Welcome to the February 12, 2021 edition of ACM TechNews, providing timely information for IT professionals three times a week.
Please note: In observance of the U.S. Presidents' Day holiday, TechNews will not be published on Monday, Feb. 15. Publication will resume Wednesday, Feb. 17. 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."
|
|
|
Pay Gap Widens Between Female, Male Scientists in North America
Nature
Chris Woolston
February 11, 2021
The pay gap between male and female scientists with Ph.D.s in North America is widening, according to research by the U.S. National Science Foundation (NSF) and the Council of Canadian Academics. The NSF polled over 55,700 Ph.D. recipients between July 1, 2018 and June 31, 2019, and found roughly 35% reported having a permanent position lined up at graduation, while 38% planned to assume postdoctoral research jobs. Men with permanent CS positions reported an expected median salary of $125,000, versus $101,500 for women. The Canadian report found male Ph.D. holders with full-time jobs had an average annual salary of slightly more than C$123,500 (about US$97,000) in 2016, while women reported an average annual salary of just more than C$107,500 (about US$84,000).
|
Google Search Can Help Predict Covid-19 Spikes
The Jerusalem Post
Zachary Keyser
February 8, 2021
Researchers from the U.K.'s University College of London (UCL) and Israel's Bar-Ilan University have found that spikes in Covid-19 cases could be predicted an average of 17 days in advance using online search data. The researchers developed an analytical model that compares Covid-19 symptom-related Google searches against spikes in Covid-19 cases. Symptoms identified by the U.K.'s National Health Service and Public Health England are used by the model and weighted according to how frequently they occurred in confirmed Covid-19 cases. UCL's Vasileios Lampos said, "We have shown that our approach works on different countries irrespective of cultural, socioeconomic, and climate differences. Our analysis was also among the first to find an association between Covid-19 incidence and searches about the symptoms of loss of sense of smell and skin rash."
|
Can AI Tackle Racial Inequalities in Healthcare?
BBC News
Cody Godwin
February 6, 2021
University of California, Berkeley (UC Berkeley) researchers used artificial intelligence (AI) to analyze x-rays to "predict patients' experienced pain" for osteoarthritis of the knee. UC Berkeley's Ziad Obermeyer said the algorithm predicted what the patient would say about their own experience of knee pain, and could detect things radiologists might overlook. How doctors evaluate pain has been associated with discrimination and even racism, as studies determined that physicians tend to underestimate self-perception of pain by certain groups. Radiologists examining seemingly similar arthritis cases said blacks reported more pain than whites, yet the UC Berkeley algorithm factored in additional undiagnosed features to indicate cases were less similar than they appeared. Emory University's Sandra Hobson said, "I think AI has an opportunity to help incorporate data, including patients from all backgrounds, all parts of the country around the world, and help make sense of all that data together."
|
Big Jump in RDP Attacks as Hackers Target Staff Working From Home
ZDNet
Danny Palmer
February 8, 2021
Researchers at cybersecurity company ESET said Remote Desktop Protocol (RDP) attacks climbed 768% last year as many employees worked from home amid the Covid-19 pandemic. ESET identified 29 billion attempted RDP hacks during 2020, with increased remote work offering malefactors a greater opportunity to infiltrate networks undetected, using legitimate login credentials. ESET's Ondrej Kubovic said RDP ports also can be misconfigured, which "in many cases leads to valuable resources, such as company servers or devices with admin rights, that represent a springboard for further, often network-wide, compromises." Actions that organizations can take to reduce the odds of successful RDP attacks include encouraging users to use strong passwords that are difficult to guess via brute-force attacks, and applying two-factor authentication network-wide. Another approach is to guarantee users are using the latest versions of operating systems and software through patching.
|
Sci-Tech Oscar Goes to 3 Former Computer Science Ph.D. Students From Saarbrücken
Saarland Informatics Campus (Germany)
February 10, 2021
The U.S. Academy of Motion Picture Arts and Sciences will confer this year's Technical Achievement award (Sci-Tech Oscar) to three former computer science Ph.D. students from Germany's Saarland University. Sven Woop, Carsten Benthin, and Ingo Wald have been cited for co-developing or conducting preparatory research on the Intel Embree Ray Tracing Library, which has been used to maximize photorealism in virtual game environments and animated films, including international blockbuster movies. Said Saarland's Philipp Slusallek at the German Research Center for Artificial Intelligence (AI). He said, "We use real-time ray tracing today to generate synthetic sensor data very quickly, which can be used to train and validate AI systems."
|
Breakthrough in Quantum Photonics Promises New Era in Optical Circuits
USC Viterbi School of Engineering
Greta Harrison
February 3, 2021
University of Southern California (USC) researchers have taken a step toward quantum optical circuits, demonstrating uniform emission of single photons from quantum dots arranged in a precise configuration. The team used quantum-dot alignment methods pioneered by USC's Anupam Madhuka to generate single-quantum dots. The researchers fabricated regular arrays of nanometer-sized mesas with a defined edge orientation, shape, and depth on a flat semiconductor substrate of gallium arsenide, then created quantum dots atop the mesas by adding appropriate atoms. USC's Jiefei Zhang said the resulting uniformity enables usage of established techniques like local heating or electric fields to refine the quantum dots' photon wavelengths to match, which is crucial for interconnections between different quantum dots for circuits. Said Zhang, "We now have an approach and a material platform to provide scalable and ordered sources generating potentially indistinguishable single-photons for quantum information applications."
|
U.S. IT Jobs Grow in January, Defying Broader Employment Struggles
Computerworld
Galen Gruman
February 8, 2021
Information technology (IT) employment consultancy Janco Associates estimates that seasonally adjusted U.S. IT job growth in January was 18,200, despite the overall economy's difficulties. A total of 55,000 new IT jobs were posted last month, up from 18,000 in December, based on revisions from the Bureau of Labor Statistics. However, U.S. IT jobs have declined by 35,800 year-over-year for a loss of about 1%. Meanwhile, CompTIA reported 78,000 new IT-related jobs across all industry sectors last month, reflecting a 2.4% unemployment rate versus 3% in December. Janco's M. Victor Janulaitis anticipates 11% growth in U.S. IT jobs through the next decade, mostly concentrated among software developers, quality assurance, and testers, and driven by work from home.
|
A Language Learning System That Pays Attention—More Efficiently Than Ever Before
MIT News
Daniel Ackerman
February 10, 2021
Massachusetts Institute of Technology (MIT) researchers have designed a specialized software-hardware combination called SpAtten to run an attention mechanism for more efficient natural language processing (NLP). MIT's Hanrui Wang said SpAtten operates similarly to the human brain by reading quickly and focusing on keywords. SpAtten's efficiency comes from cascade pruning, the elimination of unnecessary data like keywords (tokens) and computation branches (heads) from calculations. SpAtten also conducts progressive quantization by wielding data in smaller bitwidth chunks and minimizing those retrieved from memory, while its hardware uses parallelism to rank the importance of tokens and heads for potential pruning in a small number of computer clock cycles. Said Wang, "Our vision for the future is that new algorithms and hardware that remove the redundancy in languages will reduce cost and save on the power budget for datacenter NLP workloads."
|
Algorithm May Be the Key to Timely, Inexpensive Cyber Defense
Penn State News
Matt Swayne
February 3, 2021
A team of researchers led by The Pennsylvania State University (Penn State) has developed an adaptive cyber defense against zero-day attacks using machine learning. The new technique offers a powerful, cost-effective alternative to the moving target defense method used to detect and respond to cyberattacks. Reinforcement learning enables the decision maker to learn to make the right choices by choosing actions that maximize rewards. Said Penn State's Peng Liu, "The decision maker learns optimal policies or actions through continuous interactions with an underlying environment, which is partially unknown. So, reinforcement learning is particularly well-suited to defend against zero-day attacks when critical information—the targets of the attacks and the locations of the vulnerabilities—is not available."
|
As Robots Fill the Workplace, They Must Learn to Get Along
Wired
Will Knight
February 2, 2021
Singapore's Changi General Hospital is using software from Open Robotics to avoid standoffs among its robots in hallways or outside elevators. The hospital's approximately 50 robots are from eight manufacturers, and the software enables robots from different manufacturers to communicate and negotiate safe passage around each other. Open Robotics maintains the Robot Operating System (ROS) open-source software, but the software being used by Changi allows communication among robots not based on ROS. The company hopes for more widespread adoption of such free, easily modified software to facilitate greater interoperability of workplace robots. ABI Research's Rian Whitton said companies have been able to develop mobile robots more quickly thanks to ROS, and interoperability could accelerate through an open source platform.
|
Chemistry, Computer Science Join Forces to Apply AI to Chemical Reactions
Princeton University
Wendy Plump
February 4, 2021
Chemists and computer scientists at Princeton University collaborated on the development of open-source machine learning software that uses artificial intelligence to optimize chemical synthesis reactions. The software adapts key principles of Bayesian optimization to enable faster, more efficient chemical synthesis. Said lead author Benjamin Shields, who created the Python package, "In designing the software, I tried to include ways for people to kind of inject what they know about a reaction. No matter how you use this or machine learning in general, there's always going to be a case where human expertise is valuable."
|
Dynamic 3D Printing Process Features Light-Driven Twist
Northwestern McCormick School of Engineering
Megan Fellman
February 3, 2021
Researchers at Northwestern University have a developed a three-dimensional (3D) printing technique that uses a liquid photopolymer activated by light and a high-precision robotic arm that allows each layer to be moved, rotated, or dilated as the structure is being built. Said Northwestern's Cheng Sun, "Now we have a dynamic process that uses light to assemble all the layers but with a high degree of freedom to move each layer along the way.” The continuous printing process allows 4,000 layers to be printed in about two minutes. The researchers used their method to 3D-print a customized vascular stent, a soft pneumatic gripper made of one hard and one soft material, a double helix, and a mini Eiffel Tower.
|
Labels in Computer Vision Datasets Poorly Capture Racial Diversity
VentureBeat
Kyle Wiggers
February 8, 2021
A study by Northeastern University researchers posits that labels in computer vision datasets are unreliable as identity indicators. In analyzing the FairFace, BFW, RFW, and LAOFIW datasets, the authors determined that racial labels are used in computer vision either without definition, or with only loose or vague definitions. The Northeastern researchers also cited some datasets' usage of the label "Indian/South Asian" as an example of the shortcomings of racial categories. The authors wrote, "One can consider extending the number of racial categories used, but racial categories will always be incapable of expressing multiracial individuals, or racially ambiguous individuals." The Northeastern team also found faces in the analyzed datasets were systematically subject to racial disagreements among annotators, and warned these facial labeling biases could be replicated and amplified if left unresolved.
|
|
