Welcome to the November 9, 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
Will Computer Vision Help France Take a Lead in AI?
ComputerWeekly.com (11/08/16) Pat Brans
A key factor for France emerging as a leading center for research and development (R&D) in artificial intelligence (AI) is overcoming public concerns and outdated government regulations. Last year, the Facebook Artificial Intelligence Research (Fair) lab was established in Paris to leverage French talent and the support of research institutions. One of Fair's research priorities is computer vision, and former Fair director Florent Perronnin says France is a strong leader in statistics-enabled computer vision R&D. He cites various institutes concentrating on computer vision, including Inria (the French Institute for Research in Computer Science and Automation), Xerox in Grenoble, Technicolor in Rennes, and Facebook in Paris. Skills endemic to the French population, such as machine learning, dovetail with computer vision. In addition, Tractica research director Aditya Kaul says computer vision is a chief AI application field, with already existing utility in image recognition/classification on mobile phones, emotion analysis and face recognition, and cancer detection. However, France's population must warm to AI in general and overcome fears of job displacement for it to prosper. Kaul notes there currently is more dialogue in the U.S. of AI "helping to augment humans, improving the quality and efficiency of work" than in France.
Big Data Shows People's Collective Behavior Follows Strong Periodic Patterns
University of Bristol News (11/08/16)
Big data-based analysis of massive datasets of modern and historical news, social media, and Wikipedia page views can reveal periodic patterns in collective population behavior that could otherwise be overlooked. By analyzing these sources in two studies, researchers from the ThinkBIG project at the U.K.'s University of Bristol found these patterns are not only more predictable than previously thought, but are usually only uncovered by analyzing a large population's activities for a prolonged period. "What emerges is a glimpse at the regularities in our behavior that are hidden behind the day-to-day variations in our lives," says ThinkBIG leader and Bristol professor Nello Cristianini. The first study examined 87 years of U.S. and U.K. newspapers between 1836 and 1922. The researchers found the weather and seasons strongly regulated people's leisure and work, with words such as picnic or excursion peaking every summer in both countries. The second study analyzed the aggregate sentiment in Twitter in the U.K. and aggregate Wikipedia access over four years, and found seasons also have a significant influence on mental health. Researchers observed an overexpression of negative sentiment in wintertime, while anxiety and anger are overexpressed between September and April.
DMU's 48-Hour Hackathon in Brazil Leads to Winning App
De Montfort University Leicester (11/09/16)
Brazil's University of Sao Paulo recently hosted the Hackathon for Health, Assisted Living, Social care, and Communities (HALE), a 48-hour hackathon to develop six prototype applications to increase awareness of respiratory diseases and how to manage them. Students and staff from De Montfort University Leicester (DMU) in the U.K. participated in the HALE, which was organized by DS-Cubed--a company managed by DMU's Samad Ahmadi and run by student interns and volunteers. The winning app, developed by five students from the University of Sao Paulo, along with Vitor Augusto Andrioli, who had studied at DMU, is a quiz game that teaches users about asthma and how to deal with it, as well as encouraging them to interact and share experiences. "At the end of the game, users will get a certificate stating that they are capable of basic care for asthmatic children, which could also enhance their employability," Andrioli says. "We had such a brilliant response from everyone at the University of Sao Paulo, who happily gave up their time and expertise to support our students," Ahmadi says. "As a result, not only have our students benefited significantly, but we are working with two universities on developing joint research proposals."
A Slow Ride Toward the Future of Public Transportation
The New York Times (11/04/16) Henry Fountain
A self-driving electric bus relying on laser sensors, global-positioning systems, and software is undergoing testing in Helsinki, Finland. The two-year, $1.2-million Sohjoa project, is backed by researchers at several universities with cooperation and money from government agencies and the European Union. The researchers say the proliferation of autonomous public transportation will reduce street congestion and greenhouse gas emissions. "A good possible outcome is that less and less people will own personal vehicles in the cities because they really don't need them anymore," says Sohjoa project coordinator Harri Santamala. The buses do not have a steering wheel, or brake or accelerator pedals. During testing a person is stationed on the bus, which holds up to 12 people, and they can hit an emergency "stop" button if necessary. Santamala notes the Sohjoa buses drive very slowly and also are restricted in terms of lateral movement due to safety concerns. The buses, which are not as sophisticated as those being developed by Google and other companies, are "taught" a route by having operators drive them using steering and acceleration controls on a small box. The routes are then further fine-tuned using software. Santamala says the goal is to establish a real bus route in the next two years.
Software Dreams Up New Molecules in Quest for Wonder Drugs
Technology Review (11/03/16) Tom Simonite
Researchers from Harvard University, the University of Toronto in Canada, and the University of Cambridge in the U.K. have developed software that suggests new molecular structures for medicines. The team says the system can devise structures more independently of humans and without lengthy simulations by leveraging its own experience, built up by training machine-learning algorithms with data on hundreds of thousands of drug-like molecules. "It explores more intuitively, using chemical knowledge it learned, like a chemist would," says Harvard professor Alan Aspuru-Guzik. The new system is based on deep learning, which has become common in computing companies but is less established in the natural sciences. Deep learning uses a design known as a generative model, which absorbs a trove of data and uses what it learns to generate plausible new data of its own. After training the generative model on 250,000 drug-like molecules, the researchers found it could generate plausible new structures by combining properties of existing drug compounds, and being asked to suggest new molecules that strongly displayed certain properties such as solubility, and being easy to synthesize. The researchers say the project suggests deep-learning software can internalize a kind of chemical knowledge, and use it to help scientists.
Machine-Learning Algorithm Quantifies Gender Bias in Astronomy
Nature (11/04/16) Inga Vesper
Researchers from the Swiss Federal Institute of Technology in Switzerland used machine learning to determine bias against women in citation rates in the field of astronomy. The researchers estimate that papers whose first authors are women receive about 10 percent fewer citations than those that are first-authored by men. "The novelty of this paper is in dispelling the myth that gender disparity in citation can be attributed to specifics of the paper, rather than to gender," says Indiana University Bloomington researcher Cassidy Sugimoto. The Swiss researchers analyzed 200,000 papers in five journals from 1950 to 2015, and first trained a machine-learning algorithm to accurately calculate the citations for each paper first-authored by a man using as many non-gender-related factors as possible. The team then used their algorithm on the papers with female-first authors, and found this set, taken from 1985 onward, had received about 6 percent fewer citations than their male-authored counterparts. However, the algorithm predicted the papers should have gotten 4 percent more citations than those authored by men, based on other factors. "This means women and men of equal quality will have unequal records," says Yale University researcher Meg Urry.
The Ocean's Robots May Soon Enjoy High-Speed Internet
Wired (11/03/16) Eric Niiler
A wireless communications network could soon enable autonomous underwater instruments to collect information, talk to each other, and transfer data at regular Internet speeds, thanks to the collaborative Sunrise project. Researchers are experimenting with boosting the bandwidth and speed of existing underwater acoustic modems, which transmit information through water with sound waves. The Sunrise team in Italy is working on devising a common language with which underwater drones can communicate with each other. Practice experiments for the new technologies are made difficult by the ocean's salinity and temperature differences, but the systems are getting sturdier, says Johns Hopkins University professor Louis Whitcomb. He recently completed an oceanographic expedition north of the Arctic Circle using underwater drones. The drones had to surface every few minutes and send data in 64-byte packets once per minute, with each package taking six seconds to transmit. "We're not going to be able to push HD TV over an acoustic modem anytime soon," Whitcomb says. "But we have a range of technologies and we will see the development of vehicles that will bridge these communications regimes."
Chinese Characters Are Futuristic and the Alphabet Is Old News
The Atlantic (11/01/16) Sarah Zhang
Stanford University professor Tom Mullaney believes Western countries are lagging behind China technologically because they still use the QWERTY keyboard, whereas computers make Chinese--with its 75,000 individual characters in place of an alphabet--far more advantageous. Mullaney says typing English via a QWERTY keyboard "doesn't make use of a computer's processing power and memory and the cheapening thereof." He notes there are now dozens of ways to input Chinese with software-augmented shortcuts. With the arrival of the computer in the 1970s and 1980s came a boom in Chinese input options, and experimentation with pronunciation-based systems using the QWERTY keyboard yielded software-driven predictive English letter/Chinese character translation systems. Tegic's U.S. introduction of the T9 predictive texting system in early cellphones offered much less input efficiency than China's deployment because it had to conform to the traditional keyboard letter arrangement. The Chinese version allowed fewer strokes per character, says Tegic co-founder William Valenti. "In China, the proliferation of mobile over the last 10 years has meant an explosion of new users learning to type for the first time on a mobile device," notes ethnographer Christina Xu. "They don't have any attachment or experience in QWERTY, so getting them to adopt new input methods is way easier."
Brown Researchers Developing New Interactive Sleep App
News from Brown (11/01/16) Kevin Stacey
Brown University researchers developed SleepCoacher, a cellphone app using sleep analytics to generate personalized recommendations informed by scientific literature. SleepCoacher guides users through a self-experimentation framework to help them find the recommendations that best work for them. The researchers used SleepCoacher in two pilot studies, which found 80 percent of people who followed the recommendations at least 60 percent of the time reported improvement in their sleep. "This could be personalized for whether you are a night owl or morning person, a light or heavy sleeper, or even someone who needs more than the usual eight hours of sleep," says Brown professor Jeff Huang. As part of the two pilot studies, participants entered a rating of how refreshed they felt in the morning, and noted other factors that might have affected their sleep. The SleepCoacher algorithm used that data to determine what factors were correlated with three key sleep outcomes: how long it took participants to fall asleep, how many times they woke up during the night, and how refreshed they reported feeling in the morning. When a strong correlation is made, the algorithm generates a recommendation based on a collection of 117 recommendation templates developed with psychologists and psychiatrists.
New U.S. Robotics Roadmap Calls for Regulation, Research, and Education
UC San Diego News Center (11/01/16) Ioana Patringenaru
The U.S. Robotics Roadmap released last week advocates for new policy frameworks to regulate emerging technologies, increased research in the field of human-robot interaction, and improved science and technology education. Although the field of autonomous vehicles is advancing quickly, self-driving cars are faced with regulatory and operational hurdles. The roadmap says before the technology can be deployed widely, autonomous vehicles will need to reach a suitable level of autonomy and reliability and require little to no human intervention. The report also discusses the trajectory of healthcare and companion robots as the population ages, and notes companion robots will need to be equipped with better navigation systems and user-friendly interfaces, while easy-to-use interfaces also will be a necessity for manufacturing robots. The report stresses robots' dexterity must improve dramatically to be efficient in manufacturing. As robotic systems continue to permeate everyday life, the workforce also will need to know how to interact with new technologies. Training needs to happen at all levels, but education will be crucial for students in kindergarten through high school. "We need to realize that most of the interfaces we design today for robotic systems aren't easy to use," says University of California, San Diego professor Henrik I. Christensen.
MIT Researchers Are Working to Create Neural Networks That Are No Longer Black Boxes
Digital Trends (10/31/16) Luke Dormehl
Researchers at the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory have done preliminary work to demonstrate the possibility of training deep-learning neural networks so they not only offer predictions and classifications, but also justify their decisions. The team studied neural nets that were trained on textual data, and split the network into two modules: one which extracted segments of text and scored them on their length and coherence, and another that performed prediction and classification. The system was tested on a dataset comprised of reviews from a website in which users rated beers. The system corresponded with human annotations by 96 percent and 95 percent, respectively, when anticipating ratings of beer appearance and aroma, and 80 percent when predicting palate. "The question of justifying predictions will be a prevalent issue across complex [artificial intelligence] systems," says MIT professor Tommi Jaakkola. "They need to be able to communicate with people. Whether the solution is this particular architecture or not remains to be seen. Right now, we're in the process of revising this work and making it more sophisticated."
Engineer Gets Grant to Boost Security in Computer Chip Production
UT Dallas News Center (10/28/16) Miguel Perez
The U.S. National Science Foundation and Semiconductor Research Corp. have awarded a University of Texas at Dallas (UT Dallas) engineer and colleagues a $480,000 grant to pursue research designed to improve the security of computer chip production. Since the early 2000s, the semiconductor industry has steadily increased the amount of computer chips produced in outsourced plants. UT Dallas professor Jeyavijayan Rajendran and colleagues will develop new techniques in the physical design of integrated circuits to improve the security of a method known as split manufacturing. Security researchers have concerns about the method because attackers can still use its automated processes to their advantage. For example, an attacker at a foundry can insert hardware Trojans into products, and an attacker also can create unauthorized chips. "To overcome this security vulnerability in split manufacturing, we developed an automated tool that ensures security by design," Rajendran says. "This defense improves the security of split manufacturing by deceiving the [Front End of Line] attacker into making wrong connections." His team is working to mimic potential attacks by developing models that can determine a design's missing connections, just as an attacker would with the limited design information given to the untrusted foundry.
Statistics in the Era of Big Data
Stanford News (11/07/16) Taylor Kubota
In an interview, Stanford University professor Susan Holmes discusses the role of statistics in big data initiatives, noting "the data comes first. But it's also about trying to make the data as good as possible." Holmes says her primary area of interest is sophisticated types of geometry and generating their visualizations via software so people can easily and quickly re-interact with their data. "We have complicated data and the challenge is getting to the most useful visual summaries that will tell you something about that data," she says. Holmes notes this requires tools, specifically an open source software system called R and a digital repository, which facilitate reproducible research. Holmes says her specialty involves "taking a large amount of variables and lots of measurements and all kinds of information and trying to make it so that we can have pictures of the data that point out the underlying structure." She cites the computer as a major driver of change in the statistics field, to the point where "our colleagues expect us to find solutions that are easier to interpret. There is less mathematics and more visualization and computing." Holmes says these developments have opened up statisticians' engagement with many fields.
Abstract News © Copyright 2016 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: firstname.lastname@example.org