Welcome to the December 22, 2017 edition of ACM TechNews, providing timely information for IT professionals three times a week.
Please Note: In observance of the upcoming U.S. holiday, TechNews will not be published on Monday, Dec. 25. Publication will resume Wednesday, Dec. 27. 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."
|
|
|
'Listening' Drone Helps Find Victims Needing Rescue in Disasters
Tokyo Tech News
December 22, 2017
Researchers in Japan say they have developed technology capable of distinguishing simultaneous speech from multiple persons, demonstrating simultaneous meal ordering by 11 people and creating a robot game show host that manages multiple contestants answering simultaneously. The technology stems from audition research led by Satoshi Tadokoro of Tohoku University, and it has led to the creation of a system that can detect voices, mobile device sounds, and other noises from disaster victims through the background noise of a drone designed to assist in more rapid victim recovery. The system consists of a microphone array based on robot audition open source software, a three-dimensional sound source location estimation solution with a map display that facilitates assembling an easily understood visual user interface from invisible sound sources, and an all-weather 16-microphone array linked by a single cable for installation on a drone. The last component enables the execution of search and rescue missions even in adverse weather.
|
Proof of Randomness Builds Future of Digital Security
Princeton University
John Schoonejongen
December 21, 2017
Researchers at Princeton University say they have developed a technique for verifying the strength of random number generators on which most encryption systems are based. The researchers note it may be impossible to tell whether a number generator is compromised without examining its source code, as generators are usually tested by analyzing their outputs, which cannot ensure their proper function. Former Princeton researcher Katherine Ye and collaborators wrote proofs in several existing frameworks for verifying programs, and they studied a widely used pseudorandom number generator (HMAC-DRBG). They created a comprehensive and machine-checked proof of its security, meaning its output is sufficiently difficult to differentiate from true randomness. Ye says the results demonstrate the practicality of applying secure tests to other generators. The team's research was presented last month at the ACM Conference on Computer and Communications Security (CCS 2017) in Dallas, TX.
|
Robot Drummer Posts Pictures of Jamming Sessions on Facebook
Queen Mary, University of London
Rupert Marquand
December 20, 2017
Researchers at Queen Mary University of London have developed Mortimer, a drumming robot that plays along with human keyboard players and posts pictures of the sessions on Facebook. Mortimer stems from a study examining how humans interact with robots over time, and how social media can enhance that relationship. The researchers helped trigger a sense of believability by extending Mortimer's capabilities so it could take pictures during recording sessions and post them with a supporting comment to Facebook while also tagging the keyboard player. The team found the time spent with the robot increased over the study, but session length for members in the group who were Facebook friends with Mortimer reduced over time. "There are signs of high engagement, such as high self-reported repeat interaction, across all participants that strengthen previous results about the use of music as a good base for improving long-term human-robot relationships," says Queen Mary professor Peter McOwan.
|
Even Imperfect Algorithms Can Improve the Criminal Justice System
The New York Times
Sam Corbett-Davies; Sharad Goel; Sandra Gonzalez-Bailon
December 20, 2017
Researchers note computer algorithms that U.S. judges increasingly use to determine whether defendants awaiting trial must pay bail or can be released without payment are helping to combat bias and arbitrariness in human decisions. These algorithms provide scores rating a defendant's risk of skipping trial or committing a violent crime if released, bringing a modicum of consistency and evenhandedness to the process. Experiments also have demonstrated the programs' utility in informing sentencing decisions, with one study finding officers could cut back their supervision of individuals deemed low risk without increasing rates of recidivism. These and other examples show concerns of algorithms exacerbating the biases of their creators are mostly unfounded, as long as they are well designed. To avoid biased data's corruption of statistical judgments, the researchers recommend against estimating risk of arrest instead of risk of commission of a crime, and urge following the best statistical practices that avoid unequally predictive risk factors across groups.
*May Require Free Registration
|
73 Percent of Developers Who Don't Use AI Plan to Learn How in 2018
TechRepublic
Alison DeNisco Rayome
December 20, 2017
Although only 17 percent of software developers worked with artificial intelligence (AI) or machine learning this year, 73 percent of those who did not say they are planning to learn about the technologies next year, according to a DigitalOcean survey of 2,500 developers. With major companies adding AI to their products for developers to more easily integrate data, it is becoming increasingly vital for developers to familiarize themselves with these technologies. The DigitalOcean poll also found many of the AI-using developers worked with the technology with TensorFlow, sentiment analysis, image recognition, and natural-language processing. Significant AI-related issues respondents expect to be confronted with in 2018 include workflow automation and incorporating AI and machine learning within the business. Meanwhile, 89 percent of respondents listed Linux as their preferred operating system, and PHP was rated the programming language of choice by 27 percent.
|
Researchers Fooled a Google AI Into Thinking a Rifle Was a Helicopter
Wired
Louise Matsakis
December 20, 2017
Researchers at the Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory say they have successfully and reliably tricked a popular Google image-recognition algorithm into mistaking a rifle for a helicopter. The team notes they ran their experiment under "black box" conditions, with little insight into the algorithm's underlying mechanics, by designing a method for rapidly finding a way to generate black-box adversarial examples that can deceive different algorithms. They targeted the component of the Google algorithm that labels objects; each time the team attempted to fool the algorithm, they analyzed the outcomes and gradually tweaked the image until it could trick a computer into wrongly tagging one object as another. "There's no bias, we didn't choose what was easy," notes MIT's Anish Athalye. The researchers say their work demonstrates that hackers could potentially create adversarial examples capable of fooling commercial artificial intelligence systems.
|
Hidden Bridge Between Quantum Experiments and Graph Theory Uncovered Using Melvin
University of Vienna (Austria)
Alexandra Frey
December 19, 2017
Researchers at the University of Vienna in Austria and the Austrian Academy of Sciences have identified a deep connection between experimental quantum physics and the mathematical field of graph theory. The researchers say they made their breakthrough while studying the Melvin algorithm's unusual solutions. Melvin was developed for calculating technical solutions for quantum physical experiments. However, during the initial analysis of the solution calculated by Melvin, the researchers did not understand how it actually works. Eventually, they found a sequence of numbers known only in graph theory, which has been used to describe networks such as the Internet or neural networks. After further investigation, the researchers found many similarities between experimental quantum physics and mathematical graph theory. The University of Vienna's Mario Krenn says the research makes it possible to grasp quantum technology as a graph or a network in order to explore new experimental possibilities.
|
New Research Streamlines Machine Learning
MIT News
December 19, 2017
Researchers at the Massachusetts Institute of Technology and Michigan State University (MSU) say they have developed the open source Auto-Tuned Models (ATM), an automated machine-learning system that can sometimes outperform humans at 100 times the speed. ATM exploits on-demand cloud computing to execute a high-throughput search over modeling options to find the best possible solution for specific problems, while also tuning the model's hyperparameters. Thousands of models are tested in parallel, with more computational resources allocated to the most promising techniques. Testing ATM against users of the openml.org collaborative crowdsourcing platform found ATM's analysis of 47 datasets led to a solution better than the human-derived one 30 percent of the time. Meanwhile, its delivery of a near-optimal solution took less than a day, versus the human average of 100 days. MSU professor Arun Ross says ATM displays results as a distribution, thus accelerating testing and comparison of different modeling approaches without excluding human intuition.
|
Blue Waters Supercomputer Processes New Data for NASA's Terra Satellite
National Center for Supercomputing Applications
Kristin Williamson
December 19, 2017
Researchers at the University of Illinois at Urbana–Champaign's (UIUC) National Center for Supercomputing Applications used the Blue Waters supercomputer to process data from the U.S. National Aeronautics and Space Administration's (NASA) Terra satellite and create visualizations showing how the Earth has changed over 20 years. The researchers say the visualizations will showcase how the satellite samples the Earth in a visually compelling and informative way, and how the different instruments on the satellite are combined to better document how the Earth has changed. NASA's Terra data archive is about 1.2 petabytes, and Blue Waters is one of the only computers that can process that data on such a large scale in a timeframe suitable for scientists to interact and ask questions of the data, says UIUC professor Larry Di Girolamo. He notes the data is taken from five instruments that perform coincident measurements of the Earth's system and each carries out a different mission.
|
Scientists Hope to Inject Robo-Cat With AI to Help Seniors
Associated Press
Michelle R. Smith
December 19, 2017
Researchers at Brown University have received a three-year, $1-million U.S. National Science Foundation grant to study ways to add artificial intelligence (AI) to Hasbro's "Joy for All" robotic cat. The project, called Affordable Robotic Intelligence for Elderly Support (ARIES), is aimed at developing additional capabilities for the robotic cats to help older adults with simple tasks. The researchers, working out of Brown's Humanity-Centered Robotics Initiative, want to determine which tasks make the most sense to help older adults stay in their own homes longer. The robotic cat can be a tool that could make life easier for someone caring for a person with dementia, or to be used in nursing homes where pets are not allowed. The first step is conducting surveys, focus groups, and interviews to understand how older adults live. The researchers are considering how the robotic cat could help complete everyday tasks, and how it would communicate with its owner.
|
Transformational Advances in Quantum Systems
CCC Blog
Helen Wright
December 18, 2017
In a recent Dear Colleague Letter (DCL), the U.S. National Science Foundation notes it identified 10 Big Ideas for Future Investment, two of which concern the multi-pronged Quantum Leap initiative to advance basic knowledge of quantum phenomena, materials, communications, and systems, and the Convergent Research effort to integrate ideas and approaches from widely diverse disciplines. The letter encourages the submission of interdisciplinary research projects that must include at least three complementary components represented by researchers with expertise in the areas of physics, chemistry, mathematics, materials science, engineering, and computer/computational science. The proposals are required to focus on quantum functionality by evaluating aspects relevant to both fundamental and application concepts, which lead to experimental demonstrations of transformative innovations toward quantum systems and/or proof-of-concept validations. The DCL also says it "encourages various aspects of quantum communication and quantum computing" to address the need for when systems-level designs for networks including quantum components are made with higher levels of abstraction.
|
Google AI Top Scientist Calls China 'Important Country' for AI
Xinhua
Xiong Maoling; Liu Si; Huang Kun
December 20, 2017
In an interview, Google Cloud researcher Fei-Fei Li pushes for enhanced collaboration between major countries on artificial intelligence research, with China especially cited as a rising AI power. Li recently announced Google will open the Google AI China Center in Beijing, noting, "We all recognize the importance of China because of its talent, because of the incredible creativity and innovation that are already going on here." Li also says China's AI efforts are accelerating via government mandate, and notes the country has "more prominent global responsibility in technology, politics, and culture." She believes cooperation between key nations on AI will eventually yield "extraordinary benefits" for all of humanity. Li also notes although AI is seen by some as a source of international rivalry, her hope is that more transnational collaboration and communication will be fostered by the notion of scientific progress as a universal pursuit.
|
Science Application Developers Share Pioneering Strategies for Exascale Computing
HPCwire
Jonathan Hines
December 19, 2017
In an interview, three U.S. Department of Energy (DoE) scientists discuss providing scalability and performance portability to prepare for exascale computing. The Oak Ridge Leadership Computing Facility's Tjerk Straatsma stresses performance portability for transferring research between different facilities, and he says sharing these programming strategies is vital as DoE's latest generation of supercomputers is set up at labs. The National Energy Research Scientific Computing Center's Katerina Antypas cites the need for applications to leverage three central technologies to realize good performance on exascale computers--longer vector units, high-bandwidth memory, and numerous low-powered cores. The Argonne Leadership Computing Facility's Timothy Williams notes developers/computational scientists and experimentalists and theorists are becoming increasingly codependent. In this context, Straatsma says pioneering exascale research should benefit the scientific community via large computer-enabled applications and methods "that can be translated to other codes or other application domains and be used to make these applications run very well on these new architectures."
|
|
