Welcome to the July 13, 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
Google's AI Is Learning How to Save Your Life
IDG News Service (07/12/16) Katherine Noyes
Google DeepMind researcher David Silver says advancing human health via personalized medicine will be the next phase of the AlphaGo deep-learning artificial intelligence's (AI) evolution. U.K.-based DeepMind has announced an alliance with London's Moorfields Eye Hospital, which will apply machine learning to diabetic retinopathy and age-related macular degeneration, while several clinical mobile apps also are under development. Silver says the innovations that enabled AlphaGo to beat human Go players, using convolutional neural networks and reinforcement learning, also could be relevant to healthcare. He notes these techniques make the AI capable of self-education over time. "Of course beating [Go champion Lee] Se-dol was exciting, but for me, even more exciting than the achievement itself was the manner in which AlphaGo did it," Silver says. "It showed it can learn from data and self-play to figure out knowledge for itself." Silver is confident AlphaGo's achievements demonstrate the viability and practicality of reinforcement learning. "Now we can look around at many different domains," he says. "We're by no means done with AlphaGo."
Computer Hackers Don't Stand a Chance Against These Girls
The Washington Post (07/11/16) Raymond M. Lane
The GenCyber program consists of 119 summer camps sponsored by the U.S. National Security Agency (NSA) and the National Academy of Sciences. The free camps are designed for girls ranging from kindergarten through 12th grade, and some of the camps are geared toward teachers who want to integrate computer security literacy into their classrooms. This year, Virginia hosted the most camps with 11, followed by Texas with 10 and Hawaii with 9. Next year, GenCyber organizers expect to have 200 camps nationwide. The camps are for girls who want to understand how the Internet, computers, smartphones, and other wireless devices can be kept safe from bullies, hackers, spies, and terrorists. "We focused on team building, learning new words and concepts, and then just pumped it up," says GenCyber teacher Shade Adeleke, who also is a science, technology, engineering, and math instructor at Prince George's Community College. At one Maryland-based camp, students created a "honeypot," an unprotected website rigged so they could watch hackers breaking into it. Campers also visited the NSA's National Cryptologic Museum to learn how national security data protection works.
In-Ear EEG Makes Unobtrusive Brain-Hacking Gadgets a Real Possibility
IEEE Spectrum (07/07/16) Eliza Strickland
Electroencephalogram (EEG) sensors that fit inside the ear are being developed by two research teams to provide reliable brain data for use in unobtrusive brain-hacking applications. John Chuang's lab at the University of California, Berkeley has tweaked a commercial EEG headset and redirected the electrode from the forehead to the ear canal. The group aims to use this setup to transmit mental commands to control a computer, a drone, or another electronic device. The team tested five different "mental gestures" and selected the two with the clearest EEG signatures as a potential binary control system for any linked device. Meanwhile, Imperial College London's Danilo Mandic is working on an EEG interface using an in-ear sensor fashioned from an off-the-shelf noise-blocking earplug made from memory foam. Two electrodes composed of a soft silver-coated fabric are attached to the sides of the earplug, enabling the collection of high-quality EEG signals. Mandic says his team is clearing a path for a "truly wearable" EEG system, while also setting up "completely new avenues in 24/7 monitoring of the state of body and mind." Among the uses envisioned by Mandic are monitoring the progress of chronic diseases, tracking sleep patterns, and keeping tabs on military personnel's mental state and fitness for duty.
How to Keep More Girls in IT at Schools If We're to Close the Gender Gap
The Conversation (07/10/16) Karin Verspoor
The gender gap in the information technology (IT) industry is underscored by a lack of engagement of girls in computing and IT education, writes University of Melbourne professor Karin Verspoor. She notes most IT classes in public schools focus on basic skills, such as accessing educational games and quizzes online and managing spreadsheets. Programming, computational thinking, and algorithmic problem solving typically are only taught in extracurricular classes and clubs. Verspoor says opt-in training often precludes students who do not know what programming is or who do not identify with computer or gaming culture. Meanwhile, she says, gender stereotyping can dissuade girls from joining coding clubs, as parents tend to buy technology more for boys than for girls, and female students might feel unwelcome in a male-dominated club. According to Australia's Digital Careers group, the best strategy to increase girls' interest in IT is a compulsory integrated digital technologies curriculum that is gender inclusive. Schools in the U.K. and Australia are beginning to implement coding curricula to teach computing skills and the collection and interpretation of data using automated tools. In addition, the University of Adelaide is helping to prepare Australian teachers for the new curriculum by offering online courses about digital technologies.
Wide-Field Ethnography: Studying Software Engineering in 2025 and Beyond
CCC Blog (07/11/16) Helen Wright
University of Washington Bothell professor David Socha and colleagues have developed the concept of wide-field ethnography (WFE), in which large, multimodal, and multi-stream datasets of physical-social-economic-cyber systems (PSECs) are collected and analyzed in action. The researchers say the process yields datasets that help scientists to individually and collectively track work over space and time, pursue a broad spectrum of research questions, enable interdisciplinary analysis by communities of researchers, go beyond in-situ human observation, mitigate against researchers' preconceptions of what type of data to gather, and develop and assess machine learning algorithms on naturalistic datasets. Socha says his team's current WFE research concentrates on PSECs also to account for the economic aspects of such systems, as revealed through various value exchanges across multiple units of analysis. Among the fields WFE can benefit are software engineering, computer-supported cooperative work, human-computer interaction, user research, human-centered design, anthropology, sociology, and education. Socha says he and his colleagues focus on human collaboration in naturalistic settings to co-design and co-evolve complex systems. This field includes studying agile software development practices, and how software developers and students design. Socha says his team is working on machine-learning algorithms that can automatically detect events of interest for further exploration and analysis.
Duke Robotics Research Team Creates Chip to Expand Robots' Motions
Duke Chronicle (07/11/2016) Ajay Desai
Duke University researchers have developed a prototype chip that enables a robot to efficiently plan its movements. "For robots [motion planning] is really hard, and it turns out that this is a very computationally intensive problem," says Duke professor Daniel Sorin. "Effectively, what is happening is that the robot is considering all of the possible motions it could do and trying to figure out [if each motion could] cause [the robot] to collide with an obstacle in its environment." Doing all of those tests to see if a motion will cause a collision requires a vast number of computations on a typical computer. Sorin says the Duke team built special-purpose hardware to address these power and computing challenges. "It is so special-purpose that it does nothing else but motion planning," he says. "But by virtue of only doing that one thing, it can do it really, really well." Sorin says the chip runs an algorithm to simultaneously consider thousands of potential movements. He notes the prototype has been successfully tested on a robotic arm, and the team is continuing to enhance the chip for use in industrial and commercial settings. The Duke team is exploring extending applications to autonomous cars and unmanned aerial vehicles.
MIT News (07/08/16) Larry Hardesty
Researchers at the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory have developed software that could make it easier for laypeople to work with databases. The program's home screen looks like a spreadsheet, but it lets users build their own database queries and reports by combining functions familiar to spreadsheet users. Drop-down menus enable the user to pull data into the tool from multiple sources. The user can then sort and filter the data, recombine it using algebraic functions, and hide unneeded columns and rows. The researchers note the tool will automatically generate the corresponding database queries, enabling "direct manipulation" of data. MIT professor David Karger says the new tool could help an organization get up and running with a new database without having to wait for a custom interface. The tool's main drop-down menu has 17 entries, and Karger and MIT graduate student Eirik Bakke say those functions are enough to perform any database query possible in SQL-92. The tool currently enables query construction on an existing database, but it does not facilitate direct data entry or modification. The MIT team recently presented the tool at the ACM International Conference on Management of Data (SIGMOD/PODS 2016) in San Francisco.
Science on the Verge of Creating 'Emotional' Computer
National Research Nuclear University (07/08/16)
Researchers at the National Research Nuclear University's Moscow Engineering Physics Institute (MEPhI) are developing Virtual Actor, an emotion-based artificial intelligence. The researchers say the system will have both an emotional and a narrative intellect, and it will be able to understand the context of what is going on, as well as unfolding scenarios. Based on this information, Virtual Actor can make plans and set targets. "Our principal goal is to formulate the basic principles that natural intelligence in the human brain is built upon," says MEPhI professor Alexei Samsonovich. The researchers also want to create a simplified form of Virtual Actor that can be played like a video game. "A virtual agent and a real person control the figures on a computer screen, interacting with each other, thus building social rapport based on emotionally charged actions," Samsonovich says. He says the agent also must be able to study as a thinking person, instead of via programming or "carrot and stick" reinforcement learning. The researchers note Virtual Actor must be able to set learning goals, formulate questions to achieve these goals, and actively seek the answers, which requires logical thinking, all aspects of perception, decision-making, meta-thinking, and many other cognitive functions.
Machine Learning Puts New Lens on Autism Screening and Diagnostics
University of Southern California (07/06/16) Amy Blumenthal
Researchers at the University of Southern California (USC) Signal Analysis and Interpretation Laboratory (SAIL) are collaborating with autism research leaders to investigate whether machine learning might help screen for autism and guide caregiver and practitioner intervention. They applied machine learning methods to analyze how parents' responses to individual items and combinations of items contained in two standard industry tests correlated with children's overall clinical diagnoses of autism spectrum disorder (ASD), compared to non-ASD diagnoses. Thousands of caregivers' test scores were analyzed to determine redundancies in the questions, and their elimination enabled the researchers to identify five Autism Diagnostic Interview-Revised test questions that seemed able to maintain 95 percent of the instrument's performance. The researchers believe these techniques could reduce administrative time and tailor questions to find the novel challenges justifying intervention for a particular individual. They also think machine learning can furnish a clearer and more data-enhanced picture for caregivers and practitioners, which they say could be revolutionary by removing "the guesswork or subjectivity involved even in trusted, industry-wide instruments." In addition, the SAIL team is working to produce quantitative measures of human behavior based on audio, video, and physiological sensors via signal processing.
Extortion Extinction: Researchers Develop a Way to Stop Ransomware
University of Florida News (07/07/16) Steve Orlando
University of Florida (UF) researchers have developed CryptoDrop, a system they say can thwart ransomware. CryptoDrop does not prevent ransomware from starting, but it prevents malware from completing its task. "So you lose only a couple of pictures or a couple of documents rather than everything that's on your hard drive, and it relieves you of the burden of having to pay the ransom," says UF doctoral student Nolen Scaife, a founding member of UF's Florida Institute for Cybersecurity Research. The UF team describes CryptoDrop as an early-warning system, and says its results have been impressive. During a run against several hundred ransomware samples that were live, CryptoDrop detected 100 percent of the samples, and it did so after only an average of 10 files were encrypted, according to Scaife. The team also says CryptoDrop works seamlessly with antivirus software. "About one-tenth of 1 percent of the files were lost, but the advantage is that it's flexible," says UF professor Patrick Traynor. "We don't have to wait for that antivirus update. If you have a new version of your ransomware, our system can detect that."
Researchers Want to Achieve Machine Translation of the 24 Languages of the EU
Saarland University (07/07/16)
Saarland University researchers are developing an automated system for translating between the languages of the European Union (EU) so comprehensible texts are achieved for as many language combinations as possible. Due to the complexity of some EU languages, the researchers taught the computers to recognize patterns in huge text repositories and to learn from them, instead of feeding the computers with grammar rules and linguistic details. "This machine learning strategy has nothing to do with natural intelligence, but it does have similarities with the processes that occur in the human brain when we control the muscles in our bodies," says Saarland University professor Josef van Genabith. The Saarland researchers are working with QT21, a consortium of 14 leading research institutions for machine translation in Europe and Hong Kong, which includes universities, research institutions, and private companies. "Our common goal is to exploit machine learning to significantly improve automatic translation, particularly of more complex languages such as Latvian or Czech," van Genabith says. Meanwhile, van Genabith and the German Research Center for Artificial Intelligence also are leading the European Language Resources Coordination, which aims to collect suitable language datasets that will enable the European Commission's automated translation platform to be adapted and optimized for daily requirements of public administrators in all EU member states.
Startling Data Reveals Computer Science Education Conundrum–-and Why It Matters
eCampus News (07/06/16) Laura Devaney
Thousands of computer science jobs with six-figure salaries remain open across the U.S., due in large part to a deficit of students pursuing computer science education at the college level and earning degrees in the field. "Barely one in eight U.S. high schools teach [Advanced Placement] computer science, which leaves many students unable to pursue it at the college level and qualify for these high-paying jobs," says ACT/The App Association's Jonathan Godfrey. He also notes companies want to hire qualified software developers, but are having trouble finding them. A report from ACT/The App Association estimates there are more than 223,000 unfilled job openings for software developers. Salaries paid to software developers bring more than $114 billion to the economy, and the average salary for software developers nationwide is more than $104,000, according to the report. In addition, the report says 89 percent of software developers are employed outside Silicon Valley. Meanwhile, a report from the Information Technology and Innovation Foundation found although interest in computer science education, and access to it, is expanding, insufficient numbers of students are taking high-quality computer science classes at the high school and university levels.
It's Automatic: CMU Smartphone App Manages Your Privacy Preferences
Carnegie Mellon News (PA) (07/06/16) Byron Spice
Managing privacy settings is a task that can require smartphone users to make more than 100 decisions, but a personalized privacy assistant application being developed at Carnegie Mellon University (CMU) could simplify the task. The research demonstrates that people's preferences can be organized in a small number of categories or "profiles," which differ based on people's willingness to grant different types of apps access to their information. The privacy assistant uses machine learning techniques to analyze a user's response to a small number of questions focusing on the particular apps they have on their phone. The privacy assistant can learn the user's preferences and quickly recommend the most appropriate settings, such as with which app to share the user's location, or contact list. Users can accept or reject the privacy assistant's recommendations. During testing, participants accepted almost 80 percent of the recommendations made by the privacy assistant, and indicated they were more comfortable with their privacy settings than users who did not have a privacy assistant. "Our findings suggest that the personal privacy assistant does a good job of properly profiling each user and that its recommendations based on those profiles were useful," says CMU professor Norman Sadeh. He thinks such a tool eventually could become trustworthy enough to automatically make these decisions.
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.