Association for Computing Machinery
Welcome to the January 4, 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).


What Happens When Virtual Reality Gets Too Real
The Wall Street Journal (01/03/16) Jack Nicas; Deepa Seetharaman

Researchers at Stanford University's Virtual Human Interaction Lab have been studying virtual reality (VR) technology for the past 13 years, and lab director and professor Jeremy Bailenson is convinced VR experiences can change users' thinking and behavior. Researchers say the ability of VR to support more lifelike experiences than TV and other technologies, and make users active instead of passive participants, could impact their outlook and perspectives to the degree it presents an issue for the industry as it grows. One test in the Stanford lab determined test subjects who had just cut down a virtual tree used fewer paper towels when cleaning up a spill after the experience than subjects who had not felled a tree. In another example, French students developed a harrowing VR simulation of being in one of the World Trade Center towers on Sept. 11, 2001. Some developers aim to tap VR's lifelike nature to help users overcome phobias. Meanwhile, long-term disconnection could become especially relevant if users start spending prolonged periods in VR. As the technology matures, "you're going to have this alternate reality...that you start spending more and more time in," says Facebook's Brendan Iribe. "At least that's what science fiction says we're going to do."

5 Robot Trends to Watch for in 2016
Technology Review (01/01/16) Will Knight

The Chinese government wants to fill its factories with advanced manufacturing robots in an attempt to help the country retain its vast manufacturing industry. The project will require robots that are significantly more advanced and cost-efficient. For example, Guangdong province already has said it will invest $154 billion in installing robots, and new techniques and algorithms are enabling robots to learn much more quickly and effectively. Deep learning, which relies on large simulated neural networks, has proven invaluable for training robots to understand the content of images, video, and audio, and it could have a big impact in industrial robots in the near future. Another trend to watch for in 2016 is robots sharing the knowledge they have acquired with other robots, which could accelerate the learning process. However, giving a robot an engaging personal touch is difficult, and even in limited scenarios these robots will need to be designed and programmed very carefully in order to be well received by the public. Finally, 2016 is likely to be the year autonomous drones reach their potential. The U.S. Federal Aviation Administration released regulations for registering drones at the end of 2015, and the agency is testing technology that could help automate air traffic control for automated vehicles.

'Human Computation' Could Save the World Without the Risks of AI
Motherboard (12/31/15) Victoria Turk

Several scientists are urging human-machine collaboration via systems that integrate computer and human capabilities to solve the most pressing problems of today's world without the existential uncertainty of pure artificial intelligence. The authors of an editorial in the journal Science, Human Computation Institute director Pietro Michelucci and Cornell University researcher Janis Dickinson, envision a system that supplies a technical framework for ideas to be shared, analyzed, and revised until the best come to the surface. The researchers say the goal is to develop our understanding of real-world issues online, evaluate potential solutions in this computational space, and then apply new knowledge back in the real world. The basis of human computation is the fact that computers and humans bring unique proficiencies to collaboration, with the former offering rapid analysis of immense volumes of data and the latter contributing unmatched conceptualization of new ideas. Michelucci and Dickinson's model is designed to serve as a simulation of real-world scenarios, with a Wikipedia-like scheme in which millions of individuals can contribute, in combination with algorithms to create a feedback loop that promotes constant idea assessment and revision. The researchers say human computational systems already can be supported by existing technical infrastructure.

Why Some Colleges Are Better Than Others at Getting Women Into STEM Careers
MarketWatch (12/30/15) Jillian Berman

Some U.S. colleges are better at attracting more women to science, technology, engineering, and math (STEM) fields than others, according to a new analysis. A study of government data conducted by Brookings Institution fellow Jonathan Rothwell cites ways colleges can make STEM careers more appealing to women. One strategy Rothwell uncovered is to offer programs to all students that may be of special benefit to women. A more direct approach is to sponsor research focused on gender equality in business. A program at Dartmouth College specifically targets female students by giving freshman women an opportunity to work for professors in labs. Employers also are seeking new, more diverse employees through recruiting programs at certain schools, while other college programs may train women to cope with and overcome workforce discrimination. The American Society for Engineering Education estimated women received only 19.9 percent of the engineering degrees awarded in the U.S. last year. "If it continues to be the white men who are doing the best coming out of colleges then to some extent higher education is failing in its fundamental mission to create opportunity for anyone who is willing to work hard," warns Institute for Women's Policy Research executive director Barbara Gault.

How Software Developers Helped End the Ebola Epidemic in Sierra Leone
The Guardian (12/30/15) Bethany Horne

A team of open source software developers helped solve the urgent problem of distributing wages to healthcare workers on the front line fighting the Ebola epidemic in Sierra Leone. Sierra Leone received millions of dollars from international sources and it was not clear how it would be distributed, considering payroll was handled in cash. The developer team drew on existing open source software solutions for payroll management, biometrics, logistics, and accounting. The developers cannibalized existing voter registration machines to create a payroll enrollment scheme. They could not use fingerprint biometrics because it would have created a cross-contaminating risk, so they used open source facial-recognition software called OpenBR to enroll healthcare workers. They also developed a mobile money system, which substituted cellphone minutes for cash, and created an automated payment system. The team built the core system in two weeks, and NetHope's Emerson Tan says the people were paid on time. He believes the developers' efforts helped restore faith in Sierra Leone's healthcare system.

App Differentiates a Baby's Crying Sounds
Reuters (12/30/15)

Researchers at National Taiwan University Hospital (NTUH) have developed the Infant Cries Translator, an app that can differentiate between four separate crying sounds made by babies. The researchers spent two years collecting about 200,000 crying sounds from approximately 100 newborn babies, and uploaded them to an online database. An analysis of the frequency of the individual screams helped the researchers distinguish subtle differences in acoustics. The app displays an analysis of a baby's cries on the user's phone within 15 seconds, with an accuracy of 92 percent for infants under two weeks old, helping inform parents when their child is hungry, sleepy, in pain, or has a wet diaper. The analysis becomes less accurate as the baby gets older. The app constantly updates its database to a cloud drive, and its machine-learning algorithm enables parents to set up a personal setting for their infant, based on the parents' feedback. "Once the baby cries, we only need to press the recording button for 10 seconds, and the sound will be uploaded to the cloud drive," says NTUH researcher Chang Chuan-yu.

Hate Exams? Now a Computer Can Grade You by Watching You Learn
New Scientist (12/30/15) Aviva Rutkin

Researchers at Stanford University and Google have developed an algorithm that can improve a user's knowledge and do away with formal tests by analyzing students' performance on past practice problems, identifying where they tend to go wrong, and developing a picture of their overall knowledge. The researchers tested the system on more than 1.4 million student answers to math problems set on the Khan Academy online learning platform, and the corresponding scores. The researchers also trained a neural network to sort the questions by type. The system used all of the data to learn each student's capabilities on each question type. The researchers found the model could predict with up to 85-percent accuracy whether a student would get a new problem right or wrong by examining a few dozen other questions they had already answered. "Our intuition tells us if you pay enough attention to what a student did as they were learning, you wouldn't need to have them sit down and do a test," says Stanford researcher Chris Piech. The algorithm is a significant advance, according to University of Colorado, Boulder professor Tamara Sumner. "What is particularly impressive is that this approach does not require significant human input to annotate training data or hand-craft models of expertise," Sumner notes.

Robot Workers Revealed as Binary in More Ways Than One
Financial Times (12/30/15) Emma Jacobs

Assigning gender to robots carries the danger of reinforcing stereotypes in the workforce, according to University of Bielefeld research. The experiment showed two robots to a group of men and women, with each robot identical apart from distinctions in hairstyle and lip shape. The robot with longer hair and fuller lips was perceived as female while the one with short hair was seen as male. The male robot was considered to be more proficient at "masculine" tasks such as guarding a home or repairing devices, while the female robot was regarded as suitable for stereotypical jobs such as caring and housekeeping. These unconscious gender biases also can be expressed by algorithms, says University of Southampton professor (and past ACM president) Wendy Hall. For example, a study of Google's algorithms determined the search engine was more likely to show male job seekers ads for high-paying executive positions than female seekers. Meanwhile, author Martin Ford thinks industries in which men tend to work--such as finance, warehouse, and manufacturing--are especially vulnerable to automation. "In the near-term, automation seems likely to fall more heavily on men," he says.
View Full Article - May Require Paid Subscription | Return to Headlines | Share Facebook  LinkedIn  Twitter 

Power Problems Threaten to Strangle Exascale Computing
IEEE Spectrum (12/30/15) Jeremy Hsu

Exascale computers currently can only be built at astronomical cost, but Sandia National Laboratories' Erik DeBenedictis is working with the IEEE Rebooting Computing initiative and the International Technology Roadmap for Semiconductors to make exascale much more practical via innovations in three technologies. The first technology is the millivolt switch, which computer engineer Eli Yablonovitch is developing via the U.S. National Science Foundation's Center for Energy Efficient Electronics Science. Millivolt candidates he is working on include a tunnel field-effect transistor, power-conserving nanoelectromechanical switches, nanophotonics, and nanomagnetic-based logic circuits. A second potential technology is three-dimensional stacked memory chips, and Tezzaron Semiconductor's Robert Patti points to a possible migration to nonvolatile resistive random-access memory to support exascale supercomputing. The third technological solution is a specialized processor architecture, in which each part of the chip would operate at full power only when its specific function is required. The drawback to this approach is that it could limit what a particular supercomputer could execute efficiently. "If you want something else, you will have to replace the chip with one of a new design," DeBenedictis says.

Machines, Lost in Translation: The Dream of Universal Understanding
National Public Radio (12/24/15) Anne Li

A universal translator remains an elusive goal more than 60 years after the creation of one was first undertaken, and expert opinion varies on how soon one will be delivered. Microsoft Research's Vikram Dendi thinks this milestone is imminent thanks to achievements such as the Skype translator, which renders video chat speech as spoken or written translations in up to seven languages. The current method scientists use in advancing machine translation is the neural network, in which machines are trained to mimic people's thought processes. Neural networks are designed to convert each word into a simple vector, building up their accuracy as they attempt more translations. University of Montreal professor Yoshua Bengio believes the neural network technique holds more promise of supporting human-level performance by concentrating on the meaning of words. The manual input approach for teaching computers to translate between languages quickly became onerous, so in the 1980s a statistical-based model was explored in which machines were fed human-translated content so they could infer language rules and patterns themselves. Neural networks improved on this principle, and today machines can glean more information about each word and conduct better probability analysis to avoid unnaturally sounding translations.

New Questions About Many Queries for Quantum Computing
University of Waterloo (12/23/15)

Researchers in Canada have raised new concerns about the advantages of the bucket brigade model for algorithms using super-polynomial oracle queries, such as Grover's quantum searching algorithm. One of the main advantages is that it requires exponentially less active, potentially noisy processes to complete queries, allowing for more efficient energy consumption and more robust implementation. "The bucket brigade model...activates routing nodes only along the active path in the computation, it reduces the chances of error, and increases the speedup," says Vlad Gheorghiu, a postdoctoral fellow at the University of Waterloo's Institute for Quantum Computing (IQC). "It works well for algorithms that make a relatively small number of queries [i.e., polynomial], some of which may be used in quantum machine learning." A team from the IQC used the simplest form of noise, a bit flip error model, to test its robustness. The researchers found a greater number of noisy queries and the need for active quantum error correction offset the main benefits of the model approach. They observed for algorithms that make a smaller number of queries such as quantum machine learning, the scheme with a polynomially small error rate is still useful.

Robots to Help Immigrant Children Learn German
Bielefeld University (12/22/15)

A European Union-funded consortium encompassing five universities and two companies plans to launch this month a three-year project known as Second Language Tutoring Using Social Robots (L2TOR). The effort seeks methodologies to teach young immigrant children a second language using interactive robots, according to Stefan Kopp at Bielefeld University's Cluster of Excellence Cognitive Interaction Technology. In collaboration with local day-care centers, Kopp's team is testing robots' effectiveness as tutors. One application of their work is a humanoid Nao robot integrated with a tablet PC to lead children through a language course. As it teaches, the robot pays attention to the children's speech, expressions, and gestures so it can determine when pupils are having trouble understanding the lessons. "It is important that the robot recognizes how the child being taught feels, and whether he or she is frustrated or confused, for example," notes Bielefeld researcher Kirsten Bergmann. "We program the robot so that it can shape its interaction with the child so that he or she is being supported in the best way possible." Kopp says most day-care centers usually cannot offer individualized instruction in a second language, and the Bielefeld project aims to complement their existing resources.

Abstract News © Copyright 2016 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