Welcome to the June 10, 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
ICANN Endorses Plan to Cede Internet Oversight
Agence France-Presse (06/09/16) Glenn Chapman
The U.S. government has endorsed a plan to cede its oversight of the Internet Corporation for Assigned Names and Numbers (ICANN) to the broader online community. The plan aims to maintain Internet governance under a "multi-stakeholder" model that avoids control of the online ecosystem by any single governmental body. "The Internet's multi-stakeholder community has risen to the challenge we gave them to develop a transition proposal that would ensure the Internet's domain name system will continue to operate as seamlessly as it currently does," says the U.S. Commerce Department's Lawrence Strickling. The plan is a response to the U.S. government's March 2014 announcement that it would transition "stewardship" of online domain name system technical functions from the Commerce Department to a body that would fairly represent all parties with interests in a vibrant and healthy Internet. The proposal was crafted over the course of two years with input from businesses, academia, governments, and others. The plan will not affect how users interact online, but will turn over the technical supervision of the online address system to ICANN. The new system will include checks and balances so no single entity can exert control over the Internet.
Google Moves Closer to a Universal Quantum Computer
Nature (06/08/16) Philip Ball
A research team has made an experimental prototype of a universal quantum computer that can solve a wide range of problems in fields such as chemistry and physics, and has the potential to be scaled up to larger systems. The Google prototype combines the best of analog and digital approaches to quantum computing. Google computer scientists and physicists at the University of California, Santa Barbara (UC Santa Barbara) and the University of the Basque Country in Bilbao, Spain, used a row of nine solid-state quantum bits (qubits) fashioned from cross-shaped films of aluminum about 400 micrometers from tip to tip, deposited them onto a sapphire surface, and cooled the metal to turn it into a superconductor with no electrical resistance. Information could be encoded into the qubits in their superconducting state, and interactions between neighboring qubits were controlled by logic gates that steer the qubits digitally into a state that encodes the solution to a problem. The researchers say their approach should enable a computer with quantum error correction, and they predict devices with more than 40 qubits could be a reality in a couple of years. "At that point, it will become possible to simulate quantum dynamics that is inaccessible on classical hardware, which will mark the advent of 'quantum supremacy,'" says UC Santa Barbara's Daniel Lidar.
The AI Machines Undergoing Behavioral Psychology Tests
Technology Review (06/08/16)
Junhyuk Oh and colleagues at the University of Michigan (U-M) are experimenting with using mazes created in Minecraft to test the cognitive skills of artificial intelligence (AI) machines. In real-world settings, learning can be difficult for machines because objects can be partially obscured, vision and movement have to be carefully coordinated to succeed, and the resulting reward is often delayed. The online world of Minecraft is an ideal platform because mazes are easy to produce and the flexibility of the domain allows for the construction of even more challenging cognitive tasks to further evaluate architectures. The U-M team reports the best-performing AI system uses deep reinforcement learning enhanced with additional memory. "Our main empirical result is that context-dependent memory retrieval can more effectively solve our set of tasks," the researchers say. They note the research raises the prospect that AIs first will perform advanced tasks in virtual reality settings such as Minecraft, and ethical issues of AI could be explored in such a benign environment.
Psychologists Grow Increasingly Dependent on Online Research Subjects
Science (06/07/16) John Bohannon
The growing popularity of online subjects recruited for social science research, although advantageous in terms of speed and efficiency, is a concern for researchers. Participants at the Association for Psychological Science meeting in May cited the use of Amazon Mechanical Turk (MTurk) for research as an example of psychologists' excessive reliance on a commercial platform, as well as unresolved questions about the accuracy of such studies and the authenticity of the subjects participating in them. Last month, 23,000 volunteers completed 230,000 tasks on their computers via MTurk in 3.3 million minutes. In comparison, only 61 studies using MTurk were published in 2011. Among the issues with MTurk meeting attendees raised are the clunkiness of its interface, and the fact that adapting the platform for social sciences frequently demands coding skills few have. The most challenging problem for researchers has been the size and diversity of the MTurk user community, in view of concerns the number may be smaller than advertised, leading to recirculation of the same subjects across experiments, which can bias outcomes. An early effort to calculate the effective MTurk research cohort found far lower numbers of subjects willing to take part in an experiment at any one time, and far less diversity. Additional factors include the low salary MTurk subjects earn and a lack of ethical protections and anonymity.
Goal of Improved Connectivity Drives Computer Scientists
Cornell Chronicle (06/08/16) Louis DiPietro
A team led by Cornell University researchers has developed an open source, software-defined networking (SDN) platform designed to make networks more customizable, adaptive, and secure. Researchers from Cornell are working with Princeton University and the Universities of Massachusetts Amherst and Catholique de Louvain to explore software solutions to common problems facing network management, such as visibility and global control. "The main benefit is making network admins' jobs easier," says Cornell professor Nate Foster. "Especially nowadays, some networks are huge, like Google's, and customers expect incredible performance and uptime. That's the context where this will have the most impact." The controller platform, Frenetic, gives network administrators a greater ability to maintain large networks and automate certain processes. In contrast to traditional networking hardware, which is cumbersome and costly to program, SDN software enables customization. "Instead of buying a fixed function box from company x, you can define your own functionality," Foster says. "That's where the software comes in." The team plans to explore Frenetic's capability to predict network errors using probabilistic reasoning.
Using Computers to Better Understand Art
The Conversation (06/07/16) Ricky J. Sethi
Visual stylometry is a new research field designed to measure artistic style via computational and statistical methods to uncover unique insights about artists and artworks, writes Fitchburg State University professor Ricky J. Sethi. "Computer analysis of even previously well-studied images can yield new relationships that aren't necessarily apparent to people, such as Gaugin's printmaking methods," he says. "In fact, these techniques could actually help us discover how humans perceive artworks." Sethi says his team, composed of experts in computer science, art philosophy, and cognitive science, is developing a digital image-analysis tool for studying paintings called Workflows for Analysis of Images and Visual Stylometry (WAIVS). Based on the Wings workflow system, WAIVS enables users to build analyses in the same manner as drawing a flowchart. For example, to compare tonal analyses of an entire painting and the background alone would not require creating complex computer software, but instead a simple diagram of the process. "WAIVS includes not just discrete tonal analysis but other image-analysis algorithms, including the latest computer-vision and artistic-style algorithms," Sethi says. He also notes his group has incorporated into WAIVS convolutional neural network methods for separating artwork style from content.
Haptic Taco Helps You Navigate by Feel
IEEE Spectrum (06/07/16) Evan Ackerman
Yale University researchers are developing small haptic peripherals that are designed to help drivers navigate using just their sense of touch. The "Haptic Taco" is a little cube that expands and contracts in the user's hand to navigate to a predetermined destination. "The continuous contour improves the haptic impression of the device modifying its volume, rather than simple linear motion of two faces," says Adam J. Spiers, a member of Yale's Grasping & Manipulation, Rehabilitation Robotics, and Biometrics (GRAB) Lab. The device's functionality is based on growing and shrinking. To navigate with it, users pair it with a navigation app on a mobile device; the device will maximize its volume into a rectangular prism, and then slowly shrink itself back down to a cube as the user moves toward the destination. The goal is to provide intuitive but not distracting sensations, according to Spiers. The researchers tested the Haptic Taco by asking a group of users to navigate around a series of invisible points in a small room. On average, using the Haptic Taco was about half as efficient as taking a straight-line path. The researchers note the one-degree-of-freedom Haptic Taco performed just as well as the two-degrees-of-freedom Haptic Sandwich, suggesting providing less information to the user does not necessarily lead to worse navigation performance.
Flight of the RoboBee
National Science Foundation (06/07/16) Aaron Dubrow
Researchers involved with the U.S. National Science Foundation-supported "RoboBees" project recently presented work demonstrating that their aerial microrobots now can perch during flight to save energy, in the same way as bats, birds, and butterflies. The team used an electrode patch, which takes advantage of electrostatic adhesion, to enable the RoboBees to stick to almost any surface, from glass to wood to a leaf. The mechanism requires about 1,000 times less power to perch than it does to hover. "When making robots the size of insects, simplicity and low power are always key constraints," says Harvard University professor Robert Wood, who is leading the project. The RoboBees project aims to create autonomous robotic insects capable of sustained, independent flight. Wood estimates it will take another five to 10 years before the RoboBee might be ready for use in the real world. The robots could one day assist in reconnaissance, aid in remote communication, or act as artificial pollinators. The team now has begun work on making the perching mechanism omnidirectional and developing onboard power sources that could enable RoboBees to fly untethered.
Deep Learning Helps to Map Mars and Analyze Its Surface Chemistry
University of Massachusetts Amherst (06/07/16) Janet Lathrop
Researchers at the University of Massachusetts (UMass) Amherst and Mount Holyoke College are working together to apply recent advances in machine learning to analyze large amounts of scientific data from Mars. The U.S. National Aeronautics and Space Administration's (NASA) Curiosity rover has been exploring a crater on Mars since August 2012 and sending back a steady stream of specialized camera images and data on the chemical composition of rocks and dust for analysis. The researchers are exploring how machine-learning methods can be used to provide a practical and useful new tool for handling these types of large scientific datasets. "Our study will test the ability not at recognizing objects on Earth, but understanding planetary geochemistry from the Martian rock tests," says UMass Amherst professor Sridhar Mahadevan. The researchers hope within four years they will be able to show deep learning can have a much better success rate than other, previous methods of differentiation. "Nobody has actually shown that deep learning techniques can deliver what we hope for from these scientific data sets," Mahadevan says. "Somebody has to explore the question."
Walking and Talking Behaviors May Help Predict Epidemics and Trends
Penn State News (06/06/16) Matt Swayne
Mobile phone data could reveal an underlying mathematical connection between how users move and how they communicate that could make it easier to predict how diseases, or even ideas, spread through a population, according to an international team of researchers. "What we are starting to see with big data is that there is a very deep regularity underlying much of what we do," says Pennsylvania State University professor Dashun Wang. In a study, location and communication data collected from three international mobile phone carriers showed people move and communicate with predictable patterns. Since movement and communication are connected, other researchers may only need one type of data to make predictions about the other set. "What this mathematical equation allows us to do is to derive one from the other," Wang says. The researchers tested the equation on a simulated epidemic and found either location or communication datasets could be used to reliably predict the movement of the disease. "One application we showed is if we know who communicates with whom in a country, we would be able to estimate how a virus would spread within that country," Wang says. In addition, the researchers think they could use the data to predict how ideas and trends permeate through a culture.
Developer Teams Up With Facebook and Google to Make 'Machines See'
TechWorm (06/05/16) Kavita Iyer
Facebook and Google are working with a Russian developer to create an open source computer-vision platform that behaves like a teaching machine and enables machines to "see." "The project opens up the field of computer vision to a greater audience of developers," says Facebook researcher Balmanohar Paluri. The platform is designed to remove several technological barriers traditionally faced by scientists in the fields of computer vision and neural networks. The researchers have open-sourced their work, which they say means any startup in this field can launch a project in a matter of days rather than years. The researchers do not view this as a commercial project, and say its purpose is developing the community and its long-term prospects. "Integration of machine learning and computer vision in a unified development kit is an important step towards stimulating the creation of the new technologies and products in such strategic industries as robotics and artificial intelligence," says the Skolkovo Robotics Center's Albert Efimov.
USC Viterbi Alumna Designs Thinking Robots and NASA Rovers
USC News (06/03/16) Sam Corey; Daniel Druhora
Georgia Institute of Technology professor Ayanna Howard, an alumna of the University of Southern California (USC) Viterbi School of Engineering, was honored this year by the Computing Research Association with the A. Nico Habermann award "for her sustained commitment to increasing diversity, combined with her distinction in research." Howard's early work includes programming the first commercial genetic algorithm package as director of the Axcelis software team. Her tenure at the U.S. National Aeronautics and Space Administration's Jet Propulsion Laboratory included work on computer vision and mobile systems, with the goal of advancing technology for deploying robotic rovers into unknown, threatening environments. Howard's postdoctoral research at USC Viterbi focused on humanized intelligence, and her 1999 dissertation involved designing robotic hands that could bend, twist, change shape, and grip objects. USC Viterbi professor Timothy Pinkston praises Howard for contributing new ideas to robotics, as well as boosting diversity in the discipline. "Increasing the diversity of experiences in engineering is important not just because it promotes diversity of thought and perspective," Howard says. "But also because it promotes diversity of skills needed to solve our global challenges." Howard also predicts within two to three decades, "robots will work for us and with us."
Data Mining of Twitter Posts Can Help Identify When People Become Sympathetic to Groups Like ISIS
Lancaster University (06/03/16)
Researchers have used data-mining techniques to determine when Twitters users start displaying supportive behavior to terror groups such as ISIS. Lancaster University collaborated with Open University to analyze 154,000 Europe-based Twitter accounts and more than 104 million tweets in English and Arabic relating to Syria. The researchers attempted to show users of social media platforms are more likely to adopt pro-ISIS language--and therefore display potential signs of radicalization--when connected to other Twitter users who are linked to many of the same accounts and share and re-tweet similar information. They provide evidence that shows when users start sharing tweets from known pro-ISIS accounts or using extremist language, they quickly display a large change in the language they use, tweeting new words and terms, and indicating a clear shift in online behavior. Users' language changed to use religious words more frequently, such as Allah, Muslims, and Quran. "This clear change suggests that users are rejecting their prior behavior and escalating their new behavior until displaying radicalized signals," says Lancaster researcher Matthew Rowe. Still, he says more work is needed to check the robustness of the data-mining techniques due to the relatively small sample size of the analysis.
Abstract News © Copyright 2016 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.