Welcome to the September 8, 2021 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."
|
|
|
Didem Unat Is 2021 ACM SIGHPC Emerging Woman Leader in Technical Computing Award Winner
HPCwire
September 2, 2021
Didem Unat of Koç University in Turkey has won the 2021 ACM SIGHPC Emerging Woman Leader in Technical Computing (EWLTC) award. Unat was honored for innovations in the field of programming models for data locality in high-performance and scientific computing, and her leadership role in the international high-performance computing community. Unat's work on simplifying software development for current and future supercomputing architectures resulted in architecture-independent abstractions. These allow for the development of scientific software that maps to complex memory hierarchies and accelerator structures, with high-performance results. Intel's Cristina Beldica, chair of the EWLTC Committee, said, "Unat's rigorous technical work has directly impacted the productivity of application scientists. This is critical not only for high-performance computing, but science in general."
|
Japan Needs a Lot More Tech Workers. Can It Find a Place for Women?
The New York Times
Malcolm Foster
September 2, 2021
Japan is facing a severe shortage of technology workers and engineering students, and efforts are being made to encourage women to enter these fields. Although Japanese girls and boys perform similarly in math and science on international standardized tests, research shows that girls lose interest and confidence in math and science when they reach age 15 and must choose between the science and humanities tracks in high school. UNESCO indicates that women account for 14% of university graduates in engineering programs and 25.8% in the natural sciences in Japan, versus 20.4% and 52.5%, respectively, in the U.S. The nonprofit Waffle offers one-day tech camps for middle school and high school girls in an effort to reverse that trend. Said Waffle's Asumi Saito, "Our vision is to close the gender gap by empowering and educating women in technology."
*May Require Paid Registration
|
Computer Scientists Create Search Systems to Limit COVID-19 Misinformation
University of Waterloo News (Canada)
September 1, 2021
A system developed by computer scientists at Canada's University of Waterloo achieved an 80% increase in accuracy and reliability for searches related to COVID-19 and other health topics. Waterloo's Ronak Pradeep said, "Most of the [search] systems are trained on well-curated data, so they don't always know how to differentiate between an article promoting drinking bleach to prevent COVID-19 as opposed to real health information. Our goal is to help people see the right articles and get the right information so they can make better decisions in general with things like COVID." Researchers augmented their two-stage neural reranking architecture, known as mono-duo-T5, with the label prediction system Vera, which is trained to distinguish between correct and incorrect information. A search protocol relying on data from the World Health Organization and verified information is linked to the system to help rank, promote, and exclude certain online articles.
|
'The Algorithm Fired Me': California Bill Takes on Amazon's Notorious Work Culture
Los Angeles Times
Margaret Roosevelt
August 31, 2021
A bill passed by the California Assembly would challenge Amazon's algorithm-driven warehouse culture. The legislation would require warehouses to detail quotas and work-speed metrics to workers and government agencies; prohibit "time off task" penalties, such as bathroom use, that harm health and safety; and ban retaliation against dissident employees. The bill seeks to address issues of warehouse computer-directed productivity and work-speed demands that have caused serious worker injuries, as well as algorithmically directed terminations based on metrics that Amazon discloses to neither the public nor employees. Opposition to the bill by retailers and other industries is likely to make for a close vote in California's Senate. The University of Illinois at Chicago's Beth Guteliu said, "In the U.S., we are at an inflection point on the question of how technologies are used in workplaces and what rights workers have to data collected about them."
*May Require Paid Registration
|
No Cashiers, Please: Futuristic Supermarket Opens in Mideast
Associated Press
Isabel Debre
September 6, 2021
Retail giant Carrefour has opened the Middle East's first automated cashier-less store in Dubai, United Arab Emirates. The Carrefour City+ mini-market looks like a traditional convenience store but features a system that tracks the movements of customers, who must download the store's smartphone app before entering. The shelves are lined with sensors, and there are almost a hundred surveillance cameras in the ceiling. Shoppers are charged for their purchases five minutes after leaving the store. Said Hani Weiss of Majid Al Futtaim, which operates Carrefour in the Middle East, "We use (the data) to provide a better experience in the future ... whereby customers don't have to think about the next products they want. All the insights are being utilized internally in order to provide a better shopping experience." Weiss stressed that human workers will be needed, at least in the short-term, for customer support and to assist the machines.
|
Imaginary Numbers Protect AI from Very Real Threats
Duke University Pratt School of Engineering
Ken Kingery
August 31, 2021
Computer engineers at Duke University have shown that numbers with both real and imaginary components can be critical in securing artificial intelligence algorithms against threats while preserving efficiency. Including just two complex-valued layers among hundreds if not thousands of training iterations offers sufficient protection. For example, using complex numbers with imaginary components can instill additional flexibility for adjusting internal parameters within a neural network being trained on a set of images. Duke's Eric Yeats said, "The complex-valued neural networks have the potential for a more 'terraced' or 'plateaued' landscape to explore. And elevation change lets the neural network conceive more complex things, which means it can identify more objects with more precision." This enables gradient-regularization neural networks using complex numbers to arrive at solutions just as quickly as those lacking the extra security.
|
Singapore Trials Patrol Robots to Deter Bad Social Behavior
Reuters
Chen Lin
September 6, 2021
Singapore's Home Team Science and Technology Agency announced that a three-week trial will use robot patrols of public areas with high foot traffic in central Singapore to curb poor social behavior, such as ignoring COVID-19 safety measures, smoking in prohibited areas, or improperly parking bicycles. The two robots, called Xavier, are outfitted with cameras and can send real-time alerts to the command-and-control center. The robots will conduct surveillance and display messages on proper behavior. According to the agency, "The deployment of Xavier will support the work of public officers as it will reduce the manpower required for foot patrols and improve operation efficiency."
|
Astronomers Create First 3D-Printed Stellar Nurseries
UC Santa Cruz Newscenter
Tim Stephens
September 2, 2021
Researchers at the University of California, Santa Cruz (UCSC), the Flatiron Institute's Center for Computational Astrophysics, and Harvard University used 3D printing to create models of stellar nurseries. UCSC's Nia Imara said, "We wanted an interactive object to help us visualize those structures where stars form so we can better understand the physical processes." The 3D printing process incorporates the fine densities and gradients of star-forming clouds in a transparent resin, producing highly polished spheres eight centimeters in diameter with nurseries embodied as swirling clumps and filaments. Researchers designed nine simulations of different physical conditions inside molecular clouds, mirroring the effects of turbulence, gravity, and magnetic fields.
|
Smartwatches Track Our Health. Smart Toilets Aren't Far Behind
The Wall Street Journal
Brianna Abbott
September 4, 2021
Researchers and companies are developing smart toilets that go beyond smartwatches and smartphones to monitor patients with chronic conditions or allow consumers to track their health and wellness at home. These devices could alert users to gastrointestinal disease, monitor blood pressure, make dietary recommendations, and more. Stanford School of Medicine researchers have developed a smart toilet that can analyze waste with the help of cameras and a machine-learning algorithm. A smart toilet developed by researchers at Duke University uses sensors to look for blood and certain proteins and can extract stool samples that can be sent to a lab for further analysis. Meanwhile, researchers at the startup Casana are developing a toilet seat capable of measuring vital signs. However, the Stanford researchers found that a third of the 300 people they surveyed were uncomfortable with smart toilets collecting health data, mainly citing privacy concerns.
*May Require Paid Registration
|
Researchers Find Way to Check That Quantum Computers Return Accurate Answers
University of Vienna (Austria)
September 2, 2021
Results of calculations performed on one quantum computer can be verified through a related but fundamentally different calculation on another quantum computer via a new cross-check procedure. An international team of physicists used an alternative quantum computing model based on graph structures, enabling different computations from one source. Martin Ringbauer at Austria's University of Innsbruck said, "While the results may appear random and the computations are different, there are certain outputs that must agree if the devices are working correctly." The team deployed the procedure on five current quantum computers using superconducting circuits, trapped ions, photonics, and nuclear magnetic resonance. Additionally, the method could be used to check a single device against itself. As quantum computers become increasingly available, this technique may be critical for ensuring that they work as advertised.
|
Smart Helmet Rapidly Assesses Stroke Patients
IEEE Spectrum
Michelle Hampson
August 31, 2021
A new smart helmet can detect and diagnose stroke using electromagnetic (EM) waves and a signal-processing algorithm. Researchers at Italy's University of Genoa (UniGe) designed the helmet to confirm the presence of stroke and categorize it as either ischemic or hemorrhagic. The device consists of antennas that are selectively activated to direct EM waves throughout the brain, and the returning signals are quantified. The signal-processing algorithm notifies health professionals of stroke occurrence, after which a more sophisticated algorithm can define stroke type, size, and position. UniGe's Alessandro Fedeli said, "With the signal-processing approach, the overall accuracy is above 80%, which represents an interesting and an encouraging starting point."
|
ML Revolutionizes How Scientists Study Behavior
Carnegie Mellon University
Caroline Sheedy
August 31, 2021
Carnegie Mellon University's Eric Yttri and Alex Hsu have designed an unsupervised machine-learning algorithm to simplify and fine-tune behavioral study. Behavioral segmentation of open field in DeepLabCut (B-SOiD) finds behaviors by identifying patterns in the position of an animal's body, and can tell researchers what behavior is occurring at every frame in a video using computer vision software. Hsu said, "It uses an equation to consistently determine when a behavior starts. Once you reach that threshold, the behavior is identified, every time." Yttri said B-SOiD eliminates user bias as well as time cost and painstaking work, processing hours of data in minutes.
|
Using Adversarial Attacks to Refine Molecular Energy Predictions
MIT News
Vineeth Venugopal
September 1, 2021
Massachusetts Institute of Technology (MIT) researchers found adversarial attacks can improve the reliability of neural networks (NN) in predicting molecular energies by quantifying their uncertainty. Researchers used adversarial attacks to sample molecular geometries in a potential energy surface (PES), and tapped multiple NNs to forecast the PES from the same data. MIT's Rafael Gomez-Bombarelli said, "We aspire to have a model that is perfect in the regions we care about [i.e., the ones that the simulation will visit] without having had to run the full ML [machine learning] simulation, by making sure that we make it very good in high-likelihood regions where it isn't." The technique could be the most realistic approach for exploring the limits of models used to predict material behavior and chemical reactions' progress.
|
|
