Welcome to the June 15, 2016 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
Computer Science Salaries Rise With Demand for New Graduates
Network World (06/13/16) Ann Bednarz
The employment prospects for this year's computer science graduates have improved compared to last year's, with competition for new talent intense. "There are currently over 500,000 open computing jobs, in every sector...but only 50,000 computer science graduates a year," according to an open letter from the Computer Science Education Coalition and Code.org to Congress. TEKsystems' Jason Hayman says industry behemoths such as Google and Facebook are grabbing developers, coders, and engineers. He notes with demand for talent so high, companies should be willing to broaden their hiring strategies and "not just fish from one specific pool of degree holders." Demand for computer science graduates is so fierce that many graduates are entering the job market with multiple employment offers and high starting salaries. The National Association of Colleges and Employers estimates the median starting salary this year for computer science bachelor's degree graduates is projected to be $61,321, while engineering graduates are predicted to start at $64,891. Meanwhile, PayScale's College Salary Report named six computer-related majors among the top 20 majors by salary potential. PayScale reports computer science and engineering is the highest-paying major, with an average starting salary of $69,100.
The Road Ahead for AI in Cars
EE Times (06/14/16) Junko Yoshida
Shipments of artificial intelligence (AI) systems for vehicles will soar from 7 million in 2015 to 122 million by 2025, according to an IHS Technology forecast. In an interview, IHS analyst Luca De Ambroggi predicts technology advancements in AI will be "in the steady state in the next 10 years," yielding innovations automotive systems can exploit. De Ambroggi cites the advent of accurate object-recognition systems as the point when AI reached a milestone in vehicle applicability, and he notes AI's ability to recognize multiple objects and give them context also is advantageous for vehicle automation. De Ambroggi expects AI's in-vehicle applications will include a critical role in sensor fusion, while certification of AI will be a key challenge once autonomous cars start to fully leverage the technology. "Tier ones and [original equipment manufacturers] need to develop a set of unified tests--safety parameters--to certify AI," he says. De Ambroggi notes the graphics processing unit currently is the most suitable hardware for implementing AI, although he says the platform is not ideal for mass production. "For genuine deep learning, you need a big muscle in your processor," De Ambroggi says.
CSIRO Concerned With Decline in Young Females Studying Computing
ZDNet (06/15/16) Asha Barabaschow
The decline of girls studying computing at primary and secondary school will impact Australia's ability to meet future workforce needs, according to a report from the Commonwealth Scientific and Industrial Research Organization's Digital Careers education program. The study cites social pressure, lack of self-belief, and the perception computer science is unsuitable for girls as factors influencing female students. "From the 'short poppy syndrome' to persistent stereotypes about [information and communications technology] being the domain of geeky boys, the result is clear: girls are missing out on learning skills that are becoming increasingly more important and valued," says Intel Australia's Kate Burleigh. With Digital Careers pointing to years 7-8 as the time when girls' participation and interest in computing starts to wane, the report recommends focusing on those years to sustain their interest. Digital Careers also suggests efforts tied to the school curriculum have the greatest chance of success in normalizing digital technologies subjects. The program calls for introducing activities such as code clubs for girls, especially in early primary and late secondary school. Other suggestions include furnishing more role models for young females, and addressing parents and caregivers' preconceptions and encouraging them to actively expose their children to computing.
The Payoff of Investing in CS Research: Some Numbers Everyone in CS Should Know
CCC Blog (06/13/16) Greg Hager
In discussing what return on investment (ROI) the U.S. National Science Foundation's National Robotics Initiative (NRI) has yielded since launching five years ago, Johns Hopkins University professor and Computing Community Consortium chair Greg Hager lists several key statistics. He first notes the entire budget for the directorate for Computer & Information Science & Engineering (CISE), which finances about 83 percent of U.S. academic computer science research, was $933 million for fiscal year 2015, an amount which he said is actually closer to $4 billion when all IT research and development-related federal funding is included. Moreover, he says half of the 10 largest firms by market capitalization draw revenue mainly from IT innovations, while only 1 percent of these companies' net revenues exceeds the CISE budget, making the ROI at least 100 to 1. In terms of areas for future computer science investment, Hager says he envisions "a future where the disabled will walk with the aid of exoskeletons, a future where manufacturing will be more efficient, safer, and more flexible using reprogrammable robots, or a future where automated transport allows my aging mother to live independently at home for many years to come."
Artificial Intelligence Produces Realistic Sounds That Fool Humans
MIT News (06/13/16) Adam Conner-Simons
Researchers from the Massachusetts Institute of Technology's Computer Science and Artificial Intelligence Lab (CSAIL) have demonstrated an algorithm that predicts sound from videos. When shown a silent video clip of an object being hit, it can generate a sound realistic enough to fool human viewers. "An algorithm that simulates...sounds can reveal key information about objects' shapes and material types, as well as the force and motion of their interactions with the world," says CSAIL postdoctoral student Andrew Owens. The project began with the recording of about 1,000 videos of an estimated 46,000 sounds representing various objects being struck, scraped, and prodded with a drumstick. The next step was to feed the videos to a deep-learning algorithm, which broke down the sounds and analyzed their pitch, loudness, and other characteristics. "To then predict the sound of a new video, the algorithm looks at the sound properties of each frame of that video, and matches them to the most similar sounds in the database," Owens says. "Once the system has those bits of audio, it stitches them together to create one coherent sound." Human viewers chose clips with the algorithm-produced sound over those with the actual recorded sound twice as often as a baseline algorithm. The research will be presented this month at the ACM Computer Vision and Pattern Recognition (CVPR 2016) conference in Las Vegas.
Cornell and Google Research How to Block Fake Social Engagement
Campus Technology (06/13/16) Dian Schaffhauser
Researchers at Cornell University and Google have found that some YouTube videos with an inordinate number of views may be "fake engagement activities," ploys undertaken by bad actors posting fake content or artificially inflating their number of YouTube engagements through automated means, or by paying people to "like" the content or add comments. The goal of these bad actors is to game the system by inflating engagement metrics in order to obtain better rankings for videos. The researchers note the problem of fake engagement is not limited to YouTube, as it also can be found on all of the major social media sites. They developed Local Expansion at Scale (LEAS) as a way to discern fake activities from legitimate ones. LEAS analyzes the engagement behavior pattern between users and YouTube videos, and the researchers say accounts posting fake hits or comments show a "stronger lockstep behavior pattern." The researchers note LEAS creates an "engagement relationship graph," which accounts for the frequency of common engagement activities shared between two individuals within a short period of time. The graph enables the researchers to detect preconceived actions by sets of users that have a very low likelihood of happening spontaneously or organically.
California Moves to Catch Up on K-12 Computer Science Curriculum
EdSource (06/13/16) Pat Maio
A critical mass of efforts seeking to expand computer science in public schools in California could hasten statewide action. A commission that advises the State Board of Education recently updated proposed computer science curriculum standards as part of the latest review of the Next Generation Science Standards, and California students will be tested on new NGSS-Aligned assessments in the 2018-2019 school year. The Computer Science Teachers Association has begun its final review of a proposed framework for bringing computer science into all K-12 grades. Some of the state's largest school districts, including Los Angeles Unified, Oakland Unified, San Francisco Unified, and Riverside Unified, are not waiting for the state to implement standards, but are creating computer science curriculum pathways from elementary to high school. Several school districts are collaborating with national groups on funding for courses and teacher training. Moreover, legislation has been introduced in the California State Legislature to create an advisory group to develop a strategic plan for computer science instruction. "I'd say we are in the beginning stages of a formalized process of developing standards for computer science, or maybe even a framework," says Lauryn Wild, commission chair for the state Instructional Quality Commission's executive committee.
Machine-Vision Algorithm Learns to Transform Hand-Drawn Sketches Into Photorealistic Images
Technology Review (06/14/16)
Researchers at Denmark's Radboud University have trained a deep convolutional neural network to convert hand-drawn sketches of faces into photorealistic portraits. They used a dataset of 200,000 faces derived from the Internet, applying standard image-processing algorithms to render the images as line drawings, grayscale sketches, and color sketches. "We found that the line model performed impressively in terms of matching the hair and skin color of the individuals even when the line sketches did not contain any color information," the team reports. They suggest the model not only can leverage luminance differences in the sketches to deduce coloring, but also can learn color properties frequently associated with high-level face features of different ethnicities. The next test assessed the neural net on a dataset using hand-drawn sketches produced in a way the net was not trained on, and it still produced photorealistic portraits, according to the researchers. The net ran into difficulty when pencil strokes in sketches were not accompanied by shading, but "this can be explained by the lack of such features in the training data of the line sketch model," the researchers note. The final test had the net generate photorealistic images of renowned artists based on sketched self-portraits.
Yale Scientists Amplify Light Using Sound on a Silicon Chip
Yale News (06/13/16) Jim Shelton
Yale University scientists have used the power of sound to significantly boost the intensity of light waves on a silicon microchip. The new waveguide system can harness the ability to precisely control the interaction of light and sound waves, and solves the problem of how to utilize this interaction in a robust manner on a silicon chip as the basis for powerful new signal-processing technologies. "The ability to combine both light and sound in silicon permits us to control and process information in new ways that weren't otherwise possible," says Yale professor Peter Rakich. He notes progress in this field has been held back because such hybrid technologies were not efficient enough for practical applications. Rakich says the research overcomes this hurdle using new device designs that prevent light and sound from escaping the circuits. "With precise control over the light-sound interaction, we will be able to create devices with immediate practical uses, including new types of lasers," says Yale researcher Eric Kittlaus. The system is part of a larger body of research the researchers have conducted over the past five years, focused on designing new microchip technologies for light.
Can Computers Do Magic?
Queen Mary, University of London (06/10/16)
Queen Mary University of London (QMUL) researchers have found magicians could use computers to create new magic effects and find new ideas for their performances. They examined modeling particular human perceptual quirks and processes, and building computer systems that can search and find designs for new tricks based on these potential audience responses. "Where computer science and artificial intelligence can help is in conjuring new tricks, which the magician could then perform," says QMUL professor Peter McOwan. The researchers note the Internet and social media platforms provide rich sets of ready-made psychological data about how many people use language in day-to-day life, which could be exploited by computational systems that combine and search large datasets to automatically generate new tricks by analyzing the often ambiguous mental associations people have with particular words in certain contexts. For example, using clusters of words and their associated meanings could enable a magician to predict how a spectator might make connections between seemingly incongruous words in the right context and predict what they might say in a particular situation. "Magicians and trick designers...already use machines as development aids, however we point out that computers also have the potential to be creative aids, generating some aspects of the creative output themselves--though currently in a highly supervised way," says QMUL researcher Howard Williams.
Intelligent Vehicles at the Starting Line for Safer Roads and Improved Traffic Flow
CORDIS News (06/10/16)
The European Union-funded ADAPTIVE project aims to enhance the performance of automated vehicles by developing new functionality to help bolster wider public acceptance of driverless vehicles. The project's researchers recently announced their demonstrator vehicles have been successfully equipped and are ready for test scenarios. ADAPTIVE seeks to optimize the interaction between drivers and automated technologies using a variety of systems, including vehicle-to-vehicle interaction, obstacle sensors, and technologies responding to driver status. The project will test assisted, partial, conditional, and high automation with eight demonstrator vehicles, including city cars, larger passenger cars, and a heavy goods truck. The vehicles are equipped with an electric steering wheel for lateral maneuvers, driveline and brake control for longitudinal maneuvers, forward-looking long-range radar, a forward-looking camera, 360-degree sensors, and modified human-machine interface designs. The testing involves three scenarios. Close-distance testing will include maneuvering for parking or in crowded environments at speeds of under 30 kilometers an hour (about 18 miles per hour). Urban scenarios will involve testing with a range of common traffic hazards at speeds up 10 to 70 kilometers an hour (roughly six to 44 miles per hour), and highway scenarios will have vehicles traveling up to 130 kilometers an hour (81 miles per hour) and will test maneuvers such as lane changes and traffic filtering.
Closing Security Gaps in Internet-Connected Household
Ruhr-University Bochum (Germany) (06/09/16)
Ruhr University Bochum researchers are developing a new method for detecting and fixing vulnerabilities in applications that run on different devices, regardless of the processor used. The researchers note the software running on a device often remains the manufacturer's corporate secret, so they did not analyze the original source code and instead analyzed the binary code of 0s and 1s, which they can read directly from a device. However, different devices are equipped with different complexities. In order to perform processor-independent security analyses, the researchers translated the different binary languages into an intermediate language. This technique enables the researchers to look for security-critical programming errors on the intermediate-language level. The researchers plan to automatically close the gaps they detect. Although this approach does not yet work for any software, the researchers already have demonstrated the method is sound in principle. The method is expected to be completely processor-independent by the time the project concludes in 2020. "Sometimes, it can take a while until security gaps in a device are noticed and fixed by the manufacturers," says Ruhr University Bochum researcher Thorsten Holz.
Service Robots Are Coming to Help Us
National Science Foundation (06/08/16) Kelly Monterroso
The U.S. National Science Foundation's Partnerships for Innovation: Building Innovation Capacity program aims to advance and integrate robotics into people-centered service systems in homes, hospitals, and elder-care facilities. Through the program, industry and university partnerships develop and test their technologies in everyday human interactions. A partnership between the Universities of Louisville and Texas at Arlington and health and defense contractors has designed Adaptive Robotic Nursing Assistants (ARNAs) to assist in the care of long-term patients by fetching items, monitoring patients and vitals, and alerting nurses of any changes. A second iteration of ARNA will walk with a patient and assess a patient's risk of falling en route. A new model of the robot will be tested in Texas hospitals in 2018. Meanwhile, University of Pennsylvania researchers are developing a low-cost robot capable of working with the elderly in elder-care facilities. The robot would help the elderly pick up objects, as well as monitoring their health. "On one level, we will gain knowledge about the design of service robots, both hardware and software, that can perform a limited set of manipulation tasks on elders' behalf," says University of Pennsylvania professor Mark Yim. "At a macro level, the information gathered by these robots and how elders use them during field use will help us learn how robots can help create a larger data-driven, health-monitoring system."
Abstract News © Copyright 2016 INFORMATION, INC.
To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.