Welcome to the November 30, 2016 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).
HEADLINES AT A GLANCE
Quantum Keys Transmitted Over Record-Breaking 404 Km
Asian Scientist (11/29/16)
Researchers at the University of Science and Technology of China have set a new world record in quantum communications, securely transmitting information over a distance of 404 kilometers. The researchers, led by professor Pan Jianwei, used ultralow-loss optical fibers to transmit photons over distances of 102 to 404 kilometers. The team also used an optimized four-intensity decoy-state method that sends out decoy pulses of light, which prevents eavesdropping attacks. For each distance, the researchers determined the maximum speed by which cryptographic keys could be securely distributed. The researchers demonstrated a 500-fold increase in speed over earlier tests, reaching a key-distribution rate that would allow for encrypted voice transmission by telephone. The record-breaking implementation of the measurement-device-independent quantum key distribution (MDIQKD) method provides a new distance record for MDIQKD and all types of QKD systems. It also achieves a distance the traditional Bennett-Brassard 1984 QKD would not be able to accomplish with the same detection devices. The researchers say the work represents a significant breakthrough in proving and developing feasible long-distance QKD.
Europe Is Leading the Way in AI and Machine Learning (and Even Silicon Valley Wants In)
Wired.co.uk (11/30/16) Victoria Woollaston
Europe is either ahead of or on equal footing with the U.S. in terms of artificial intelligence (AI), machine learning, and other technologies. Europe has invested $2.3 billion into deep tech since the start of 2015, and this year investment is nearing $1 billion. Since 2011 the U.K. has invested $1.3 billion in deep tech, more than any other European country, according to Dealroom.co data. Berlin is Europe's leading hub for AI, virtual reality, and augmented reality developments. "A new generation of entrepreneurs has been inspired by Europe's strong AI companies, and founders are now building their own their own machine intelligence, computer vision, and other deep tech startups," notes Atomico partner Siraj Khaliq. "They've found that there's plenty of talent here in Europe to make it happen, and the ecosystem is ready to power this deep tech revolution." U.S. companies are establishing European branches, with Alphabet setting up Swiss and London hubs and Apple building bases in France, Germany, Sweden, and the U.K. "I'm seeing this deep thinking, not just in the U.K., but also in Berlin, where they're bringing together the AI community and crypto scene to rethink some of the core questions around decentralized systems," says Exponential View founder Azeem Azhar.
Improved Algorithm Produces More Efficient Traffic Management for Cloud Data
The Stack (UK) (11/29/16) Nicky Cappella
Researchers from Ghana and Nigeria improved the Max-Min algorithm to increase efficiency in cloud computing via effective resource allocation and traffic management. The update applies calculated average mean time to execute for a group of tasks and optimizes the allocation of resources, resulting in better efficiency and reduced operating time compared to the existing Max-Min algorithm. The algorithm consistently places the tasks requiring the most time to execute on the fastest available machine or resource, which completes that task in the shortest possible timeframe. However, this also increases the total length of the makespan, the time required to complete all tasks, by delaying tasks requiring the minimum time to execute. The researchers adjusted the Max-Min algorithm to reduce makespan, narrowing the performance gap between service providers and users. The revision assembles all requests in increasing order of the time required to execute, and the researchers then implemented arithmetic and geometric means to calculate the average execution time for assigned tasks. The tasks were assigned based on which tasks approached the calculated mean time to execute, with those tasks requiring average time to execute receiving priority and statistical outliers completed last. Experimental testing showed using mean calculations and assigning tasks based on proximity to the mean results in a better makespan and improved resource optimization.
Eight States Have Fewer Than 10 Girls Take AP Computer Science Exam
THE Journal (11/29/16) Richard Chang
Nearly 55,000 students this year took the Advanced Placement (AP) computer science A exam in the U.S., a 17-percent gain over last year. Of this group, 23 percent of test takers were girls, up from 22 percent in 2015. Although the total number of high school girls taking the annual exam is increasing nationwide, fewer than 10 girls took the May 2016 exam in eight states, and not a single girl took the test in Mississippi or Montana, according to an analysis of the data from the College Board by the Georgia Institute of Technology's (Georgia Tech) Barbara Ericson. In South Dakota, only one out of 26 test takers was female, while in Wyoming two out of six test takers were girls. Ericson cites studies showing girls are often discouraged from entering the computer science field by their teachers and parents. Meanwhile, 36 percent of AP computer science test takers in West Virginia were girls, marking the highest percentage of any U.S. state. The pass rate for girls stood at 61 percent, three points below the national pass rate of 64 percent. However, in New Jersey, 69 percent of girls passed the test, while 67 percent of girls passed in California and Illinois, and 66 percent of girls passed in Massachusetts.
'Hackathon' Attempts to Stem Proliferation of Fake News
Financial Times (11/28/16) Hannah Kuchler
A hackathon inspired by worries about the influence of fake news over the U.S. presidential election seeks to bring the technology industry and leading media thinkers together to incentivize the prioritization of truth. Dartmouth College professor Hany Farid sees one solution in adapting algorithms designed to recognize and flag inappropriate content, such as extremist propaganda and child pornography, on social networks. "It is disingenuous to say this is too hard a problem because as soon as there is a financial incentive, they get a lot smart," Farid says. "Social media and the Internet reach millions and millions of people, do incredible harm, propagate hatred and influence elections, they have to take that responsibility seriously." Others contend fact-checking requires greater human participation, and Wikimedia Foundation executive director Katherine Maher cites the community-vetting Wikipedia model, with its emphasis on transparency, as a template for the proliferation of reliable information. She says this is in marked contrast to Facebook, whose algorithm supplies content that is often devoid of context. Betaworks executive director John Borthwick and City University of New York professor Jeff Jarvis are urging collaborative efforts against fake news, suggesting the development of technologies that trace back to the root source of news items and memes.
U.S. Students Still Lag Many Asian Peers on International Math and Science Exam
The Washington Post (11/29/16) Emma Brown
Results from the Trends in International Mathematics and Science Study (TIMSS) show U.S. students still lag behind many of their Asian peers. Scores have risen at both the eighth- and fourth-grade levels in math and science since the TIMSS exam was first administered in 1995, but U.S. students are still behind their peers in 10 other countries, including Singapore, Hong Kong, South Korea, Taiwan, and Japan. Half of students in Singapore scored high enough to reach the "advanced" benchmark in math, compared with only 14 percent of U.S. students. U.S. fourth graders scored the same as nine other systems and higher than 34 systems. Results of TIMSS Advanced also were released Tuesday and represent the performance of high school seniors taking advanced courses in physics and math. U.S. students' performance on this exam remained unchanged since 1995, the only other year in which the U.S. participated in the exam. However, the gender gap among the advanced high school seniors has widened. Male students scored 46 points higher than their female peers in physics and 30 points higher in math.
Driverless Cars to Hit the Road in Ontario Early Next Year
The Toronto Star (11/28/16) Michael Lewis
The University of Waterloo's Center for Automotive Research in Canada will begin testing an autonomous vehicle, a Lincoln MKZ hybrid sedan nicknamed the Autonomoose, on Ontario's public roads early next year. The province also has approved road testing for two more groups developing self-driving cars. The road tests will give the groups the opportunity to evaluate how the vehicles operate under various weather and road conditions, speeds, and degrees of automation. Ontario Transportation Minister Steven Del Duca on Monday said an ecosystem needs to be developed to align Wi-Fi and other infrastructure with communications systems in vehicles. A year ago, Ontario announced a 10-year project that provided the opportunity for it "to be a world leader in automated technology." At that time, officials said nearly 100 companies and institutions were involved in research and development for the connected vehicle and automated vehicle industry in the province. "The ability to take this research work to the next level while safely testing on all kinds of roads in Ontario represents a significant leap forward in this field," says University of Waterloo lead researcher Krzysztof Czarnecki.
Inside the Project to Rebuild the EDSAC, the World's First General Purpose Computer
ZDNet (11/24/16) Danny Palmer
Volunteers at the U.K.'s National Museum of Computing are physically reconstructing the Electronic Delay Storage Automatic Calculator (EDSAC) mostly from photos in the absence of a complete blueprint. The room-sized machine was one of the first practical general-purpose computers, capable of performing 650 instructions a second, and computing more than 1,500 times faster than the mechanical calculators it supplanted when it ran its first program in 1949. Project leader Andrew Herbert says part of the challenge in recreating the pioneering system is the fact that its designers and programmers approached its creation from different technical backgrounds than today's digitally savvy scientists. Chief designer Chris Burton says ensuring the replica's authenticity requires the volunteers to work exactly as the original design team worked. Essential to the effort's success was consulting with surviving users of the original EDSAC to gain insight into its daily operations. The project team hopes the EDSAC's replication will demonstrate its importance in the history of computing and inspire later generations. The machine's value lies in the fact it shares the same basic processes and components as modern computers, and can be used to teach basic computer functionality to future coders.
The Next Generation of HPC Technology Dominates Green500
Scientific Computing World (11/24/2016)
The latest Green500 list of the most energy-efficient supercomputers demonstrates recent significant progress in high-performance computing (HPC). As new multicore processors from Intel and NVIDIA start penetrating top systems, performance-per-watt numbers are rising. The DGX SATURNV, NVIDIA's first petascale supercomputer, earned the top spot with a rating of 9.46 gigaflops per watt, 40-percent higher than the top system six months ago. The supercomputer's superior efficiency is due to its DGX-1 server, which houses eight P100 graphical-processing units (GPUs). Piz Daint, a Cray XC50 system, achieved a mark of 7.45 gigaflops per watt; based on the same processor as the DGX SATURNV, the Piz Daint uses a lower ratio of GPUs to central-processing units. Most of the remaining top 10 systems relied on homogenous manycore processors in a standalone configuration. Five of those machines are powered by Intel's Knights Landing Xeon Phi processors, which exhibit energy efficiencies of between 3.84 and 5.61 gigaflops per watt. The spikes in energy efficiency this year have been due to recent releases of new processor technology, so it is unlikely there will be a similar jump in efficiency before the next generation of GPUs, Xeon Phi processors, and manycore processors are available in late 2017 to early 2018.
A New Perovskite Could Lead the Next Generation of Data Storage
Swiss Federal Institute of Technology in Lausanne (11/24/16) Nik Papageorgiou
Researchers from the Swiss Federal Institute of Technology in Lausanne (EPFL) have developed a perovskite material whose magnetic order can be rapidly changed without disrupting it due to heating. The researchers say the new material demonstrates some unique properties that could be used to help build next-generation digital storage systems. "We have essentially discovered the first magnetic photoconductor," says EPFL postdoctoral researcher Balint Nafradi. The new crystal structure combines the advantages of ferromagnets, in which magnetic moments are aligned in a well-defined order, and photoconductors, where light illumination generates high-density free conduction electrons. The combination of the two properties resulted in the "melting" of magnetism by photo-electrons, which are electrons that are emitted from a material when light hits it. In the new material, a red light-emitting diode is enough to "melt" the material's magnetic order and generate a high density of traveling electrons, which can be freely and continuously tuned by changing the light's intensity. In addition, the timescale for shifting the magnetic field in this material is very fast, needing only quadrillionths of a second. "This study provides the basis for the development of a new generation of magneto-optical data storage devices," Nafradi says.
Researchers Run Largest Known Transparent Checkpointing Process
Northeastern University News (11/23/16) Shandana Mufti
A team led by Northeastern University researchers has completed what is believed to be the largest known instance of transparent checkpointing. The MVAPICH software supporting the Message Passing Interface was used to run the High Performance Conjugate Gradients program for linear algebra in parallel over 32,768 central-processing unit (CPU) cores on 2,048 computers. It used total memory of 38 terabytes, and was checkpointed in 10 minutes, 53 seconds. A second program, Nanoscale Molecular Dynamics, was run in parallel over 16,368 CPU cores on 1,024 computers, using total memory of 10 terabytes. The Distributed Multi-Threaded CheckPointing software, now in its second decade, is responsible for checkpointing, which enables computer scientists and engineers working on large projects to save and reopen programs without modifying any code. The team carried out the processes on the Stampede supercomputer at the Texas Advanced Computer Center (TACC). "These results show how the Extended Collaborative Support Services from the [U.S.] National Science Foundation-supported Extreme Science and Engineering Discovery Environment can help scientists and developers improve the scalability and efficiency of their code on high-performance computing clusters," says TACC research associate Jerome Vienne.
For Wearable Electronic Devices, NIST Shows Plastic Holes Are Golden
NIST News (11/23/16) Chad T. Boutin
A U.S. National Institute of Standards and Technology (NIST) research team has developed a way to build safe, nontoxic gold wires onto flexible, thin plastic film. The team says a flexible plastic membrane on which wearables would be built might work better if the membrane had microscopic holes in it. When gold wiring is built upon a plastic membrane with holes, bending the plastic produces only minor cracks in the gold, and the wire is still able to conduct electric signals. The finding could be a major step toward creating electronics that are flexible enough to be worn comfortably on or inside the human body without exposing a person to harmful chemicals in the process. In addition, the devices and will last long enough to be useful and convenient. "This thin membrane could fit into very small places, and its flexibility and high conductivity make it a very special material, almost one of a kind," says NIST biomedical engineer Darwin Reyes-Hernandez. Moreover, he notes gold is a good option because it does not corrode and is nontoxic. Reyes-Hernandez also says the porous membrane's electrodes exhibit even higher conductivity than their counterparts on rigid surfaces.
Positive Language Is on the Decline in the United States, Study Finds
USC News (11/22/16) Michelle Boston
A new study from researchers at the universities of Southern California (USC) and Michigan explored the phenomenon called language positivity bias (LPB). Research shows people are more likely to use positive rather than negative words on the whole in their communications. The new study found that people's tendency to use positive language has been on the decline in the U.S. over the past 200 years. The researchers analyzed writing published over the past two centuries from Google Books and The New York Times, and calculated the ratio of positive to negative words. The computations found LPB has waned significantly. The researchers say the study demonstrates the value of newly available data sources to address long-standing scientific questions. "In addition to showing that it's important to look at language as a dynamic feature of human psychology, we're also showing that techniques from computer science--the big data of psychology--can have many important applications in studying and verifying long-standing theories about human psychology, and about social sciences more generally," says Morteza Dehghani, who runs the Computational Social Science Laboratory at the USC Dornsife College of Letters, Arts, and Sciences.
Abstract News © Copyright 2016 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: [email protected]
|
