Welcome to the September 26, 2022, 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."
|
|
LinkedIn Ran Social Experiments on 20 Million Users over Five Years
The New York Times Natasha Singer September 25, 2022
The professional networking platform LinkedIn ran experiments on more than 20 million users over five years that could have impacted some people's livelihoods, according to a multi-institutional analysis. LinkedIn randomly altered the prevalence of weak and strong contacts suggested by its "People You May Know" algorithm without informing users. Researchers observed that relatively weak social ties on the platform were twice as effective in obtaining employment as stronger social ties, but the LinkedIn experiment has drawn some criticism for its potential to alter lives. The goal of the experiment was to roll out a better algorithm for everyone’s jobs prospects "rather than anointing some people to have social mobility and others to not," said Massachusetts Institute of Technology's Sinan Aral, lead author of the study.
*May Require Paid Registration
|
Detecting Earliest Population Changes of COVID-19 Activity with Wearable Devices
Scripps Research Institute September 23, 2022
Wearable devices can be used to identify early population changes in COVID-19 activity, according to researchers from the Scripps Research Translational Institute, the Rockefeller Foundation, and software company CareEvolution. The researchers employed a mobile application to collect sensor and self-reported information from 39,931 participants across the U.S. They assessed factors such as average resting heart rate and step count data and detected "anomalous" sensor data if someone had a higher resting heart rate or lower step count compared to baseline data. This data substantially improved the seven-day average COVID-19 infection forecasts in California as well as the 12-day national predictions.
|
A New Kind of Number Improves the Math of AI
IEEE Spectrum Dina Genkina September 25, 2022
Scientists at Spain's Complutense University of Madrid have developed a processor core that can deploy the posit standard, a numerical representation that improves over standard floating-point arithmetic processors, in hardware. Complutense researchers synthesized the hardware in a field-programmable gate array to compare computations performed using 32-bit floats alongside 32-bit posits. Posits exhibited a four-order-of-magnitude improvement in matrix multiplication accuracy without losing computation time. "It's possible that posits will speed up [artificial intelligence] training because you're not losing as much information on the way," said Complutense's David Mallasén Quintana.
|
How to 3D-Print One of the Strongest Stainless Steels
U.S. National Institute of Standards and Technology September 22, 2022
Scientists from the U.S. National Institute of Standards and Technology, the University of Wisconsin-Madison, and the Department of Energy's Argonne National Laboratory have 3D-printed a strong stainless steel. The 17-4 precipitation hardening stainless-steel alloy identified by the researchers retains the properties of its conventionally manufactured iteration. The team used high-energy X-rays from a particle accelerator to extract high-speed data about the printing process. They bombarded steel samples with X-rays during printing and mapped how the crystal structure changed, gaining insights into controllable factors.
|
Software Could Soon Watch Over Your Brain while You Rest in Intensive Care
ScienceAlert David Nield September 23, 2022
Algorithms could be trained to monitor changes in the consciousness of critically ill patients in intensive care, according to a study by computer scientists at the Stevens Institute of Technology. Researchers built an algorithm to monitor feeds from heart rate, breathing rate, brain temperature, blood oxygen level, and other sensors. A study of 239 patients with hemorrhages found that the algorithm could classify patients' states of consciousness as well as trained clinicians and nearly as well functional magnetic resonance imaging machines and other equipment. "If you only check patients once per day, you just get one data point. With our algorithm, you could track consciousness continuously, giving you a far clearer picture," said Stevens Institute of Technology's Samantha Kleinberg.
|
Soft Robots that Grip with the Right Amount of Force
MIT News Rachel Gordon September 22, 2022
A robotic system that can grasp tools and exert the right amount of force for a task has been developed by researchers at the Massachusetts Institute of Technology (MIT) and the Toyota Research Institute. The Series Elastic End Effectors (SEED) system, which comprises soft bubble grippers and embedded cameras, maps how the grippers deform over a six-dimensional space and apply force to a tool. 3D images from the grippers enable the system to reconstruct the tool's position. A learned model, based on the robot's previous experience determining the grippers' stiffness by disturbing a force torque sensor, maps the tool's position to the measured force.
|
Spoofing Cyberattack Can Make Cameras See Things that Are Not There
New Scientist David Hambling September 26, 2022
Using radio waves, Sebastian Köhler at the U.K.'s University of Oxford and colleagues were able to trick image-recognition systems into seeing nonexistent things. Digital cameras include sensors that render light as electrical impulses, and post-transducer signals can create the false impression of actual images. The researchers fooled a barcode scanner from 0.5 meters (1.6 feet) away, adding noise to photos the scanner captured and inducing failure 99% of the time. Köhler said his team has introduced shapes in such attacks like readable text, while more refined exploits to deceive object-recognition software into seeing unreal things also are feasible. "An attack from tens of meters is possible with reasonably sized hardware," Köhler said.
*May Require Paid Registration
|
Designing Quantum Materials on the Computer
Technical University of Wien September 19, 2022
An international team including researchers from Rice University and Austria's Technical University of Vienna (TU Wien) have developed a theoretical method for identifying novel materials with special electronic properties necessary for quantum computers. No supercomputer has yet been able to calculate the behavior of the strongly interacting electrons in topological semimetals. However, the researchers developed a method that combines simplified model calculation, mathematical symmetry considerations, and a database of known materials to identify promising candidates that could have the theoretically expected topological properties.
|
Philadelphia Tech Salaries See Biggest Jump in the U.S.
The Philadelphia Inquirer Erin Arvedlund September 23, 2022
Technology job-search marketplace Hired.com's annual worker/employer poll found Philadelphia tech salaries grew fastest this year, rising nearly 12% to $142,000 compared with 15 other U.S. metro areas. Wages rose even higher for remote tech roles in Philadelphia, averaging $157,000 in 2022. Dallas/Fort Worth and Denver joined Philadelphia in posting the highest average "local" salary hikes across all U.S. markets, with +11% each. Engineers and other tech experts' salaries saw exponential growth during the pandemic, compounded by inflation in goods and services.
|
Lossless Data Management Platform for ML, Sharing of Experimental Information
Waseda University September 21, 2022
A lossless data-management platform that describes relationships between properties, structures, and experimental processes in electronic laboratory notebooks has been developed by researchers at Japan's Waseda University and the National Institute for Materials Science. The notebook represents experimental events and related environmental parameters as knowledge graphs. Researchers incorporated raw data from more than 500 experiments on superionic conductivity in organic lithium-ion electrolytes into the notebook; an artificial intelligence algorithm rendered the knowledge-graph data as machine-learnable datasets and uploaded them into a public archive. "By sharing raw experimental data among researchers across the globe, novel functional materials could be discovered more quickly," said Waseda's Kan Hatakeyama-Sato.
|
More than Microscopes Can Show
Max Planck Gesellschaft September 22, 2022
Researchers from Germany's Max Planck Institute for Molecular Biomedicine (MPI) modeled how the stem cell protein Oct4 opens pieces of DNA wrapped around nuclear storage proteins known as histones. The researchers visualized Oct4's binding to nucleosomes via a computational nanoscope. The protein's DNA-opening mechanism "fills a knowledge gap in the understanding of how factors like Oct4 induce cell fate transitions,” said former MPI researcher Vlad Cojocaru. “Understanding these mechanisms will eventually provide means to optimize and control such transitions for successful use in therapies. And computer simulations will be at the heart of these future discoveries."
|
AI Message Decoder Based on Bacterial Growth Patterns
Duke University Ken Kingery September 23, 2022
Biomedical engineers at the Duke University Pratt School of Engineering have deciphered bacterial growth patterns with an artificial intelligence (AI) decoder. The researchers produced a virtual bacterial colony, and controlled growth and the numbers and sizes of bacterial dots, to create an alphabet based on the colonies' appearance in a Petri dish. This emorfi encoding scheme encrypts messages by assembling a movie of a series of patterns, each correlating to a specific letter that a machine-learning algorithm can identify. Interlopers should be unable to crack the code without their own AI, provided the receiver knows the set of initial conditions that led to the code's creation.
|
|