Sithole says. "In South Africa we also have that ambition to accelerate areas where we are competitive in industry and science."

Intelligent Technology--the Evolution and Future of Automation (08/31/16)

Researchers at the Chinese Academy of Sciences in Beijing, China, argue core principles of automation and artificial intelligence (AI) must be reconsidered as the world experiences an information technology (IT) shift. The researchers' comments follow AlphaGo's victory against an internationally ranked Go player last year. "AlphaGo is not only a milestone in the quest for AI, but also an indication that IT now has entered a new era," says Chinese Academy of Sciences professor Fei-Yue Wang. Wang describes the progress of robotic and neural machine-human interaction in a timeline of five "control" eras. Wang says automation evolved from the pure mechanics of ancient water clocks and steam engines to the development of electric circuits and transfer functions, which in turn gave way to power grids. Digital computers and microprocessors represent the third shift and paved the way for the Internet and the World Wide Web, Wang notes. "In Control 5.0...only association revealed by data or experience is available, and causality is a luxury that is no longer attainable with limited resources for uncertainty, diversity, and complexity," Wang says. He predicts recognizing these three worlds and the dual learning roles of each will be essential in the fifth era of intelligent technology.

Researchers Discover Machines Can Learn by Simply Observing, Without Being Told What to Look For
University of Sheffield (08/30/16) Kirsty Bowen

Researchers at Britain's University of Sheffield have discovered machines can learn how natural or artificial systems function via simple observation. "Our study uses the Turing test to reveal how a given system--not necessarily a human--works," says Sheffield's Roderich Gross. He says the team monitored a robot swarm to determine the rules governing their movements, using a second swarm to observe them; the movements of all the robots were recorded, and the motion data given to interrogators. "Unlike the original Turing test, however, our interrogators are not human but rather computer programs that learn by themselves," Gross notes. He says their task was to differentiate robots from either swarm, and they were rewarded for correctly classifying the motion data from the original swarm as authentic, and those from the other swarm as inauthentic. "The learning robots that succeed in fooling an interrogator--making it believe their motion data were genuine--receive a reward," Gross says. He notes this approach makes it unnecessary to have humans tell the machines what behaviors to seek to identify. Gross thinks these insights could lead to innovations in how machines infer knowledge and use it to predict behavior and abnormal divergences from behavior.

Here's How Russian Hackers Could Actually Tip an American Election
The Washington Post (08/30/16) Craig Timberg; Andrea Peterson

Voting technology experts such as Princeton University researcher Andrew Appel have long warned of endemic insecurity in U.S. election systems, including hackers' ability to delete or alter data on voter rolls to manipulate electoral outcomes. "There are computers used in all points of the election process, and they can all be hacked," Appel says. "So we should work at all points in that system to see how we make them trustworthy even if they do get hacked." Rice University professor Dan S. Wallach points to the destruction of voter registration databases as another way to sow chaos, and this threat has sparked considerable concern with recent allegations of Russian intrusions into several states' voting systems. Experts agree the most secure voting systems combine optical scanners and paper ballots voters fill out manually, creating both a paper trail and an electronic count. Systems that only tally digital records offer numerous hacker targets, including polling machines, counting stations, and the computers that compile and count jurisdictional results. The use of systems that produce both print and digital records is trending nationally, with Verified Voting Foundation estimating 35 states and many counties elsewhere have deployed such systems.
View Full Article - May Require Free Registration | Return to Headlines | Share Facebook  LinkedIn  Twitter 

One in Two Users Click on Links From Unknown Senders
Friedrich-Alexander University Erlangen-Nurnberg (Germany) (08/25/16)

Researchers at Friedrich-Alexander University Erlangen-Nurnberg (FAU) in Bavaria, Germany, investigated why many users click on links from unknown senders even through they know these types of links can be dangerous. They found most Facebook users clicked on a link from an unknown sender mainly out of curiosity. The researchers sent about 1,700 FAU students��� emails or Facebook messages under a false name, and they adapted the messages to the target groups by signing them with one of the 10 most common names for the group's generation. If the recipient clicked on the message's link they were directed to a page with the message "access denied," which enabled the researchers to register the click rates. In one study the researchers addressed the test subjects by their first names, while in a second study they did not address the subjects personally but gave more specific information about the occasion on which the photos were supposedly taken. In the first study 56 percent of the email recipients and 38 percent of the Facebook message recipients clicked on the links. In the second study the percentage of email recipients who clicked on the link dropped to 20 percent, and the percentage of Facebook users who did so rose to 42 percent.

What Robots Can Learn From Babies
Technology Review (08/30/16) Will Knight

New software can predict how objects captured by a computing device's camera will most likely behave. Researchers at the Allen Institute for Artificial Intelligence (Ai2) in Seattle developed the system, which combines machine learning and three-dimensional (3D) modeling to draw conclusions about the physical properties of a scene. Ai2's Roozbeh Mottaghi and colleagues converted more than 10,000 images into scenes rendered in a simplified format using a 3D physics engine, and fed the images and 3D representations into a computer running a deep-learning neural network. Mottaghi says the computer gradually learned to associate a particular scene with certain simple forces and motions. When shown unfamiliar images, the system could suggest the various forces that might be in play. Mottaghi notes the system does not work perfectly, but more often than not it will draw a sensible conclusion. For example, for an image of a stapler sitting on a desk, the system can determine that if it is pushed across the desk it will fall abruptly to the floor. Mottaghi says the system could potentially help make robots and other machines less prone to error.

Rise of the Robo-Journalists? Google Teaches an AI the Art of Writing a Good Headline
ZDNet (08/26/16) Liam Tung

The Google Brain Team says they have made strides in teaching computers to summarize text, and have developed a machine-learning algorithm that can write "very good" headlines. "We've observed that due to the nature of news headlines, the model can generate good headlines from reading just a few sentences from the beginning of the article," says the Google Brain Team's Peter Liu. He notes the team used Google's TensorFlow software library as a basis for the model, which has generated several passable headlines using extracts from articles. For example, the algorithm converted "Australian wine exports hit a record 52.1 million liters worth 260 million dollars in September, the government statistics office reported on Monday" into "Australian wine exports hit record high in September." Liu says the model was trained on data from John Hopkins University's massive Annotated English Gigaword dataset, and Google used a deep-learning model called sequence-to-sequence learning to mimic humans' text summarization. Liu acknowledges the technique is less effective when producing a summary requires reading an entire document. However, he says the method can function as a baseline, and Google has made the models publicly accessible on GitHub.

3D Printed Structures 'Remember' Their Shapes
MIT News (08/26/16) Jennifer Chu

A team from the Singapore University of Technology and Design (SUTD) and the Massachusetts Institute of Technology (MIT) has printed shape-memory structures with finer details than those produced with conventional three-dimensional (3D) printers. MIT professor Nicholas X. Fang and colleagues employed a 3D printing process they have pioneered, called microstereolithography, in which light from a projector is used to print patterns on successive layers of resin. They were able to print micron-scale features as small as the diameter of a human hair--dimensions that are at least one-tenth the size of what others have been able to achieve. Even after being stretched, twisted, and bent at extreme angles, the structures sprang back to their original forms within seconds of being heated to a certain temperature. The researchers ultimately want to use body temperature as the trigger. The process can be thought of as 4D printing, as structures are designed to change over the fourth dimension--time. "This will advance 4D printing into a wide variety of practical applications, including biomedical devices, deployable aerospace structures, and shape-changing photovoltaic solar cells," says SUTD professor Qi Ge.

UC Berkeley Launches Center for Human-Compatible Artificial Intelligence
Berkeley News (08/29/16) Jeffrey Norris

The University of California, Berkeley has launched the Center for Human-Compatible Artificial Intelligence (AI), which will work to ensure the most sophisticated AI systems of the future will be beneficial to humans. One of the biggest obstacles facing AI development is a machine's capability to understand human objectives and behave in a manner that is aligned with human values. The center will address this issue through inverse reinforcement learning, in which a robot can learn about humans by observing their behavior. For example, a robot can learn about the value of caffeine to humans by watching people wake up and brew their morning coffee. However, it will be difficult to program an AI system that can understand the objectives of a wide variety of people. "People are highly varied in their values and far from perfect in putting them into practice," notes Stuart Russell, an AI expert (and recipient, in 2005, of the ACM Karl V. Karlstrom Outstanding Educator Award) who will lead the new center. "These aspects cause problems for a robot trying to learn what it is that we want and to navigate the often-conflicting desires of different individuals." The center, which was launched with a $5.5-million grant from the Open Philanthropy Project, expects to add collaborators in the fields of economics, philosophy, and other social sciences.

Virtual Peer Pressure Works Just as Well as the Real Thing
New York University (08/25/16)

Virtual pressure from a computer-simulated peer is just as motivating as real peer pressure, according to researchers at New York University's Tandon School of Engineering. Moreover, the researchers say "fake" competition can be used for the good of science. The team formulated a mathematical model of human behavior that successfully predicted group responses across conditions. The researchers then designed an experiment to test whether virtual peer pressure could boost individual participation in a citizen science project. They reworked the interface of a citizen science project in which users view and tag images, adding an indicator bar at the top of the screen to show the number of times another participant had tagged the same image. This was the performance of the virtual peer, and the researchers developed five different scenarios for the virtual peer's performance. They say the highest-performing group of real participants were those who saw a virtual peer that consistently outperformed them. Conversely, the group who saw a virtual peer that underperformed them contributed less than any other group. In addition, the group whose virtual peer matched their own level of activity also outperformed the control group, potentially indicating the mere presence of a peer leads to increased performance.

Secret Commands in Online Videos Could Hack Your Smartphone
CNBC (08/29/16) Jennifer Schlesinger; Andrea Day

Researchers at Georgetown University and the University of California, Berkeley found voice commands hidden inside videos that play on mobile devices could be used to hack the smartphone using voice recognition. "[The command could be in] some popular YouTube video that has this strange noise in the background that a human being would just dismiss as an oddity, but that at the same time that noise could be controlling a cellphone that just happens to be located next to a computer," says Georgetown professor Micah Sherr. Because of differences in how computers and the human brain understand speech, the researchers constructed sounds understood as human speech only by computers, but not by a human brain. They demonstrated the technology by playing an audio clip that sounds like a muffled deep voice, but which the computer understood as a command to open Facebook. The researchers note opening Facebook is not going to hack a smartphone, but the technology can be used with more dangerous intentions. "They can cause your phone to open a website, a malicious website that has some malicious software on there that runs on your phone, takes over your phone," Sherr says. He notes the commands work because the voice-recognition systems are always listening for commands.

IoT Early Warning System Helps Save People From Mudslides
Network World (08/24/16) Deepak Puri

A team of El Salvadorian electronics, community development, and disaster relief specialists collaborated with local villagers and global experts to develop an Internet-of-Things (IoT)-based early warning system for disasters such as floods and mudslides. The Reaccion group determined the system needed to include input from village elders with experience in spotting signs of imminent catastrophe. It also had to be battery-powered, rugged, affordable, and easy to use for people lacking formal education. The team developed a prototype with the local FabLab, a design-thinking lab conceived by Massachusetts Institute of Technology professor Neil Gershenfeld to "solve local problems with local actors through technological innovation and the use of digital manufacturing tools." Villagers can alert each other with the IoT system even in the absence of power and telephone communications, using a battery-operated device with color-coded buttons to signal both an impending disaster and its severity. The device uses IoT weather sensors and accelerometers to read tremors. Reaccion links villages across mountainous terrain with an Early Warning System via a mesh network where each village has its own alerting device, which serves as a node. The devices store and update data information across the network in a peer-to-peer manner.

Abstract News © Copyright 2016 INFORMATION, INC.
Powered by Information, Inc.

To submit feedback about ACM TechNews, contact: [email protected]