Association for Computing Machinery
Welcome to the October 5, 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).


Machine Learning Needs Rich Feedback for AI Teaching
ZDNet (10/05/16) Chris Duckett

Researchers at Monash University in Australia say artificial intelligence (AI) systems need rich feedback to know why certain answers are incorrect. "How do we design systems so that they can take rich feedback and we can have a dialogue about what the system has learnt?" asks Monash professor Tom Drummond. He says researchers must be able to enrich AI feedback because otherwise machine learning is doomed to take much longer and generate much less satisfactory answers. Drummond notes one problematic feature of AI systems is the objective function that sits at the heart of their design. He says this raises a deep question of ethics about when to use machine-learning systems and when to hold back. Meanwhile, the next challenge for robot vision is to move beyond the structured environments that have been engineered for them in the past, and move out into the unstructured real world, especially via the use of self-driving cars. Robots also will need to be able to understand how to interact with objects and be able to infer intent. "All those kinds of things are about being able to model what will happen in the future to entities in my environment, the need for safe operation and helpful operation," Drummond says.

EU Project Successfully Deploys Robots Following Italy Earthquake
CORDIS News (10/03/16)

Following the massive earthquake that struck the Italian town of Amatrice and the surrounding area on Aug. 24, 2016, Italian officials requested the help of the European Union's TRADR project, which quickly deployed two unmanned ground vehicles (UGVs) and unmanned aerial vehicles (UAVs) to the region. The team wanted to use the UGVs and UAVs to provide three-dimensional (3D) textured models of two churches in Amatrice. The project team deployed two UGVs to the San Francesco Church, teleoperating them entirely out of line of sight, with one UGV providing a view of the other one. The arrangement allowed for easy maneuvering in a very constrained space with low connection bandwidth. One of the UGVs operated in the church continuously for four hours, and a UAV was present for a short time in parallel and provided additional views of the UGVs. The project team reported entering the building with a UAV was a tough challenge, which was overcome thanks to collaboration between three UAVs operating in parallel. The mission also fulfilled its goal to collect enough data for the construction of high-quality textured 3D models, which will be used to assess and repair the damaged churches.

Infosys Foundation USA, ACM, and CSTA Announce Awards for Teaching Excellence in CS
Computer Science Teachers Association (10/03/16)

ACM, the Computer Science Teachers Association (CSTA), and Infosys Foundation announced the rollout of the Awards for Teaching Excellence in Computer Science, which will allocate up to 10 annual awards of up to $10,000 each, with funding provided by Infosys. "Computer science [CS] teachers in K-12/pre-university represent an emerging profession," says CSTA executive director Mark R. Nelson. "By founding this award, Infosys Foundation USA is sending a powerful message to these computing educators worldwide that what they are doing is indeed important." ACM President Vicki L. Hanson says the new award "reinforces our long-held goals of recognizing the contributions of computer science teachers and building a framework that supports their professional development." CSTA, founded by ACM in 2004, has expanded to more than 23,000 members in 146 countries and is committed to addressing the need for qualified CS teachers. "As technology transforms every aspect of our lives, computer science teachers shoulder a tremendous responsibility as stewards of our children's CS education that will influence and shape our digital future," says Infosys Foundation USA chairperson Vandana Sikka. "Infosys Foundation USA is pleased to celebrate and honor outstanding computer science teachers throughout the world." The window for award applications and nominations runs from Oct. 1 to Nov. 1, 2016.

Making Construction Sites Safer
Ruhr University Bochum (Germany) (10/04/16) Julia Weiler

Researchers from Ruhr University Bochum (RUB) in Germany want to make construction sites safer with interactive virtual reality (VR) training courses. The researchers are developing technology to make it possible to experience sources of danger virtually. They say the technology is intended to help occupational health and safety experts check construction sites for critical areas in advance and plan appropriate safety measures. In addition, the researchers note construction workers could be trained in VR and sensitized to dangers. The team can use common three-dimensional models of construction sites as the basis for their representation in VR, which is the same technology found in video games to ensure the environment looks as realistic as possible. VR glasses enable users to explore the construction sites for which they will later be responsible. The RUB researchers say users also can interact with the environment, such as by lifting and carrying heavy objects, using controllers. "As you have an unlimited number of lives in virtual reality, we can observe there precisely how the test subjects react before and after fatal accidents and when learning effects come in," says RUB postdoctoral researcher Thomas Hilfert.

Nijmegen Provides Solution for Secure Processing of Patient Data
Radboud University Nijmegen (09/30/16)

Large-scale research involving medical data can be conducted without imperiling the security of the information or patient privacy thanks to a new method developed by researchers at Radboud University in the Netherlands. The method will be employed for a new, large-scale study of Parkinson's disease. Radboud professors Bart Jacobs and Eric Verheul devised the Polymorphic Encryption and Pseudonymization (PEP) technique, which pseudonymizes and encrypts data so it cannot be accessed even by the party who stores the data, while still allowing analysis. The data also is carefully regulated and monitored. In addition to blocking data access by data managers, PEP enables study participants to decide for each study if they want to allow their data to be used, and researchers who use the data are provided a unique key. Participants also are given a unique pseudonym for each researcher to prevent researchers from accessing restricted data by another route. "The study of Parkinson's should demonstrate the usefulness of PEP," Jacobs says. "With this showcase as an example, PEP could grow to become the international standard for storing and exchanging privacy-sensitive medical data."

Physicists Create World's First Time Crystal
Technology Review (10/04/16)

University of Maryland (UMD) researchers say they have created the world's first time crystal, which could lead to quantum memory systems. The researchers focused on quantum systems that are not in equilibrium, specifically a line of ytterbium ions with spins that interact with each other. They say the spin of the ions can be flipped up or down using a laser, so flipping the spin of one ion causes the next to flip, and so on. These spins oscillate at a rate that depends on how regularly the laser flips the original spin, which means the driving frequency determines the rate of oscillation. However, the researchers also found after allowing the system to evolve, the interactions occurred at a rate that was twice the original period, which can only be explained by the time symmetry being broken, thereby enabling these longer periods. The researchers then measured some of the properties of the resulting time crystals. For example, they found changing the driving frequency did not change the frequency of the time crystal. UMD professor Chris Monroe says these time crystals could be used for quantum information tasks, such as implementing a robust quantum memory.

Memristors Promise More Precise and Affordable Neuroprosthetics
CORDIS News (10/04/16)

Memristors could enable neuroprosthetics to process data on neuronal cell activity in real time. The use of the electrical component also could ease restrictive requirements on bandwidth, energy, and computational capacity for neuroprosthetics. Memristors regulate the flow of electrical current in a circuit and can remember the amount of charge that was flowing through it and retain data, even when the power is turned off. Scientists involved in the European Union-funded Real neurons-nanoelectronics Architecture with Memristive Plasticity (RAMP) project developed a nanoscale Memristive Integrative Sensor (MIS) and fed it a series of voltage-time samples, which replicated neuronal electrical activity. Acting like brain synapses, the metal-oxide MIS encoded and compressed neuronal spiking activity recorded by multi-electrode arrays up to 200 times. The RAMP team says the approach addresses bandwidth constraints and is power-efficient in that power needed for each recording channel was up to 100 times less when compared to current best practices. "Our work can significantly contribute towards further enhancing the understanding of neuroscience, developing neuroprosthetics and bio-electronic medicines by building tools essential for interpreting the big data in a more effective way," says Isha Gupta, a postgraduate researcher at the U.K.'s University of Southampton.

UVA Computer Science Professor Applies Genetic Engineering Principles to Cybersecurity
UVA Today (10/03/16) Charlie Feigenoff

Researchers at the universities of Virginia and New Mexico have developed Double Helix, a cybersecurity system based on structured diversity. The researchers say this type of system creates several functionally equivalent versions of a mission-critical system, but adjusts the binary code of some of the clones so the properties needed for successful attacks are missing, says University of Virginia (UVA) professor Jack Davidson. He notes when a cyberattack occurs, the behavior of the unprotected clones diverges from the protected ones, at which point Double Helix takes action to recover from the attack by modifying the affected clones. In the first phase of the project, the researchers used catalogs that classify attacks by their vulnerable property as a guide and modified clones to eliminate these targets. Phase two of the project will concentrate on attack recovery. "We must be able to analyze the state of the clones, determine the nature of the attack, and take remedial action, all in milliseconds," Davidson says. The team plans to use a combination of techniques to achieve this goal, rolling the replicas back to a known good state before the attack while creating a clone that is immune to the attack.

3D-Printed Robots With Shock-Absorbing Skins
MIT News (10/03/16) Adam Conner-Simons

Researchers at the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory have developed a new technique for three-dimensionally (3D) printing soft materials. Called "programmable viscoelastic material" (PVM), the technique can be used to program every part of a 3D-printed object to the exact level of stiffness and elasticity desired. The team says the soft materials have the potential to make robots safer and more precise in their movements. The researchers used the technique to outfit a cube robot--which moves by bouncing--with shock-absorbing "skins" that use less than half of the energy that would normally be transferred to the ground. Working with a standard 3D printer, the researchers used a solid, a liquid, and a flexible rubber-like material called TangoBlack+ to print both the cube and its skins. "It's hard to customize soft objects using existing fabrication methods, since you need to do injection molding or some other industrial process," says MIT postdoctoral researcher Jeffrey Lipton. The team thinks the PVM technique could improve the durability of everything from drones and phones to shoes and helmets.

Game Theory Research Reveals Fragility of Common Resources
Purdue University News (09/29/16) Emil Venere

Purdue University researchers have used game theory to study human decision-making when it comes to the utilization of common resources. The team applied the Nash equilibrium as well as prospect theory, which describes how people make decisions when there is uncertainty and risk. The researchers say the Nash equilibrium is key because it captures the idea that nobody is forcing someone to do the right thing. The findings reveal resources have a higher likelihood of failure at the Nash equilibrium under prospect theory, meaning people will over use common-pool resources, even if it risks failure of the system, to the detriment of society as a whole. The findings have implications for the management of engineered systems such as the power grid, communications systems, distribution systems, and online file-sharing systems, along with natural systems such as fisheries. "We are surrounded by large-scale complex systems, and as engineers we are trying to figure out how to design systems to be more robust and secure," says Purdue professor Shreyas Sundaram. He says the team plans to apply the approach to cybersecurity to study decision-making under risk.

Google Translate Gets a Deep-Learning Upgrade
IEEE Spectrum (10/03/16) Jeremy Hsu

Google has launched a Google Translate upgrade utilizing enhanced deep-learning techniques to produce more accurate translations. The neural machine translation system considers the entire sentence as one unit to be translated. The system relies on a recurrent neural network algorithm consisting of layered nodes, and a network of eight layers acts as the encoder and transforms the input into a list of vectors representing all possible meanings of each word. The second eight-layer network acts as the decoder and generates the translation one word at a time. Meanwhile, an attention network connects the encoder and decoder by directing the decoder to refer back to certain weighted vectors. Rare words are broken into a set of smaller, common subunits the system can handle more easily. Google researchers say this approach has reduced the number of translation errors by at least 60 percent versus the phrase-based method. The researchers used a chip designed for deep learning to make translations three times faster than ordinary chips without sacrificing accuracy. Google Translate already has begun using neural machine translation for its 18 million daily translations between English and Chinese, with improved translations for many more language pairs rolling out in the coming months.

Using Big Data to Monitor Societal Events Shows Promise, but the Coding Tech Needs Work
Northeastern University News (09/29/16) Thea Singer

Northeastern University researchers have analyzed the effectiveness of four global-scale databases and found they are falling short when tested for reliability and validity. The researchers studied Lockheed Martin's International Crisis Early Warning System (ICEWS), Georgetown University's Global Data on Events Lan­guage and Tone (GDELT), MITRE's Gold Stan­dard Report, and the University of Illinois' Social, Polit­ical, and Eco­nomic Event Data­base (SPEED). The researchers first tested the systems' reliability, trying to determine if they all detected the same protest events in Latin America. They found ICEWS and GDELT rarely reported the same protests, while ICEWS and SPEED agreed on only 10.3 percent of the events. The team then assessed the systems' validity, asking if the reported protest events actually occurred, and found only 21 percent of GDELT's reported events referred to real protests. Although ICEWS' track record was better, the system reported the same event more than once, increasing the protest count. The systems also were found to be vulnerable to missing news. "If some­thing doesn't get reported in a news­paper or a sim­ilar outlet, it will not appear in any of these data­bases, no matter how impor­tant it really is," says Northeastern professor David Lazer.

UNC Professor Training Internet to Recognize Images
Daily Tar Heel (NC) (09/28/16) Lorcan Farrell

University of North Carolina (UNC) researchers are teaching computers to learn and gain information from images. "The goal is effectively to make the computer see, to understand the images as we would understand them when we look at them," says UNC professor Jan-Michael Frahm. One of Frahm's recent projects involved using computers to recreate models of famous landmarks around the world. Frahm and a team of researchers from the Stevens Institute of Technology created software that processes images and turns them into three-dimensional (3D) recreations of landmarks around the world. "We took 100 million images that Flickr or Yahoo released and turned this effectively into a virtual version of the real world," Frahm says. He also is working to find a way to bypass facial recognition-based security software by creating a 3D model of a person's face using photos found online. "We want to show where the flaws lie right now so people can develop systems that actually are not vulnerable to these problems," Frahm says. "The problem is there are wrong assumptions in the security systems on what's hard to do and that's why they fail. They make unreal assumptions of the attack."

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

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