Association for Computing Machinery
Welcome to the September 4, 2015 edition of ACM TechNews, providing timely information for IT professionals three times a week.

Please note: In observance of the Labor Day holiday, TechNews will not be published on Monday, Sept. 7. Publication will resume Wednesday, Sept. 9.

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).


Silicon Valley, Seeking Diversity, Focuses on Blacks
The New York Times (09/03/15) Patricia Leigh Brown

The U.S. technology industry is courting African-Americans to join its workforce in Silicon Valley in order to become more diverse, as blacks are the most underrepresented minority in the sector. Initiatives to address this problem include many recently established grassroots efforts, such as the Hidden Genius Project, which exposes black high-schoolers to programming, Web and app design, team-building, and other advantageous skills for the tech world. Partly fueling the emphasis on greater diversity were findings by the U.S. Census Bureau and tech giants that representation of African-Americans and Hispanics has been consistently low in science, technology, engineering, and mathematics (STEM) careers. Like other grassroots projects, #YesWeCode seeks to cultivate a generation of black entrepreneurs who create profit-making apps, according to program founder Van Jones. Lack of access to venture capital is a hindrance many blacks are confronted with, which new networking groups are trying to remove by matching potential investors with African-American engineers and other tech-oriented specialists and entrepreneurs. "The most important ingredient for a tech company is talent," notes Saama Technologies chairman Ken Coleman. "It's short-sighted to overlook talent anywhere." In addition, promoting entrepreneurship and boosting STEM graduates is now a pressing issue for historically black colleges and universities.
View Full Article - May Require Free Registration | Return to Headlines | Share Facebook  LinkedIn  Twitter 

More Students Taking AP Physics, Computer Science Exams
Education Week (09/03/15) Liana Heitin

Participation rates for Advanced Placement (AP) science exams rose sharply over the last year, with computer science seeing the second-largest overall increase, according to the College Board. AP computer science courses saw participation rates rise 25 percent over the last year, from 39,200 to 49,000 students. National Science Teachers Association executive director David L. Evans says the increase is likely due to a rise in both interest and availability. "There's been a lot of effort to increase kids' awareness about coding and computer science in various ways," Evans says. "We're seeing a lot more schools offer AP computer science courses." Boys continued their dominance of AP Computer Science, making up about 78 percent of the test takers over the last year. However, the participation rate for girls rose 2 percent, while underrepresented minorities increased their participation by half of a percent. The AP Computer Science exam has long had issues with diversity. Some states have seen no female, African-American, or Hispanic students take the test in recent years. The College Board is attempting to address this disparity with a new course, AP Computer Science Principles, which will focus on a broader range of skills and provide teachers with greater flexibility. The new course is expected to debut in 2016.

A New Design for Cryptography's Black Box
Quanta Magazine (09/02/15) Kevin Hartnett

In the two years since the proposal of indistinguishability obfuscation (IO) shook the cryptography world, barriers to realizing this milestone have hindered its use--but new studies in recent months hint at a practical IO deployment within a decade, if not sooner. IO involves positing two programs that use different approaches to compute the exact same outputs, and encrypting them so users cannot distinguish between which version they have. Extremely slow speed and a lack of mathematical security made the concept impractical, but research has shown how to go from single-circuit obfuscation to obfuscation of a general-purpose computer via punctured programming. Punctured programming can obfuscate longer, open-ended calculations as a series of small, practical steps. "The main technical achievement applies IO for circuits to the local steps of a computation and ties things together so you are protecting the computation globally," says Columbia University professor Allison Bishop. A year after the IO proposal, Bishop and other researchers published papers breaking IO down into simple assumptions about multilinear maps, which computer scientists are attempting to substitute with a better-understood mathematical obstacle, such as learning with errors. Researchers currently are watching for the establishment of a simpler mathematical foundation for IO security.

Great Innovative Idea--End-to-End Training of Deep Visuomotor Policies
CCC Blog (09/02/15) Helen Wright

University of California, Berkeley researchers propose end-to-end training of deep visuomotor policies to improve robotic learning.  They say manually designed components constrain the applicability of robotic learning to relatively controlled scenarios, and their learning algorithm can be used to obtain robotic control policies that map directly from raw inputs to raw outputs.  The researchers note deep-learning advances can facilitate this process.  They also address the challenge of high dimensionality of learned function approximators via guided policy search, in which the reinforcement learning problem is converted into a supervised learning problem, with supervision supplied by an optimal control "teacher" that solves simplified versions of the task to produce training data for a neural network policy.  The researchers suggest this type of end-to-end robotic learning has potential for making robots capable of handling complex, unstructured environments.  They also say their method, when used to train machines' perception and control systems, can span the gap between perception and control and give robots autonomously learned sensorimotor skills.  The researchers say this learning method would enable robots to accrue refined behaviors they need to succeed in complex, unstructured settings.  The researchers also concentrate on approaches to effect robotic learning from humans via learning from demonstration, computer vision and perception techniques incorporating active decision-making, and robotic control methods that leverage and reconfigure prior experience to accommodate new tasks.

This AI Creates Interactive Fiction by Reading Other People's Stories
Motherboard (09/02/15) Emanuel Maiberg

An artificial intelligence (AI) developed by Georgia Institute of Technology (Georgia Tech) researchers can generate interactive narratives by reading and learning from stories written by human authors. Scheherazade-IF was programmed to read several hundred stories about either going on a date to the movies or robbing a bank. The AI then used pattern recognition to ascribe importance to events in the stories, as well as learning that certain events tend to occur before others, to plot out a story map. "Humans are pretty good storytellers and possess a lot of real-world knowledge," notes Georgia Tech professor Mark Riedl. "Scheherazade-IF treats a crowd of people as a massively distributed knowledge base from which to digest new information." Scheherazade-IF's writing skills were tested by comparing it to a version of the system that produced random stories, and one that had a "perfect" script coded by a single human storyteller. The researchers then asked people to play the different versions and report how often the story lapsed into nonsense. In some cases, Scheherazade-IF's performance was comparable to the human-programmed version, and in other scenarios it did approximately 83 percent as well.

Your Smartphone Can Tell If You're Bored
Technology Review (09/02/15) Rachel Metz

Researchers at Telefonica Research in Spain have developed an algorithm that enables a smartphone to determine whether or not its user is bored.  The algorithm is designed to look at the mobile activity of the user and consider factors such as the last time the user had a call or text, the time of day, and the level of smartphone use.  Such data provides a reliable prediction of boredom 83 percent of the time, according to the researchers.  Using machine learning to infer the state of mind of people, and doing so reliably via a smartphone, could be powerful.  For example, an app could predict the user is bored, know the location of the user, and then try to provide content it thinks the user would like in that particular context.  The researchers determined characteristics of boredom using an Android app to ask study participants to rate their level of boredom several times a day over two weeks; the responses were compared with other data captured from the phones. To validate the algorithm, researchers developed another Android app that inferred on its own whether the user was bored and, when it did, sent an alert to their phone asking if they wanted to read an article on BuzzFeed's news app. The researchers plan to present their work at the ACM UbiComp2015 ubiquitous computing conference in Japan next week.

Intel Promises $50 Million for Quantum Computing Research
IDG News Service (09/03/15) Peter Sayer

Intel has announced it will provide $50 million to fund research into quantum computers being conducted in the Netherlands. The funds will go to QuTech, a research unit at the Technical University of Delft. Intel says the funds will be distributed over a period of 10 years, and the company also will provide QuTech with staff and equipment to support its work. QuTech says the partnership will help the research unit bring its theoretical work on quantum computers to life by enabling it to tap Intel's manufacturing expertise to produce larger-scale quantum computing devices. Intel CEO Brian Krzanich says the company is supporting QuTech because of its long experience in the field of quantum computing, particularly its work on the interconnects that are used to link parts of quantum computers together. Intel expects the collaboration will help the company develop a working quantum computer within about 12 years. Other leading technology companies also are interested in quantum computers. For example, IBM earlier this year announced several quantum computing breakthroughs, including a larger quantum computing chip and two new ways of detecting and correcting for quantum errors.

Facebook's New Spam-Killer Hints at the Future of Coding
Wired (09/01/15) Cade Metz

Facebook's software engineers chose the relatively obscure Haskell programming language to serve as the platform for rebuilding the company's anti-spam system. Facebook software engineer Louis Brandy says Haskell is a perfect choice because it is extremely proficient at multiple and simultaneous task execution as well as providing engineers tools for coding the various tasks on the fly. "We want to run as many checks in the shortest amount of time, and that's where Haskell helps us," he notes. Haskell's prowess at concurrency is rooted in its nature as a "purely functional programming language," in which programs are built around a series of functions, with each function running independently of all the others. Haskell fulfilled Facebook's need for a language to help engineers craft "rules" for spotting spam on the social network, without being too concerned about their manner of execution. "If we're working in a purely functional language that we know doesn't have side effects, the faster we're able to move," says Instagram's John Edstrom. Facebook's strategy could point to the future of programming, as newer languages follow Haskell's design in that they enable developers to build massively parallel code at speed. Brandy also observes other projects are constructing Haskell-like software libraries for similar languages.

Self-Driving Golf Carts
MIT News (09/01/15) Larry Hardesty

The Singapore-MIT Alliance for Research and Technology (SMART) recently conducted an experiment in which self-driving golf carts carried 500 tourists around winding paths filled with pedestrians and bicyclists.  SMART also tested an online booking system that enabled visitors to schedule pickups and drop-offs at any of 10 locations scattered around the garden, automatically routing and redeploying the vehicles to accommodate all of the requests.  "We would like to use robot cars to make transportation available to everyone," says Massachusetts Institute of Technology professor Daniela Rus, one of the project's leaders. "The idea is, if you need a ride, you make a booking, maybe using your smartphone or maybe on the Internet, and the car just comes." After the experiment, 98 percent of participants said they would use the autonomous golf carts again, and 95 percent said they would more likely visit the gardens if the golf carts were a permanent fixture.  "We believe that if you have a simple suite of strategically placed sensors and augment that with reliable algorithms, you will get robust results that require less computation and have less of a chance to get confused by 'fusing sensors,' or situations where one sensor says one thing and another sensor says something different," Rus says. The researchers will describe their experiment at the International Conference on Intelligent Robots and Systems later this month.

Magnetic Fields Provide a New Way to Communicate Wirelessly
UCSD News (CA) (08/31/15) Liezel Labios

University of California, San Diego (UCSD) researchers have demonstrated a new wireless communication technique that works by sending magnetic signals through the human body. The researchers say their method could be developed into an ultra-low-power wireless system that can transmit information around the human body.  "We're trying to find new ways to communicate information around the human body that use much less power," says UCSD professor Patrick Mercier.  The technique relies on magnetic field human body communication, which uses the body as a vehicle to deliver magnetic energy between electronic devices.  The researchers showed the path losses associated with magnetic field human body communication are upwards of 10 million times lower than those associated with Bluetooth radios. In addition, lower power consumption leads to a longer battery life. "We hope to significantly reduce power consumption as well as how frequently users need to recharge their devices," says UCSD Ph.D. student Jiwoong Park. Magnetic field human body communication also makes the communication link less vulnerable to eavesdropping.  "Increased privacy is desirable when you're using your wearable devices to transmit information about your health," Park notes.

Tinder-Tinkering Artificial Intelligence Could Lessen Left-Swiping
UdeM News (09/01/15) Martin LaSalle

A University of Montreal researcher has developed an artificial intelligence program to better target the people that Tinder users might like.  Harm de Vries decided to program the software after being disappointed with the service's results.  He taught the software to recognize the women that he likes by extracting nearly 10,000 images from Tinder and processing them using algorithms.  De Vries determined his own preferences to establish the program's success rate. "I labeled all 10,000 images from Tinder. Eight thousand were used to train the program, and the rest were used to evaluate the performance of the program," he says. "The results of the first version were hardly better than chance, because it seems that a sample of 10,000 photos was too little, and because predicting attraction is more complex than a computer determining whether or not there's a person in the image." De Vries refined the analysis by applying deep-learning techniques, and after a few weeks his software was able to distinguish men from women in 500,000 photos retrieved from the OKCupid dating site with 93-percent accuracy.  In comparison, de Vries only achieved an accuracy rate of 95 percent when he personally undertook the task.  He then built the data from this learning analysis into the original program to test it again on recognizing Tinder photos he would like, realizing a success rate of 68 percent.

A Destined Duo: UI Professor to Develop Robots That Aid Elderly
Daily Illini (08/31/15) Emily Scott

University of Illinois (UI) researchers are working on the Automation Supporting Prolonged Independent Residence for the Elderly (ASPIRE) project, an interdisciplinary initiative to create mobile device-controlled robots and drones that can help senior citizens perform daily tasks.  The ASPIRE project recently received a $1.5-million grant from the U.S. National Science Foundation as part of its National Robotics Initiative.  The project aims to develop assistive robots with the flexibility, safety, and usability required to become normal parts of the homes of senior citizens.  "We want to figure out if a few things are moving around [and] how to make sure they never collide," says UI professor Naira Hovakimyan.  The researchers think virtual reality could help solve this problem, and they want to build a virtual-reality environment and collect data from experiments that will assist in the design of the robots.  The researchers anticipate having competition in the assistive technology sector.  "We are not going to be unique, we are not going to be different; we are going to have fun while developing it," Hovakimyan says.

4 Research Projects That Could Double Wireless Bandwidth
Network World (08/31/15) John Edwards

University of Texas at Austin (UT Austin) researchers have developed a magnetic-less radio wave circulator that aims to double wireless bandwidth.  "The final goal is to realize a compact device that can enable full-duplex for wireless communications, transmitting and receiving at the same time on the same frequency channel," says UT Austin professor Andrea Alu.  The circulator is free from dependence on magnetic effects, which enables it to offer a much smaller footprint and uses less-expensive materials than current circulators.  Meanwhile, University of Bristol researchers have created a new full duplex transceiver architecture that can estimate and cancel out interference from a user's own transmission, enabling a mobile device to simultaneously transmit and receive on the same channel.  "The transceiver combines electrical balance isolation and active radio frequency cancellation to suppress interference by a factor of over 100 million," says Bristol researcher Leo Laughlin.  Columbia University researchers have developed full-duplex radio integrated circuits that can be implemented in nanoscale complementary metal-oxide semiconductors to enable simultaneous transmission and reception on the same frequency.  Rice University researchers say they have developed an approach that relies on Multiple-Input Multiple-Output technology to achieve full duplex wireless communication simply and inexpensively by adding a tiny extra antenna and new software to future phone and other mobile communication devices.

Abstract News © Copyright 2015 INFORMATION, INC.
Powered by Information, Inc.

To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.
Non-Members: Unsubscribe