Welcome to the July 27, 2018 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).
To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version.
The online version now has a button at the top labeled "Show Headlines."
|
|
|
This Bomb-Simulating U.S. Supercomputer Broke a World Record
Wired
Sarah Scoles
July 23, 2018
The Trinity supercomputer at the Los Alamos National Laboratory helped Brad Settlemyer and colleagues at Carnegie Mellon University's Parallel Data Laboratory to model a plasma. The researchers used a vector particle-in-cell code, developed to monitor individual particles, which nuclear researchers often utilize to understand how plasma mixes with plasma. Under classic simulation schemes, the supercomputer would have to capture images of 1 trillion particles simultaneously, find the most energetic particles, and then reverse their trajectories to extract those particles' pathways. This was impractical, so Settlemyer's approach was to generate more files with less information, using one file for each particle, tracing each through the whole of the simulation, then storing those files within a searchable index. Settlemyer says this method can retrieve data up to 5,000 times faster, and has enabled Trinity to produce 1 trillion files in two minutes.
|
Python Has Brought Computer Programming to a Vast New Audience
The Economist
July 19, 2018
The Python programming language's user population has grown dramatically in the three decades since its invention by Dutch researcher Guido van Rossum. Stack Overflow estimates about 40% of Python users are professional developers, but its popularity among amateur developers and aspiring learners is staggering. Enthusiasts have added more than 145,000 software packages to Python's repository, running the gamut from astronomy to game development. Python's most compelling features are simple syntax that eases code learning/sharing and a massive assortment of third-party packages. Van Rossum says Python has become the preferred language for artificial intelligence researchers, given its neural network creation packages. Education in Python is mostly limited to science, technology, engineering, and math students, but there are suggestions to encourage its use by younger learners by making computer science mandatory in primary schools.
|
Hubble Researcher Focuses on Blockchain for Space Data Processing
CoinDesk
Nikhilesh De
July 24, 2018
Space Telescope Science Institute astronomer Josh Peek says he is testing a decentralized blockchain network for processing the massive volumes of data produced by the Hubble Space Telescope (HST). Peek is using an AIKON interface to enter the data, which is then processed and rendered via computers forming Hadron's blockchain network. AIKON's Marc Blinder says computing time is paid for with AIKON's "CPU" tokens. Volatility is reduced by tying the value of a token to the average cost of computing power charged by cloud hosting services, while Peek says the system streamlines the process of tapping into a distributed network. Peek says, "AIKON is trying to...make this as simple as possible with no cryptocurrency nonsense."
|
Computing Power Solves Molecular Mystery
Gemini
Pernille Feilberg
July 24, 2018
Norwegian University of Science and Technology (NTNU) researchers have come up with a new technique for probing molecular behavior. The team sought to determine the mechanism that induces the division of water molecules by combining two previously non-integrated simulation methods. NTNU's Anders Lervik said, "We started looking at these 10,000 simulation films and analyzing them manually, trying to find the reason why water molecules split." The team employed machine learning to discover the reaction's underlying causes, finding factors describing what triggers the reactions. The work has significant potential for enhancing hydrogen production, and boosting the speed and efficiency of industrial chemical reactions.
|
DARPA Seeks Underground Mapping Capability
Federal Computer Week
Mark Rockwell
July 24, 2018
The U.S. Defense Advanced Research Projects Agency (DARPA) in September will host competitions in a limestone cavern in Louisville, KY, to advance geotechnology's subterranean mapping capabilities. The agency says it is seeking the ability to locate people quickly in complex underground environments where rescuers have no clear paths and live operations are too hazardous. DARPA's SubT Challenge tasks technology providers with developing solutions "that will dramatically outperform" current methods that rely on labor-intensive and potentially dangerous manual search and mapping capabilities. Participants must assemble core systems that possess autonomy, perception, networking, and mobility, with an overarching goal of integrated systems and software solutions. DARPA says it has established dual systems and virtual contest tracks: the former will devise physical strategies to be tested underground, while the latter aims to develop software and algorithms to virtually build subterranean systems, environments, and terrain for simulations targeting larger-scale scenarios.
|
Email Malware Detector Can Outperform the Top 60 Antivirus Engines
TechRepublic
Conner Forrest
July 18, 2018
Researchers in the Malware Laboratory of Ben-Gurion University of the Negev (BGU) in Israel have developed a method for detecting malicious emails which they say is more effective than the top 60 antivirus engines on the market. The new Email-Sec-360° system is built on machine learning principles and operates without Internet access; it relies on 100 email features to detect a malicious message. The researchers built the detection model using 12,835 malicious emails and 20,307 benign emails, collected between 2013 and 2016. During testing, the researchers found the system beat the next-best solution by 13%. BGU’s Nir Nissim says the researchers hope to extend their research by "integrating analysis of attachments such as PDFs and Microsoft Office documents within Email-Sec-360°, since these are often used by hackers to get users to open and propagate viruses and malware."
|
Future Electronic Components to Be Printed Like Newspapers
Purdue University News
Kayla Wiles
July 19, 2018
Purdue University researchers have developed roll-to-roll laser-induced superplasticity, a manufacturing technique that can make electronics faster than conventional methods. The process is similar to newspaper printing, but it can form smoother and more flexible metals that are needed to make ultrafast electronic devices. The method uses a rolling stamp such as those used to print newspapers at high speed, and it can briefly induce "superelastic" behavior in different metals by applying high-energy laser shots, enabling the metal to flow into the nanoscale features of the rolling stamp. Purdue's Ramses Martinez says this process could enable the creation of touchscreens covered with nanostructures that can interact with light and generate three-dimensional images.
|
AI Plus a Chemistry Robot Finds All the Reactions That Will Work
Ars Technica
John Timmer
July 18, 2018
Researchers at Glasgow University in the U.K have coupled a machine learning algorithm with a robot to run and analyze its own chemical reaction, resulting in a system that can determine every possible reaction that can be induced from a given set of starting materials. The algorithm was fed a series of 72 reactions with known products so it could generate predictions of the outcomes of further reactions; afterwards, it randomly selected reactions from the remaining list of options to ascertain whether they also created products. The algorithm predicted the results of untested reactions with more than 80% accuracy by the time it had sampled 10% of the total possible reactions, without bias of which reactions would be viable. The researchers added a neural network given data on specific chemical studies and trained the system on about 3,500 reactions; it yielded an error rate of only 11% when predicting the outcomes of another 1,700 reactions from the research.
|
RMIT to Provide Students With Blockchain-Based Digital Credential Platform
ZDNet
Asha McLean
July 24, 2018
The Royal Melbourne Institute of Technology (RMIT) in Australia, in partnership with digital credentialing platform Credly, has launched an initiative to provide students with blockchain-enabled digital credentials. The initiative will give students the option to publish data about their skills and capabilities to blockchain, as well as permitting them to share that information to social networks such as LinkedIn. RMIT's Belinda Tynan says the initiative is "yet another way that we are working with industry to demonstrate we are providing real-world benefits and meaningful student outcomes." In addition, the university has established the RMIT Blockchain Innovation Hub, which it describes as the world's first research center focused on the social science of blockchain. The hub aims to develop an interdisciplinary research team focused on the economic, cultural, and social implications and impacts of blockchain technologies.
|
Educators Explore How to Bring Computer Science Learning to All Students
Cornell Chronicle (NY)
Melanie Lefkowitz
July 24, 2018
Representatives of four New York state school districts participating in the two-year U.S. National Science Foundation-funded CS Visions project to develop consistent, meaningful computer science (CS) curricula came together at a recent workshop to address the challenges they face. Representatives from those districts, along with the Tompkins-Seneca-Tioga Board of Cooperative Educational Services (TST BOCES) and Putnam/Northern Westchester BOCES, discussed the past year's progress with CS Visions, a two-year project led by the CSforAll Consortium and New York University. Cornell Tech's Diane Levitt said participants of the workshop were "so knowledgeable, they’re so driven, they’re so passionate about bringing this to their kids. Last year they saw the barriers more than they saw the opportunities, and that has completely switched.” Educators say the relationship has opened up new opportunities, knowledge, and support so they can coherently and thoughtfully integrate CS learning into their curricula. CSforAll's Leigh Ann DeLyser says her group's mission is to help each district devise an approach aligned with its own goals, and CS Visions will inform the education CSforAll provides to districts throughout the U.S.
|
Evolutionary Algorithm Outperforms Deep-Learning Machines at Video Games
Technology Review
July 18, 2018
Evolutionary computing can match the performance of deep learning machines in outperforming humans at arcade video games, according to researchers at the University of Toulouse in France. Evolutionary computing aims to produce computer code that solves a specific problem by starting with many randomly generated codes, and preserving only the parts that work best for the "next generation" of codes. Each new generation of codes is then tested, and again the best pieces are preferentially reproduced into another generation. Over time and after many generations, the code can become better than any human coder can design. Evolutionary computing codes are smaller than those used in deep learning techniques, making it easier to see how they work; the research suggests computer scientists should not focus exclusively on deep learning.
|
Speed Up Solving Complex Problems: Be Lazy and Only Work Crucial Tasks
Aalto University
July 19, 2018
Research teams at Aalto University in Finland and KU Leuven in Belgium have developed an approach to "lazy grounding" that could solve hard-set and complex issues in freight logistics, routing, and power grids by significantly reducing computation times. Conventional methods to "ground" such computations free up memory, but may cause the system to get stuck in searching for a solution and suddenly require an unreasonable amount of time. The new method pinpoints the small subset of decisions that contribute to a wrong turn somewhere in the computation, and ignores the rest. Aalto's Antonius Weinzierl says, "Our approach essentially draws a local part of the map on demand and allows you to pinpoint where exactly the initial wrong turn was and how to get straight back on track."
|
A Fault-Tolerant System for Stopping 'Leaks' in Quantum Computers
YaleNews
Jim Shelton
July 24, 2018
Researchers from Yale University have designed a system to prevent future quantum computers from "leaking" errors in ancillary quantum bits (qubits) to logical qubits. Such leakage introduces more errors than the ancillary qubits are able to correct. The researchers say they have addressed this problem by enhancing the system's fault tolerance. Yale's Philip Reinhold says this "essentially amounts to making a robust system out of faulty parts," opening the way for error correction that increases quantum computation's fidelity. Yale's Serge Rosenblum says the method reduces the odds of catastrophic failure during error detection in quantum systems by a factor of five.
|
|
