Association for Computing Machinery
Welcome to the January 25, 2017 edition of ACM TechNews, providing timely information for IT professionals three times a week.

ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).


HiPEAC Vision Report Advocates Reinvention of Computing
Inside HPC (01/23/17)

The European Network on High Performance and Embedded Architecture and Compilation (HiPEAC) on Monday published a report calling for "the reinvention of computing." The HiPEAC Vision 2017's authors contend the arrival of a "Centaur Era," in which the boundaries between the traditional roles of man and machine will start to erode, is upon us. "The evolution from desktop PC will not stop at smartphone and tablet: the devices and systems that will allow us to automate key infrastructures, such as transport, power grids, and monitoring of medical conditions, are bringing us into the age of artificial intelligence," says Koen De Bosschere, a professor at Ghent University in Belgium and coordinator of the HiPEAC network. Vision lead editor Marc Duranton at France's Commission for Atomic Energy and Alternative Energies (CEA) says a crossroads has been reached, with current computers and software reaching the limits of what they can achieve in a dynamic environment. "What we now describe as "cyber-physical systems...entangle the cyber and physical worlds," he notes. "The increasing use of the artificial intelligence required to make them work means that we, the humans, really need to invent solutions so that we can trust systems and develop better ways to cope with the challenges of safety, security, privacy, energy efficiency, and increasing complexity."

Mummy Visualization Impresses in Computer Journal
Linkoping University (01/23/17) Monica Westman Svenselius

Researchers at Linkoping University in Sweden have developed visualization technology that can be used by visitors to the British Museum to reveal the murder of the mummified Geberlein Man 5,500 years ago. "It was challenging to obtain sufficiently high performance of the visualization such that visitors can interact with the table in real time, without experiencing delays," says Linkoping professor Anders Ynnerman. He notes several thousand images of the mummy taken by computer tomography (CT) are stored in the visualization table. Rapid graphics processors can then create three-dimensional volumetric images in real time to display what the visitors want to see. The degree of reflection and absorption of the x-rays by the mummy is recorded by the CT scanner and converted with the aid of a specially developed transfer function to different colors and degrees of transparency. "Sixty times each second, virtual beams, one for each pixel on the screen, are projected through the dataset and a color contribution for each is determined," says Linkoping's Patric Ljung. "We use the latest type of graphics processor, the type that is used in gaming computers."

Big Brother Will Have Some Difficulty 'Watching You' in Future
University of the Witwatersrand (South Africa) (01/24/17)

Real-time error correction in quantum communication is possible, according to researchers at the University of the Witwatersrand, Johannesburg in South Africa. Their research demonstrates that sometimes nature has difficulty distinguishing between specific kinds of laser beams and quantum entangled photons, a discovery that should lead to the creation of more secure long-distance quantum communication links. "In the quantum world...waves sometimes look like particles, particles like waves, and measurements change the properties of the very thing you are trying to measure," notes Witwatersrand professor Andrew Forbes. He says the gray area where the classical and quantum are indistinguishable creates the possibility of conducting experiments with "classically entangled" light. The team showed preparing and transmitting a classically entangled beam was the same as sending a quantum state. This enables reversal of the observed quantum entanglement decay caused by noise in the link, clearing a path for advances in secure quantum links in fiber and free space. "We exploited a particular type of laser beam, called vector beams, that have the property of being non-separable and sometimes called 'classically entangled,'" says lead study author Bienvenu Ndagano. The researchers say their breakthrough means long-distance quantum links can be established and tested with classically entangled light, with all error-correction measurements performed in real time without disrupting the quantum information.

DeepMind's AI Experts Have Pledged to Pass on Their Knowledge to Students at UCL
Business Insider (01/24/17) Sam Shead

University College London (UCL) students beginning this month will have the opportunity to learn from Google's DeepMind researchers, as senior DeepMind staff aim to pass on their knowledge to students enrolled in UCL's machine learning master's programs. The course will be overseen by DeepMind researcher and UCL professor Thore Graepel, and will focus on areas such as deep learning, reinforcement learning, and natural-language understanding. "This new partnership in state-of-the-art (artificial intelligence) AI is an excellent example of research-led teaching, for which UCL is renowned," says UCL professor John Shawe-Taylor. DeepMind staff also teach students at Oxford University, with the company launching its Deep Learning for Natural Language Processing advanced course in the university's Department of Computer Science this January. "We see the links between company research labs and academia as central to the future of AI," says DeepMind cofounder Demis Hassabis. "By continuing to share talent, expertise, and breakthroughs--not just on technical subjects, but also on the broader set of questions around ethics, safety, and societal impact--we believe we'll all make better progress in the development of artificial intelligence and its application for positive social benefit."

Trump Likely to See the Birth of an Exascale System
Computerworld (01/20/17) Patrick Thibodeau

An exascale supercomputer will likely be realized within the Trump administration's first term, which could be a tipping point for the U.S. Supercomputing is viewed as essential to national competitiveness because of the increasingly virtual nature of research and product development. Europe has set an exascale delivery schedule of 2022 along with a $749-million commitment, while both China and Japan aim to have a system ready by 2020. China is using its own microchips, while a European system in development uses ARM processors. The Obama administration initially set a 2023-2024 target date for exascale, but amended it in its final weeks to 2021, with a projected budget of $3.1 billion to $5.7 billion. Argonne National Laboratory's Paul Messina says the U.S. Department of Energy's Exascale Computing Project "is now a seven-year project, not a 10-year project, but it will cost more." China currently has the world's fastest supercomputer, running at about 125 petaflops. Although the U.S. exascale project's goals include contributing to the country's economic competitiveness and supporting national security, another objective is developing a software stack, in collaboration with vendors, that smaller systems in industry and academia can utilize.

Your (Social Media) Votes Matter
Notre Dame News (01/24/17) Jessica Sieff

A study by University of Notre Dame professor Tim Weninger published in ACM's Transactions on Intelligent Systems and Technology demonstrates how a single up/down vote on Reddit dictates the content users see on the site. "[Voters] become the content editors...and they're the ones who are responsible for what's trending on the site," Weninger says. "Anyone can vote on social media, so it's relatively easy for a handful of bad actors to manipulate the news and opinions and commentary that you see in social media." Weninger and Notre Dame's Maria Glenski tracked new Reddit posts for five months, randomly assigning an up-vote, down-vote, or no vote to each post. Up-vote posts received a significantly higher final score than posts with no vote, and down-vote posts got a significantly lower score. "Moreover, the posts that we up-voted were 24 percent more likely to reach the front page than those that we did not up-vote," Weninger says. He notes the content on Reddit is based on the aggregate opinion of those who vote, which comprises a relatively small portion of users. "It is critical that we understand the dynamics of how social rating systems curate the media that we all see and hear in our daily lives," Weninger says.

Your Android Device's Pattern Lock Can Be Cracked Within Five Attempts
Lancaster University (01/23/17)

Researchers at Lancaster University and the University of Bath in the U.K., and Northwest University in China, have found attackers can crack Android's Pattern Lock security system within five attempts by using video and computer-vision software. An attacker can covertly record the owner drawing their pattern lock shape to unlock their device, and then use software to track the owner's fingertip movements relative to the position of the device. Within seconds, the algorithm produces a group of candidate patterns to access the Android phone or tablet. The researchers also found the attack works even without the video footage being able to see any of the onscreen content, regardless of screen size. The team evaluated the attack using 120 patterns collected from independent users, and they cracked more than 95 percent of patterns within five attempts. Although complex patterns are used to make it harder for observers to replicate, the researchers found these shapes are in fact easier to crack because they help the fingertip algorithm to narrow down the possible options. "Contrary to many people's perception that more complex patterns give better protection, this attack actually makes more complex patterns easier to crack and so they may be more secure using shorter, simpler patterns," says Northwest University's Guixin Ye.

Girls Who Code Closing Computer Science Gender Gap
Associated Press (01/23/17) Jillian Ward

The gender gap in computer science today is worse than it was in the 1980s, but Girls Who Code is working to change that and boost diversity in the profession. "Currently, women pose only 18 percent of the computer science field," says Oregon high school senior Cayce Hill, youth liaison for the Southwestern Oregon Workforce Investment Board (SOWIB). "By 2020, there will be 1.4 million job openings and women will only fill 3 percent of that demand. That's worse than it was in the 1980s, and that gap is growing larger." SOWIB adopted the Girls Who Code program when program manager Kyle Stevens saw a clear local need. The club notes interest in computer science programming among girls is very high between 6th and 8th grades, but that interest wanes when they become high schoolers. "We want to educate them at that early age so they can stay interested and close the gender gap," Hill notes. Samantha Buckley leads the weekly Girls Who Code classes in Coos Bay, OR, and she sees the group's emphasis on fellowship and education as positive. "It's important for girls to stick together and support each other," Buckley says. Advocates say attendees also can list their participation on resumes and use their programming skills to secure industry jobs.

Brief Interventions Help Online Learners Persist With Coursework, Stanford Research Finds
Stanford News (01/19/17) Alex Shashkevich

Assuring learners from less-developed countries they belong and affirming their core values can help them succeed in massive open online courses (MOOCs), according to a study led by Stanford University. The study found achievement rates in online courses vary greatly by geographical location, with learners living in countries low on the United Nations' Human Development Index less likely to complete the course. The researchers see a psychological barrier contributing to the geographical gap in MOOCs. They collected data on 2,286 learners enrolled in computer science MOOCs, of whom 16 percent lived in less-developed countries. Learners were asked to complete an online activity before beginning the MOOC, with some learners randomly assigned a social belonging activity while others were asked to write about how taking the course affirms their most important values. Both interventions had a significant effect on the performance of learners in less-developed countries, doubling their persistence and virtually eliminating the global achievement gap. The affirmation intervention raised completion rates for learners in less-developed countries from 17 percent to 41 percent. "It is an impressive result, which suggests that social identity threat can be a barrier to performance in international learning contexts, even in online environments with little social interaction," says Stanford's Rene Kizilcec.

Robot Skin Senses Warm Bodies Like a Snake Locating Nearby Prey
New Scientist (01/20/17) Edd Gent

A temperature-sensitive film inspired by natural snakeskin membranes can help robots detect when humans are near. The technology was developed by Raffaele Di Giacomo and colleagues at ETH Zurich in Switzerland. The flexible coating is made of pectin, a low-cost plant material that relies on currents of ions instead of electrons to detect small variations in temperature. The film can sense temperature changes as small as 10 millikelvin and detect a warm body the size of a rabbit from several feet away. Sudden variations in temperature cause the film's resistance to change, which is sensed by electrodes along the edges of the film and transmitted to a computer. The ETH team is developing algorithms to map temperature across complex surfaces, and they say applying the film around the robot's mechanical parts could provide 360-degree thermal sensing, enabling machines to navigate crowded areas or help locate humans in rubble and smoke-filled areas. The film also could be used in prostheses to give the wearer sensory feedback. "The most important thing about combining (artificial intelligence) AI and humanoid robots is that this AI needs to be shaped by its senses like we are," Di Giacomo says.

SMART Automation
MIT News (01/19/17) Catherine Marguerite

Researchers at the Singapore-Massachusetts Institute of Technology (MIT) Alliance for Research and Technology (SMART) have spent the last several years developing autonomous vehicles as part of the Future of Urban Mobility Research Program. Since the public trial, the group has introduced a self-driving city car, a self-driving scooter, and a self-driving wheelchair, each of which was designed in three phases. In the first phase, the vehicle was converted to drive-by-wire control, which enables a computer to control driving functions. In the second phase, the vehicle drives through its operation environment and creates a map using features detected by its sensors. Finally, the vehicle uses the map to determine a path from the customer's pick-up point to the customer's destination and drives along the path, localizing continuously to avoid obstacles. The vehicles also use traffic data sourced from Singapore's Land and Transportation Agency to model traffic patterns. Last April, the researchers conducted a public trial at MIT to test their self-driving scooter's ability to drive indoors as well as outdoors. The researchers envision a future in which a mobility-impaired user could schedule combinations of autonomous wheelchairs, scooters, and cars to reach their destination.

Your 'Anonymized' Web Browsing History May Not Be Anonymous
Princeton University (01/19/17) John Sullivan

Researchers at Princeton and Stanford universities have found a specific person's online behavior can be identified by linking anonymous Web browsing histories with social media profiles. Although the U.S. Federal Communications Commission recently adopted privacy rules for Internet service providers permitting them to store and use consumer information only when it is "not reasonably linkable" to individual users, the study suggests pseudonymous browsing histories fail this test. The researchers wanted to determine if it was possible to de-anonymize Web browsing and identify a user even if the browsing history did not include identities. They limited themselves to publicly available information, with social media profiles that include links to outside Web pages offering the strongest possibilities. The team created an algorithm to compare anonymous browsing histories with links appearing in people's public social media accounts. "Given a history with 30 links originating from Twitter, we can deduce the corresponding Twitter profile more than 50 percent of the time," the researchers note. "All the evidence we have seen piling up over the years showing the strong limits of data anonymization, including this study, really emphasizes the need to rethink our approach to privacy and data protection in the age of big data," says Imperial College London professor Yves-Alexandre de Montjoye.

AI Software Learns to Make AI Software
Technology Review (01/18/17) Tom Simonite

Several research organizations, including Google Brain and DeepMind, are working to create artificial intelligences (AI) that can in turn develop machine-learning software. In many cases, the results coming from machines programming other machines match or exceed work done by humans. If self-programming AI techniques become practical, they could increase the pace at which machine learning is adopted throughout the economy without requiring more machine-learning experts, who already are in short supply. One set of experiments from DeepMind suggests self-teaching methods could alleviate the problem of AI software needing to consume massive amounts of data on a specific task. Researchers challenged their software to create machine-learning systems for collections of multiple, related problems. The software produced designs that demonstrated an ability to generalize and adopt new tasks with less training. A team at the Massachusetts Institute of Technology (MIT) plans to open source the software behind their experiments, in which an AI designed deep-learning systems that matched systems made by humans on standard tests for object recognition. However, these techniques require extreme computer power and are not yet viable replacements for machine-learning experts. MIT Media Lab's Otkrist Gupta believes companies will be motivated to find ways to make automated machine learning practical.

Abstract News © Copyright 2017 INFORMATION, INC.
Powered by Information, Inc.

To submit feedback about ACM TechNews, contact: [email protected]