Welcome to the November 2, 2020 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."
|
|
NASA Rocket Would Be the Most Powerful Ever, but Officials Worry About Its Software
The Washington Post Christian Davenport October 31, 2020
Officials have concerns about the software controlling the flight of the U.S. National Aeronautics and Space Administration (NASA) Space Launch System (SLS) rocket. NASA's Aerospace Safety Advisory Panel raised issues about the disjointed approach to the system's design and testing; NASA and its contractors had apparently not heeded lessons from the 2019 flight of Boeing's Starliner spacecraft, which failed to dock with the International Space Station due to software errors. A review of the SLS software detected an issue with how the rocket's second stage interpreted data from the first stage, which NASA’s Dan Mitchell said is "benign," and requires no modifications at present.
*May Require Paid Registration
|
Facing Up to the Reality of Politicians' Instagram Posts
UGA Today Cal Powell October 29, 2020
The University of Georgia's Yilang Peng used computer vision to analyze images from U.S. politicians' Instagram accounts to determine which types of posts resonated most. Peng found posts displaying politicians' faces in nonpolitical settings generated more audience engagement than posts showing them in professional or political settings. The analysis involved more than 59,000 images posted on Instagram in the fall of 2018 by 159 U.S. politicians, classifying the setting of the images as professional/political, text/illustration, personal, and architecture/landscape. Among other things, the study found that images in the personal setting category received about 20% more likes than those in the professional and text/illustration categories.
|
Programmable Filament Gives Even Simple 3D Printers Multi-Material Capabilities
IEEE Spectrum Evan Ackerman October 29, 2020
Researchers from Japan’s Meiji and Osaka universities, and Texas A&M University, have developed a technique that enables most three-dimensional (3D) printers to print multiple materials, with no upgrades to the hardware. The method involves first printing a filament from different materials, then using that to print the multi-material object. The researchers were able to print with up to six materials into such a filament. After the last material is added to the filament spiral, a final pass is made by the printer to stitch together all the transition points between the materials. To create the final object, the printed filament is lifted off the print bed, the printer is reset, and the filament is fed into the printer.
|
How Fish-Recognition Tech Is Assisting Demand for Canned Tuna
Bloomberg Anuradha Raghu October 28, 2020
The coronavirus pandemic has spurred remote monitoring of fishing vessels to determine whether tuna catches are sustainable. The Nature Conservancy's Mark Zimring said some vessels are using video cameras, sensors, and systems that use algorithms to recognize different types of marine life, similar to Facebook's facial recognition technology. Satellite imagery, machine learning tools, and artificial intelligence (AI) also are being used to ensure vessels are not misreporting the contents and volumes of their catches and are safely releasing at-risk species caught by accident. The effort initially is being focused on large scale fisheries because the monitoring systems cost $14,000 to $24,000 per vessel per year.
|
Australian, Korean Researchers Warn of Loopholes in AI Security Systems
ZDNet Aimee Chanthadavong October 23, 2020
Researchers at Australia's Commonwealth Scientific and Industrial Research Organization's Data61, the Australian Cyber Security Cooperative Research Center, and South Korea's Sungkyunkwan University warn that certain objects could be used as triggers to permit a subject to digitally disappear from artificial intelligence (AI) security systems. The researchers tested the popular YOLO object-detection camera, and found the camera could detect a subject initially, but putting a red beanie on it allowed it to be undetected by the camera. Data61's Sharif Abuadbba cited the adversarial nature of AI models, which pose a security risk if they are not trained to detect all possible scenarios. Abuadbba said, "If you're a sensitive organization, you need to generate your own dataset that you trust and train it under supervision ... the other option is to be selective from where you take those models."
|
Smartwatch App Alerts d/Deaf and Hard-of-Hearing Users to Birdsong, Sirens, and Other Desired Sounds
University of Washington Sarah McQuate October 28, 2020
A smartphone app developed by researchers at the University of Washington (UW) can notify d/Deaf and hard-of-hearing people of nearby sounds. When a sound of interest to the user is detected by the SoundWatch app, the user is sent a buzz/alert with information about the sound it detected. The watch sends the sound to the user's phone, and the results are then sent back to the watch. Said UW's Dhruv Jain, "This technology provides people with a way to experience sounds that require an action—such as getting food from the microwave when it beeps. But these devices can also enhance people's experiences and help them feel more connected to the world."
|
Path-Planning Algorithm Enables Autonomous Multi-Drone Aerial Surveys of Antarctic Penguin Colonies
Stanford News Taylor Kubota October 28, 2020
Researchers from Stanford University, in collaboration with the National Science Foundation and the U.S. Antarctic Program, deployed a multi-drone imaging system at the McMurdo Station in Antarctica to help survey colonies of about 1 million Adélie penguins. The system generated detailed visual surveys of about 300,000 nesting pairs of Adélie penguins over 2 square kilometers at Cape Crozier and about 3,000 nesting pairs at Cape Royds. Each round of the survey took about 2.5 hours, compared to about two days for the previous human-piloted drone surveys. The faster time can be attributed to a route planning algorithm that coordinated two to four autonomous drones and ensured efficient coverage while limiting backtracking and redundant travel. Said Stanford’s Mac Schwager, “I think that teams of autonomous robots can really be powerful in helping us manage our changing world, our changing environment, at a scale that we never could before.”
|
Tool Simplifies Data Sharing, Preserves Privacy
Carnegie Mellon University College of Engineering Daniel Tkacik October 28, 2020
Researchers in the CyLab of Carnegie Mellon University (CMU) and IBM have come up with a tool for creating synthesized data that simplifies data sharing while maintaining privacy. The DoppelGANger tool employs generative adversarial networks (GANs), which apply machine learning to synthesize datasets with the same statistics as training data. Models trained with DoppelGANger-generated synthetic data had up to 43% greater accuracy than models trained on synthetic data from rival tools, the researchers found. CMU's Vyas Sekar said, "We believe that future organizations will need to flexibly utilize all available data to be able to react to an increasingly data-driven and automated attack landscape. In that sense, any tools that facilitate data sharing are going to be essential.”
|
System Uses Floor Vibrations to Detect Building Occupants
EPFL (Switzerland) Nathalie Jollien October 27, 2020
A building system developed by researchers at the Swiss Federal Institute of Technology (EPFL) uses sensors installed on floor slabs to detect the number of occupants in the building at any given time, and track their movements. Instead of working like other tracking systems that raise privacy issues through their reliance on cameras or occupants' mobile phones, this system uses sensors to measure the vibrations from footsteps to calculate how many people are in a building, their location, and their trajectory. Support vector machines are used to classify signals recorded by the sensors, helping to weed out background noise and identify the footstep characteristics of specific occupants. EPFL's Slah Drira said the system needs only one sensor for every 15 to 75 square meters, and does not require floor slabs to have uniform rigidity.
|
Smart Tablecloth Can Find Fruit, Help with Watering the Plants
Dartmouth College October 29, 2020
Researchers at Dartmouth College, working with colleagues at Microsoft Research, have designed a prototype electrode-laden smart fabric that can detect non-metallic objects by sensing changes in electrical charges. The charge differential reflects the type of material, the object's size, and the shape of the contact area. Information detected on the electrical charge is compared with data stored in the system using machine learning methods. The "smart tablecloth" was used on 20 objects, including a water glass and a bowl to assess its reliability in recognizing the fullness of a container; the system achieved 94.5% accuracy in testing overall, and was especially accurate in differentiating between fruits and different types of liquids.
|
Tiny Brain Implants Hold Big Promise for Immobilized Patients
The Wall Street Journal Daniela Hernandez; Mike Cherney October 28, 2020
Companies and academic labs worldwide are working to develop next-generation devices and artificial intelligence capable of monitoring and decoding brain activity to help people with mobility issues. Rather than rely on devices that require major brain surgery and can cause inflammation over time, researchers hope to access brain activity through noninvasive sensors placed on the skull, in the ear, or in the large blood vessels of the brain. In Australia, University of Technology Sydney's Tara Hamilton said, "There's not too many options out there for allowing someone to have a little more autonomy without major surgery or without having that mental strain of eye tracking. We're still trying to find a better way to interface with the brain."
*May Require Paid Registration
|
The New Fitness Coach Might Not Have a Body
News@Northeastern Laura Castañón October 27, 2020
Computer scientists at Northeastern University have developed a digital fitness coach that can be as effective as a trained, human fitness advisor. Entirely digital, Carmen appears on a screen speaking English or Spanish and asks users short questions about their day, problems, and goals, to build a rapport. The researchers studied 245 Latino participants ages 50 to 87 who said they were not active enough; half met with a human fitness advisor, and half with Carmen. Over the course of a year, those who interacted with Carmen increased time spent walking by 154 minutes per week, compared with 132 minutes for those working with a human coach. Said Northeastern's Timothy Bickmore, "Older adults perhaps have lower computer literacy, less experience with computers in general, but we find that when we present them with something that looks and feels like a face-to-face conversation, that it's much easier for them to use."
|
Translating Lost Languages Using ML
MIT News Adam Conner-Simons October 21, 2020
Researchers at the Massachusetts Institute of Technology (MIT) have developed a machine learning system that can automatically translate a lost language, without advanced knowledge of its relationship to other dialects. The system applies principles based on historical linguistic insights, including the fact that languages generally evolve in certain predictable patterns. MIT's Regina Barzilay and Jiaming Luo developed a decipherment algorithm that can segment words in an ancient language and map them to words in related languages. The algorithm infers relationships between languages, and can assess proximity between languages.
|
|