Welcome to the August 17, 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."
|
|
|
Common Wi-Fi Can Detect Weapons, Bombs, and Chemicals in Bags
Rutgers Today
Todd B. Bates
August 14, 2018
Rutgers University-New Brunswick researchers found common Wi-Fi technology can easily identify weapons, bombs, and explosive chemicals in bags at public venues. Their inexpensive detection system needs a Wi-Fi device with two to three antennas and can be integrated into existing Wi-Fi networks. The system analyzes what occurs when wireless signals penetrate and bounce off objects and materials. Detection is 99% accurate for dangerous objects, 98% accurate for metal, and 95% accurate for liquid. Rutgers's Yingying Chen says the accuracy rate tops 95% for most backpacks, falling to about 90% when objects inside bags are wrapped. She says the next phase of the project will involve boosting the system's accuracy in identifying objects by imaging their shapes and calculating liquid volumes.
|
Oracle Open-Sources Graphpipe to Make It Easier to Deploy Machine Learning Models
VentureBeat
Khari Johnson
August 15, 2018
Oracle has released an open-source tool called Graphpipe to facilitate the deployment of machine learning models. Graphpipe will simplify the use of machine learning for mobile apps and IoT devices, end user services, and internal corporate functions. Using Graphpipe could obviate the need for developers to create custom APIs to deploy AI models or to tailor their work to a specific framework. Says Oracle's Vish Abrams, "Graphpipe is an attempt to standardize the protocol by which you speak to a remotely deployed machine learning model, and it includes some reference servers that allow you to deploy machine learning models from existing frameworks very easily in an efficient way." Although developers have several framework choices to build AI models, fewer options exist for how to serve or deploy AI models, Abrams says.
|
Artificial Intelligence Tool 'As Good as Experts' at Detecting Eye Problems
The Guardian
Samuel Gibbs
August 13, 2018
DeepMind, Moorfields Eye Hospital NHS Foundation Trust, and University College London in the U.K. have developed a machine learning system as effective as top human experts at spotting eye problems and referring patients for treatment, according to researchers. The system analyzes the optical coherence tomography (OCT) scans of patient retinas. Five independent machine learning systems, trained on 877 clinical OCT scans, generate maps of the scans, which are analyzed by a second series of five systems, trained on maps produced from 14,884 OCT scans from 7,621 patients. The second series interprets the maps and each makes a referral decision. The decisions are blended into one result, visualized as a percentage-based confidence rating. Scientists say the system can correctly refer patients with more than 50 different eye diseases for further treatment with 94% accuracy.
|
Replacing Your Boss With a Cruel Robot Could Make You Concentrate More
New Scientist
Leah Crane
August 15, 2018
Researchers at the University of Clermont Auvergne in France say a mean robot could motivate human employees to improve their cognitive performance, compared to supervision by a friendly robot or no robot supervision. They employed the Stroop task, a test in which words printed in different colors are displayed on a screen, with test subjects required to identify the color while ignoring the word. Into this task Clermont Auvergne's Nicolas Spatola and colleagues presented a humanoid robot perceived either as friendly or mean. Participants paired with mean robots were faster and made fewer errors than those paired with friendly ones or with no robot at all. Although participants paid more attention to the mean robot, Spatola says it would be "wrong and a bit dystopian to conclude that if we put a bad robot in every place everyone will perform better." He says, "Imagine a robot sitting in your office insulting you every day—it will not be good for your long-term performance."
*May Require Paid Registration
|
Irish Researchers Make Key Breakthrough on 'Brain-Like' Computers
The Irish Times
Kevin O'Sullivan
August 13, 2018
Researchers at AMBER, the Science Foundation Ireland-funded materials science center, have demonstrated the effectiveness of nanowire networks to reveal a key pathway that may optimize the flow of data in neuromorphic computers. With participants from Trinity College and University College Cork, both in Ireland, and Duke University, the team developed "neural network systems with brain-inspired architectures." The human brain develops preferred communication pathways that connect different brain circuits or loops to quickly and efficiently complete specific tasks, and the research shows evidence that the same behavior can be induced in a nanowire network. Trinity College's John Boland says, "Think of walking through a university campus or business park with some grassy areas and paths connecting the different buildings. There will be foot-worn short cuts in the grass that people take to save time and energy. The combination of frequently-used paved and unpaved pathways are the most practical or preferred pathways for moving efficiently."
|
Programming Languages May Finally Be Reaching a Status Quo
Wired
Klint Finley
August 13, 2018
Programmers' interest in Apple's Swift and JetBrains' Kotlin might be starting to slow, according to a report by analyst firm RedMonk. Since 2011, when RedMonk began tracking programmers' interest, Swift and Kotlin grew more rapidly than any other language the firm followed. Earlier this year, Swift tied with Apple's established Objective-C language for tenth place in RedMonk's list, but Swift slipped to 11th place in the latest rankings. Kotlin dropped from 27th place to 28th place. RedMonk's Stephen O'Grady says small changes in the rankings do not indicate a major shift. However, RedMonk's rankings have remained stable in recent years, especially towards the top of the list.
|
Rethinking Social Networks
Carnegie Mellon University
Daniel Carroll
August 8, 2018
Carnegie Mellon University (CMU) researchers have developed a way to show how social networks change and develop over time. The core of the Weighted Betweenness Preferential Attachment (WBPA) model focuses on the notion of "node betweenness," a quality of being between communities, while previous models have centered on the amount of connections that an individual has. The researchers found node betweenness is actually a greater attractor and driver for the formation of social ties than node degree or other measures of centrality. Instead of examining only the amount of connections a single node has, WBPA places more emphasis on community formation and the quality of node connections. CMU's Radu Marculescu says, "The new model builds on the idea that humans are better at observing qualitative aspects than quantitative ones, which is why people typically favor investing in fewer qualitative social ties rather than numerous lower quality ties."
|
New Interactive Machine Learning Tool Makes Car Designs More Aerodynamic
IST Austria
August 14, 2018
Researchers have developed software that uses machine learning to compute model flow around interactively designable three-dimensional objects, which was presented this week at the ACM SIGGRAPH 2018 conference. Nobuyuki Umetani, formerly at Autodesk Research and now at University of Tokyo, and Bernd Bickel at the Institute of Science and Technology Austria say their tool can help improve the aerodynamic characteristics of automobile designs by making streamlines and parameters available in real time. Umetani says, "With our machine learning tool we are able to predict the flow in fractions of a second." He also says the idea of using polycubes to make shapes manageable for machine learning is essential. In this approach, a model begins with a small number of large cubes which are refined and divided into smaller ones; this gives similarly-shaped objects a similar data structure that machine learning techniques can manage and compare.
|
UTSA Researcher Studies Disparities in STEM Career Aspirations for Underrepresented Students
UTSA Today
August 9, 2018
The University of Texas San Antonio's Guan Saw has measured disparities and changes in career aspirations in science, technology, engineering, and math (STEM) fields among high school students in collaboration with Texas A&M University's Chi-Ning Chang and Hsun-Yu Chan. They sought to define the sociodemographic gaps in these aspirations by analyzing a longitudinal study from the National Center for Education Statistics. They estimated that nationally about 11.4% of students were interested in following a STEM career upon entering high school, which fell to 10% after three years. They also found traditionally underrepresented groups including women, African-Americans, Hispanics, and students of low socioeconomic status were less likely to maintain and cultivate an interest in STEM careers. The researchers also learned that 17.9% of white males from a higher socioeconomic status aspired to STEM careers, versus 1.8% of black females from lower socioeconomic levels.
|
Long-Sought Carbon Structure 'Schwarzite' Joins Graphene, Fullerene Family
Nanowerk
August 13, 2018
University of California, Berkeley (UC Berkeley) researchers have demonstrated that carbon structures recently produced in South Korea and Japan are long-sought "schwarzites," which they expect to offer unique electrical and storage properties similar to those found in buckminsterfullerenes, nanotubes, and graphene. The structures were constructed within the pores of zeolites, and UC Berkeley's Efrem Braun and colleagues identified these materials as schwarzites according to their negative curvature, and devised a method for predicting which zeolites can be used to generate schwarzites. Braun simulated zeolite-templated carbon (ZTC) structures computationally using the known structures of zeolites, and collaborated with Senja Barthel of the Swiss Federal Institute of Technology of Lausanne in Switzerland to ascertain which of the minimal surfaces the structures resembled. The new form of carbon has potential applications in technology, partly because schwarzites are expected to hold unusually large amounts of electrical charge, and could lead to improved capacitors in electronics.
|
Artificial Intelligence Model 'Learns' From Patient Data to Make Cancer Treatment Less Toxic
MIT News
Rob Matheson
August 9, 2018
Massachusetts Institute of Technology researchers have developed a machine learning system that determines the fewest, smallest medication doses that could shrink glioblastoma brain tumors without harming the patient. The self-learning model examines treatment regimens currently in use, and iteratively adjusts doses to find an optimal treatment plan with the lowest possible dose potency and frequency. The reinforcement learning method employs agents that complete actions in an unpredictable, complex setting to reach a desired outcome, and receive a reward or penalty depending on whether the action works toward the outcome. Afterwards, the agent adjusts its actions to achieve that outcome. The model designed treatment cycles that reduced potency to 25% or 50% of nearly all doses while maintaining the same tumor-shrinking potential in simulated trials of 50 patients. The research could improve patients' quality of life by decreasing toxic doses of chemotherapy and radiotherapy.
|
Evaluating Creativity in Computational Co-Creative Systems
Tech Xplore
Ingrid Fadelli
August 8, 2018
Researchers at the University of North Carolina, Charlotte (UNC Charlotte) and the University of Sydney in Australia have developed a framework for evaluating creativity in co-creative systems in which humans and computers collaborate on creative tasks. There are three main strategies by which the role of computers in creative systems can be characterized: fully autonomous systems, creativity support tools, and co-creative systems. The team focused on co-creative systems, in which computers and humans collaborate with one another to produce shared creative works. The researchers define a co-creative system as an "interaction between at least one [artificial intelligence] agent and at least one human where they take action based on the response of their partner and their own conceptualization of creativity during the co-creative task." They determined that compared to autonomous systems, co-creative systems benefit from human interaction, introducing human insight and perception of the creative product during the process.
|
New 3D-Printed Device Could Help Treat Spinal Cord Injuries
University of Minnesota
August 9, 2018
University of Minnesota researchers have developed a three-dimensionally (3D)-printed device that might help restore function to people with spinal cord injuries. Starting with any kind of cell from an adult, the researchers can reprogram the cells into neuronal stem cells via bioengineering techniques. The team built a printed guide made of silicone which serves as a platform atop which neuronal stem cells are 3-D printed. The same printer is used to print both the guide and the cells, with the template keeping the cells alive and allowing them to transform into neurons. The implanted guide would be installed at the site of the spinal injury and serve as a "bridge" between living nerve cells above and below the damage. The idea is that this would relieve pain and help patients regain functions including muscle, bowel, and bladder control. The University of Minnesota's Michael McAlpine says, "The fact that we were able to keep about 75 percent of the cells alive during the 3D-printing process and then have them turn into healthy neurons is pretty amazing."
|
Researchers Help Close Security Hole in Popular Encryption Software
Georgia Tech News Center
John Toon
August 9, 2018
Georgia Institute of Technology (Georgia Tech) researchers have helped patch a security flaw that could have enabled the theft of encryption keys from OpenSSL software by briefly eavesdropping on "side channel" signals from smartphones. The hack involved using intercepted electromagnetic signals from the phones that could be analyzed with a small and inexpensive portable device that listened in on a single decryption cycle. Georgia Tech's Milos Prvulovic and Alenka Zajic eavesdropped on two different Android phones using probes in close proximity to the devices without any physical contact. Their hack analyzed signals in a 40-MHz-wide band around the phones' processor clock frequencies, which are nearly 1 GHz. Prvulovic says, "Once we got the attack to work, we were able to suggest a fix for it fairly quickly. Programmers need to understand that portions of the code that are working on secret bits need to be written in a very particular way to avoid having them leak."
|
|
