Welcome to the October 5, 2015 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
Quantum Information Workshop Draws Experts From Around the World
Diamondback (MD) (10/04/15) Michael Brice-Saddler
Leading experts in quantum information convened last week at the University of Maryland's (UMD) Joint Center for Quantum Information and Computer Science to participate in a workshop to discuss the newest developments in their field. "We want people to learn about what's going on and encourage them to start new projects," says center co-director and UMD professor Andrew Childs. He notes among the topics covered in the workshop were quantum cryptography, quantum computing, communication protocols, and quantum complexity theories. "To have a chance to take these world experts on the theory side and show them how things work in the lab...it was a really beautiful chance to show them the practical side of what they're designing and thinking about," says center co-director Jacob Taylor. He also says one of the expected results from the workshop is identifying subjects that smaller, more specialized workshops can concentrate on in coming years. "With every workshop like this, you're hoping for people to learn about the latest developments in the field so they can build upon them and so that it will lead to new collaborations," says U.S. National Institute of Standards and Technology physicist and center fellow Stephen Jordan.
A Manifesto for Algorithms in the Environment
The Guardian (10/05/15) Victor Galaz
Stockholm University researcher Victor Galaz and colleagues outline precepts for an in-progress Biosphere Code Manifesto, a recommendation for using algorithms borne out of growing awareness that they so deeply permeate our technology "they consistently and subtly shape human behavior and our influence on the world's landscapes, oceans, air, and ecosystems." The manifesto stipulates those implementing and using algorithms should weigh those programs' effects, and algorithms should be designed to consider human needs and the biosphere while also enabling transformations toward sustainability via the support of ecologically responsible innovation. A third principle calls for fair distribution of algorithms' benefits and risks, and a fourth principle advises making them flexible, adaptable, and context-aware in the event serious impacts or unforeseen outcomes emerge. Another precept is for algorithms to be used in such a manner as to augment people's capacity to deal with unexpected results, such as problems caused by errors or misbehaviors in other algorithms. Keeping data collection transparent and meaningful, and validating the datasets that feed into algorithms, is the sixth principle. The final outlined principle is to use algorithms to improve human creativity and playfulness, and to produce new kinds of art. "We should encourage algorithms that facilitate human collaboration, interaction, and engagement--with each other, with society, and with nature," Galaz says.
Study Rates UW CSE Software and Engineering Research Most Practically Relevant of the Past Five Years
University of Washington News and Information (10/01/15)
A tool developed by University of Washington (UW) researchers to improve collaboration between software developers has been judged the most practically relevant software engineering research of the last five years. The recognition comes from an industrial relevance study conducted by Microsoft Research and Singapore Management University, which asked more than 500 software developers to rate the relevance to their daily work of 571 research papers. The greatest number of respondents rated the UW project, which generated the Crystal collaboration tool, as an "essential" addition to the practice of software development. The UW research team, led by professors Michael Ernst and the late David Notkin, developed Crystal as a way to help developers who are working on a team in parallel avoid making changes that might be in conflict with each other. Crystal does this by continuously merging every developer's changes into the software so conflicts become apparent and can be quickly addressed. Crystal prevents wasting time returning to the code to rectify conflicts and problems after the fact. The paper on proactive conflict detection was part of the speculative analysis project, led by Ernst at UW's Programming Languages & Software Engineering group.
IQ Test Result: Advanced AI Machine Matches Four-Year-Old Child's Score
Technology Review (10/01/15)
Scientists led by the University of Illinois' Stellan Ohlsson tested the intelligence quotient (IQ) of the Massachusetts Institute of Technology's ConceptNet 4 artificial intelligence (AI) machine using a verbal IQ test designed for children. The test is designed to quantify children's performance in the categories of information, vocabulary, word reasoning, comprehension, and similarities. The test questions were fed to ConceptNet 4 with some modification so they could interface with the computer's structure of knowledge about the world. "ConceptNet does well on Vocabulary and Similarities, middling on Information, and poorly on Word Reasoning and Comprehension," the researchers say. They note the answers the machine yielded were very sensitive to its interpretation of the question. The researchers also observed the system tended to perform better when it was forced to only consider single concepts. Ohlsson and his colleagues found ConceptNet 4's performance on the IQ test was equal to that of an average four-year-old child, but below that of average five- to-seven-year-olds. A likely factor in this performance is ConceptNet 4's basis on knowledge gathering, instead of being learning-driven like more modern AI systems.
Rise of Concerns About AI: Reflections and Directions
CCC Blog (10/01/15) Thomas G. Dietterich; Eric J. Horvitz
In the latest issue of the Communications of the ACM, Tom Dietterich and Eric Horvitz, respectively the current and former president of the Association for the Advancement of Artificial Intelligence, discuss the state of artificial intelligence (AI) research and the recent rise in anxiety about AI. Although Dietterich and Horvitz conclude current fears about runaway AI surpassing or destroying humanity are, at best, distant concerns considering the current state of AI, they do highlight several more realistic concerns that should be given more attention. First, they caution AI systems, like all software, are vulnerable to bugs, and it is important to find ways of ensuring the reliability of AI software. Similarly, AI developers should think about cybersecurity issues, and find ways to protect their AI software from abuse by hackers. Dietterich and Horvitz also say attention must be paid to the potential for "Sorcerer's Apprentice" situations, in which limits to an AI's ability to understand human instructions could result in unintended, and potentially dangerous, outcomes. The fourth area of concern comes from the area of shared autonomy, in which humans working hand-in-hand with AI systems can lead to unforeseen dangers and complications. Finally, Diettrich and Horvitz say AI researchers need to work together with colleagues across multiple disciplines to better understand the potential socio-economic impact of AI technology.
Got 'Em! Researchers Steal Crypto Keys From Amazon Cloud
InfoWorld (09/30/15) Fahmida Y. Rashid
Worchester Polytechnic Institute (WPI) researchers have demonstrated how to use one instance of Amazon EC2 to recover the full 2,048-bit RSA key from a separate Amazon instance. "We exploit the [last-level cache (LLC)] to recover the secret key of a modern sliding-window exponentiation-based implementation of RSA, across cores and without relying on deduplication," the researchers say. They note malicious hackers could use this strategy to intercept the targeted entity's encrypted communications and extract potentially valuable information. For this attack to work, both the attacker's Amazon account and the target Amazon account containing the private RSA key must be on the same hardware chip or chip set. "Everything must work in concert together and it is highly difficult to pull off," notes Comodo's Robin Alden. The researchers say their technique highlights the need for deploying stronger isolation techniques in public clouds. Experts recommend providers patch the weaknesses that make these types of attacks possible, and smarter cache management policies for hardware and software could prevent side-channel leakages and future exploits. "A more random placement policy would make it tougher for attackers to land on the same [central processing unit] or hardware as that of the intended target," says Ciphercloud's Sundaram Lakshmanan.
More-Flexible Machine Learning
MIT News (10/01/15) Larry Hardesty
Massachusetts Institute of Technology (MIT) researchers at the Annual Conference on Neural Information Processing Systems in December will present a new machine-learning technique that enables semantically-related concepts to reinforce each other. "Because there are actually semantic similarities between those categories, we develop a way of making use of that semantic similarity to sort of borrow data from close categories to train the model," says MIT graduate student Chiyuan Zhang. The researchers quantified the notion of semantic similarity using an algorithm that mined Flickr images for identifying tags that tended to co-occur. Semantic similarity of two words was a function of the co-occurrence's frequency. In essence, the researchers say the system gives the algorithm partial credit for incorrect tags that semantically correspond with the correct tags, and the Wasserstein distance metric is employed to handle the calculations. Experimentation showed a machine-learning algorithm trained with this method predicted the tags human users applied to images on the Flickr website with better accuracy than it did with a conventional training approach. "I think this work is very innovative because it uses the Wasserstein distance directly as a way to design learning machines," says University of Kyoto researcher Marco Cuturi.
Activist Bots Recruit Humans to Their Cause on Twitter
New Scientist (09/30/15) Aviva Rutkin
The Botivist program uses Twitter to rally people to social causes, thanks to the efforts of Saiph Savage and colleagues at West Virginia University. The team's eight activist bots were designed to concentrate on corruption in Latin American governments. When Botivist detects someone tweeting about the issue, a bot sends them an inspirational quote or call to action. The bots provoked replies from 175 people out of more than 1,000 people to whom they sent inquiries. Eighty-one percent of that segment responded to the bots' call to contribute ideas. "One of the biggest benefits of activism is that it can guide citizens to take action," Savage says. "This is taking people that were complaining about corruption, now they're thinking about things they can do to change it." Savage's team is currently working on a public Botivist iteration, and also planning another round of Twitter queries to determine how far the bots can inspire people.
Picture This: An App for Blind Photographers
UCSC NewsCenter (09/30/15) Peggy Townsend
University of California, Santa Cruz (UCSC) researcher Justin Adams has developed VizSnap, a smartphone application that enables visually-impaired users to recognize their photos and to organize and share them through social media. VizSnap lets a user make an audio recording of the environment in which the photo is being taken in order to help identify the shot. The app also records the data, time, and location, while enabling the photographer to add a voice memo about the captured image. The app won first prize in the design competition at the MobileHCI conference two years ago, as well as second prize at the Big Ideas @ Berkeley competition. Adams is still improving the app and is looking for more users for a long-term study with the technology. Meanwhile, other UCSC researchers are developing a video game to help children with corrected cleft palates and lips learn to speak more clearly. The researchers also are developing an app that enables those with visual impairments to independently format documents, and a special glove that can ease the symptoms of those with hand tremors.
Identifying Problems With National Identifiers: Supposedly Encrypted Numbers Can Be Easily Decrypted
Harvard University (09/29/15)
Harvard University researchers have used a pair of experiments to show Resident Registration Numbers (RRNs) used in South Korea can be decrypted to reveal a range of personal information. In the experiments, the researchers were able to decrypt more than 23,000 RRNs using both computation and logical reasoning. The findings suggest that although such identifiers are encrypted to protect privacy, they remain vulnerable to attack and must be designed to avoid such weaknesses. The researchers showed each number in the RRN could be replaced with a letter in a recognizable pattern, which could then be used to decrypt thousands of RRNs, which could reveal personal information about their users. They also found the final RRN digit is a weighted sum of prior digits, meaning it is possible to decrypt the numbers and then use arithmetic to confirm the accuracy of the information. "Our study shows that weak encoding systems, which refer to the very design of the number, render encryptions as poor methods of protecting privacy," the researchers note. The findings are timely, because South Korea is currently debating a redesign of RRNs and other nations, including the U.S., have discussed the use of a single identifier for medical records, according to Harvard professor Latanya Sweeney.
Researchers Develop Deep-Learning Method to Predict Daily Activities
Georgia Tech News Center (09/28/15) Tara La Bouff
Researchers from the Georgia Institute of Technology's (Georgia Tech) School of Interactive Computing and the Institute for Robotics and Intelligent Machines have developed a new method to train computers to recognize and comprehend a wide range of human activities in a single day. More than 40,000 photographs were captured every 30 to 60 seconds over six months by a wearable camera, and this was fed to the computer so it could learn to categorize images across 19 activity classes. The participant wearing the camera could review and annotate the photos at the end of each day to ensure their correct categorization. The system predicted with 83-percent accuracy in which activity that person was engaged. The research team believes it has accumulated the largest annotated dataset of first-person images to demonstrate deep learning can understand human behavior and the habits of a specific individual. "This work is moving toward full activity intelligence," says Georgia Tech researcher Edison Thomaz. "At a technical level, we are showing that it's becoming possible for computer-vision techniques alone to be used for this." Thomaz says the research could potentially contribute to the development of improved personal assistant applications, as well as help researchers explain connections between health and behavior.
On What Facebook Knows--An Interview With the Man Behind Facebook's Personality Experiment
Social Media Today (10/01/15) Andrew Hutchinson
Earlier this year, researchers from the University of Cambridge and Stanford University released a report detailing how people's Facebook activity could be used to measure their psychological profile with surprising accuracy. In an interview, Michal Kosinski, one of the co-lead researchers responsible for the study, discusses its results and implications. The researchers' experiment involved developing an app that administered a psychological questionnaire and also tracked Facebook activity. About 86,000 participants used the app, and Kosinski says many of the findings were surprising. Facebook likes could, for example, predict whether or not a given user's parents were divorced, their smoking and drinking habits, and sexual orientation, as well as religious and political views. "Actually, everything we tried predicting was predictable, to a degree, and quite often it was very accurate," Kosinski says. However, he notes the most profound insights required multiple datapoints. Although knowing how many times a person logged into Facebook could be revealing, Kosinski says that data was only useful when it was accompanied by other data, which enables "predictive systems to establish very accurate profiles of who you are." He notes his experience with the study has not changed his own personal use of Facebook or social media, although he does think there are ways companies could give users more direct control over what information is being gathered about them.
Abstract News © Copyright 2015 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.