Welcome to the February 22, 2017 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).
|
China's First 'Deep Learning Lab' Intensifies Challenge to U.S. in Artificial Intelligence Race
South China Morning Post (Hong Kong)
Meng Jing
February 21, 2017
China's government has authorized the establishment of the country's first "national laboratory for deep learning," with the goal of exceeding the U.S.'s artificial intelligence (AI) accomplishments. The lab will be a virtual online research network instead of a physical facility, with Baidu commissioned by the National Development and Reform Commission to spearhead the lab's development. Baidu confirms the lab's research areas will include machine-learning-based visual recognition, voice recognition, novel human-machine interaction, and deep-learning intellectual property. "As an open platform itself, the national lab will help more Chinese researchers, companies, and universities to access the most advanced AI technologies in China," says Yu Kai, the past head of Baidu's deep-learning institute. Among the lab's leaders will be Baidu's current deep-learning institute chief Lin Yuanqing and scientist Xu Wei. Baidu already has made significant progress with AI technologies such as its autonomous driving systems and chatbots.
|
Japan Announces AI Supercomputer
Scientific Computing World
February 20, 2017
Japan's Tokyo Institute of Technology (Tokyo Tech) has announced plans to build TSUBAME3.0, a new supercomputer designed to accelerate artificial intelligence (AI) research. TSUBAME3.0 will use NVIDIA's Pascal-based Tesla P100 graphical-processing units to deliver about 12.2 petaflops, more than twice the performance of TSUBAME2.5, its predecessor. TSUBAMA3.0 is based on AI computation and is expected to deliver more than 47 PFLOPS of AI data-analyzing processing capability. In addition, the new system could be operated concurrently with the TSUBAME2.5 to upgrade performance to about 64.3 PFLOPS, making it Japan's highest-performing supercomputer for AI applications. The TSUBAME3.0 system will be used for education and high-technology research at Tokyo Tech, and it will be accessible to other private-sector researchers. The system also will serve as an information infrastructure center for leading Japanese universities.
|
Will the Public Accept the Fatal Mistakes of Self-Driving Cars?
The Washington Post
Steven Overly
February 20, 2017
U.S. policymakers are wrestling with whether people will accept autonomous vehicles, despite their potential for fatal errors. "The artificial intelligence systems on which autonomous vehicle technology will depend are presently and unavoidably imperfect," notes Toyota Research Institute CEO Gill Pratt. "So, the question is 'how safe is safe enough' for this technology to be deployed." Members of Congress are starting to weigh measures to facilitate wider adoption of self-driving autos without compromising safety, while the U.S. Department of Transportation published guidelines for autonomous vehicles to keep up with automakers' plans to deploy them in a few years. "We penalize and distrust [machines] more [than people] when they make mistakes" because people have no proper mental models of what machines can and cannot do, notes Massachusetts Institute of Technology professor Iyad Rahwan. He says the issue may come down to how much humans trust self-driving vehicles, no matter how many lives they can save.
|
Virtual Reality Game Could Help Early Detection of Schizophrenia
CORDIS News
February 17, 2017
Researchers working on the European Union's ALTEREGO project have developed artificial intelligence (AI) technology that can interact with patients suffering from disorders such as schizophrenia. The researchers say the AI technology could provide a reliable, adaptable, and affordable way of diagnosing, monitoring, and treating schizophrenic patients. ALTEREGO is a "mirror game" in which patients mimic a computer avatar. The game can detect schizophrenia patients accurately by identifying and analyzing user impairments in movement and social interaction. ALTEREGO employed statistical-learning techniques to analyze the data, using multiple features and applying a majority vote rule. The researchers found the program distinguished between schizophrenia patients and control groups with 93-percent accuracy and 100-percent specificity. The mirror test is more effective than other motor-assessment tools because it is quicker, less expensive, and less invasive. The researchers suggest the gamification of the task may make ALTEREGO a useful diagnostic tool for children.
|
Power Efficient Electronic Innovations Recognized at Prestigious ISSCC 2017 Conference
University of Southampton
February 17, 2017
Researchers and engineers from around the world assembled at the IEEE International Solid-State Circuits Conference earlier this month with a focus on the theme of "Intelligent Chips for a Smart World." Researchers from the University of Southampton in the U.K. presented two papers at the Analog Techniques and Harvesting and Wireless Power sessions. The first paper demonstrated an oscillator design comparable to state-of-the-art technology for watts-per-kilohertz as well as exhibiting temperature and voltage stability. The design has been validated as part of the Pipistrelle test-chip family. The second Southampton paper described a self-tuning resonant inductive link transmit driver with phase-switched fractional capacitance. "It is testament to the cutting-edge work led by our academics...that two papers have been accepted into one of the most respected international conferences for electronics design," says Southampton professor Bashir Al-Hashimi.
|
Computer Crashes May Be Due to Forces Beyond Our Solar System
IDG News Service
Megdalena Petrova
February 17, 2017
Vanderbilt University professor Bharat Bhuva on Friday gave a presentation on single-event upsets (SEUs) at the annual meeting of the American Association for the Advancement of Science. SEUS occur when cosmic rays and the electrically charged particles they generate interfere with the operation of the microelectronic circuitry in electronic devices. Bhuva says during an SEU, particles alter an individual bit of data stored in a chip's memory, which can have effects ranging from altering a single pixel in a photograph to bringing down a passenger jet. Although there are serious examples of SEUs causing significant damage, Bhuva notes they are still fairly rare events. However, he says as the number of transistors being used in new electronic systems increases, so does the probability of an SEU failure on the device level. "This is a really big problem, but it is mostly invisible to the public," Bhuva warns.
|
New Internet Security Device Launched to Safeguard Schools Against Child Abuse
Plymouth University
Andrew Merrington
February 20, 2017
Researchers at the University of Plymouth in the U.K. have developed ICAlert, a new device that provides constant monitoring against online child abuse and radicalization for primary and secondary schools. The device plugs directly into a school's network, monitors Web traffic, and generates an alert if there is any attempted access of illegal material. "ICAlert offers long-term, effective monitoring of a school's network and is flexible enough to deal with increasing Internet traffic or number of users," says Plymouth professor Bogdan Ghita. ICAlert, after being plugged into a school's network, requires no additional management by the school and receives periodic updates, which include the most recent list of banned Web content links. In addition, the researchers note the device provides an alternative solution to filtering, in which access to banned sites is simply denied and the attempt to perpetrate an offense goes unnoticed.
|
How We Could Close Tech's Gender Gap in a Decade
The Washington Post
Hayley Tsukayama
February 17, 2017
The inaugural Girls Who Code Governor's Summit in California last week emphasized solving the U.S. technology industry's lack of gender parity by developing standards and strategies for quantifying efforts to support women and girls interested in tech and science. A 2016 Accenture/Girls Who Code study estimated women comprise only 18 percent of computer science graduates. Girls Who Code founder Reshma Saujani sees this as a call to mobilize and step up her push to find the roots of this decline in collaboration with state governments. Saujani says one underlying reason could be girls not finding any science-interested peers or role models in high school, while others believe cultivating an interest in science should start at a much earlier age. Saujani thinks the tech gender gap could be closed within 10 years, if proponents direct their energy toward collecting solid data, setting metrics, and following through on policy goals.
|
Can Artificial Intelligence Predict Earthquakes?
Scientific American
Annie Sneed
February 15, 2017
Paul Johnson's research team at Los Alamos National Laboratory is applying artificial intelligence to earthquake prediction, using machine-learning algorithms, supercomputers, and big data storage and analysis. The team is feeding machines vast datasets of measurements taken continuously before, during, and after lab-simulated earthquakes. The algorithm mines the data for patterns that reliably signal when an artificial quake will occur. The team also has started applying the machine-learning analysis to raw seismic data from actual quakes. Via experiments run on Pennsylvania State University's earthquake simulator, the researchers found the algorithm detected reliable "creaking and grinding" noises that change in a very specific way as the artificial tectonic system gets closer to a simulated quake. "Not only could the algorithm tell us when an event might take place within very fine time bounds--it actually told us about physics of the system that we were not paying attention to," Johnson says.
|
Tiny 3D-Printed Camera Lens Could Give Drones Vision Like Ours
New Scientist
Matt Reynolds
February 15, 2017
Researchers at the University of Stuttgart in Germany created a camera that combines four three-dimensionally (3D)-printed lenses to mimic natural vision, with extremely sharp focus in a central area and wider peripheral vision. The four 3D-printed lenses have different focal lengths, with the longer ones capturing high detail over a narrow field of view and the lenses with shorter focal lengths capturing low detail over a wider field of view. The researchers used software to combine the images and create a circular photo that is highly detailed in the center, but becomes less detailed toward the edges. The entire system measures less than 300 micrometers square, and would have been impossible to create without 3D printing, according to Stuttgart researcher Harald Giessen. A camera using such a lens would require less power and processing time than conventional cameras because only the center of the image requires intensive processing.
|
Breakthrough in 'Wonder' Materials Paves Way for Flexible Tech
University of Warwick
Luke Walton
February 15, 2017
Researchers at the University of Warwick in the U.K. say they have developed a new technique to measure the electronic structures of stacks of two-dimensional (2D) materials for the first time. Multiple stacked layers of 2D materials, known as heterostructures, create highly efficient optoelectronic devices with an ultrafast electrical charge, which can be used in nano-circuits. The researchers say the materials also are stronger than materials used in traditional circuits. The new technique uses the photoelectric effect to directly measure the momentum of electrons within each layer and shows how this changes when the layers are combined. The researchers say the technology paves the way for the development of highly efficient nano-circuitry, and smaller, flexible, more wearable devices. In addition, solar power could be revolutionized with heterostructures because the atomically thin layers permit strong absorption and efficient power conversion with a minimal amount of photovoltaic material.
|
Big Improvements to Brain-Computer Interface
SDSU NewsCenter (CA)
Michael Price
February 15, 2017
The new glassy-carbon electrodes developed by the Center for Sensorimotor Neural Engineering (CSNE) have the potential to restore movement in people with spinal cord injuries. Brain-computer interface technology uses electrodes to record brain signals at the moment a person intends to make a movement, enabling the interface to learn the signal pattern for that movement and transmit the pattern to the limb's nerves. The current material used in these devices' electrodes is thin-film platinum, which fractures over time. The new electrodes are made out of glassy carbon, which is about 10 times smoother than thin-film platinum. In addition, this form of carbon corrodes less easily under electrical stimulation, lasts longer, and produces a much clearer signal than platinum or other metal electrodes. CSNE researchers are using the improved brain-computer interface to record neural signals both along the brain's cortical surface and from deeper inside the brain at the same time.
|
Leveraging the Wisdom (and Ignorance) of Crowds
Government Computer News
Susan Miller
February 14, 2017
The U.S. Intelligence Advanced Research Projects Activity (IARPA) has announced funding to support large-scale collaboration methods that will improve intelligence analysts' ability to provide accurate and timely analyses of complex issues. The Crowdsourcing Evidence, Argumentation, Thinking, and Evaluation (CREATE) program awards contracts to projects that combine crowdsourcing with structured techniques. One CREATE awardee is the Smartly-assembled Wiki-style Argument Marshalling (SWARM) project from the University of Melbourne in Australia. SWARM is a cloud-based platform that uses algorithms to build a statistical summary of a user's reasoning strengths and biases. Meanwhile, the Trackable Reasoning and Analysis for Collaboration and Evaluation (TRACE) is a Web-based application from Syracuse University that uses crowdsourcing to improve the division of labor and reduce errors. Finally, George Mason University's Cogen Argumentation System with Crowd Elicitation (Co-Arg) is a software-based cognitive assistant that evaluates hypotheses, evidence, and facts.
|
|
