Welcome to the June 29, 2022, edition of ACM TechNews, providing timely information for IT professionals three times a week.
Please note: In observance of the U.S. Independence Day holiday, TechNews will not be published on Friday, July 1, or Monday, July 4. Publication will resume Wednesday, July 6. 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."
|
|
|
Period-Tracker Apps Aim for Anonymity Following Roe v. Wade Decision
The Wall Street Journal
Cordilia James; Shara Tibken
June 26, 2022
Period-tracking and fertility apps like Flo, Clue, Natural Cycles, and Apple's Health are used by women to help them become pregnant, avoid pregnancy, and track their periods. In the wake of the U.S. Supreme Court ruling overturning Roe v. Wade, the developers of some of those apps are looking to prevent the collection of users' identifying information. Said Natural Cycles' Raoul Scherwitzl, "The goal is to make it so no one—not even us at Natural Cycles—can identify the user." Flo is expected to roll out "Anonymous Mode," to enable its users to remove personal information from their accounts. However, stressed University of Houston's Leah Fowler, "Deleting your app from your phone does not always mean you've deleted your data anywhere other than your device."
*May Require Paid Registration
|
Measuring AI's Carbon Footprint
IEEE Spectrum
Matthew Hutson
June 26, 2022
Researchers at the Allen Institute for AI have developed a method of calculating the carbon emissions of artificial intelligence (AI) systems that boasts greater accuracy. The new approach records the energy usage of server chips as a series of measurements and aligns that data with a series of data points detailing local emissions per kilowatt-hour (kWh) of the energy consumed. The researchers found graphics processing units training the biggest model to only 13% completion emitted nearly as much carbon as powering a U.S. home for a year does. In contrast, training the smallest models generated as much carbon as charging a phone.
|
Schools Spend Billions on High-Tech Defenses for Mass Shootings
The New York Times
Natasha Singer
June 26, 2022
U.S. schools are investing heavily in technology-based defenses against mass shootings despite a lack of evidence of their effectiveness. Market research company Omdia estimates U.S. schools spent roughly $3.1 billion on such products last year, versus $2.7 million in 2017. A 2016 report from Johns Hopkins University researchers found little hard evidence of the effectiveness of such technologies, which include student-monitoring cameras and facial recognition software, gun-detection scanners, and artificial intelligence-enabled threat detectors. Civil liberties experts say surveillance technologies not only make students feel less safe, but they also make no headway against widespread availability of assault weapons and a national mental health crisis.
*May Require Paid Registration
|
Researchers Develop Algorithm to Study Connectivity in Brain
The Hindu (India)
June 27, 2022
Researchers at India's Indian Institute of Science (IISc.) have developed the Regularized, Accelerated, Linear Fascicle Evaluation (ReAl–LiFE) algorithm to gain more insights into connectivity between different areas of the human brain. The graphic processing unit (GPU)-based algorithm can help analyze connectivity data produced by diffusion magnetic resonance imaging (dMRI) scans 100 to 150 times faster than conventional desktops or state-of-the-art algorithms. An IISc. press release said the researchers modified the LiFE optimization algorithm to reduce computational load by eliminating redundant connections, and also reworked it to operate on GPUs. "Tasks that previously took hours to days can be completed within seconds to minutes," said IISc.'s Devarajan Sridharan.
|
Supercomputers Take Over NOAA's Weather Forecast Models
Nextgov
Frank Konkel
June 29, 2022
The U.S. National Oceanic and Atmospheric Administration (NOAA)'s National Weather Service on Tuesday began using twin supercomputers developed by General Dynamics Information Technology (GDIT) to run its weather forecast models. The Dogwood and Cactus systems operate at a speed of 12.1 petaflops, and provide roughly triple the computing capacity for weather monitoring of the supercomputers previously constituting NOAA's Weather and Climate Operational Supercomputing System program. GDIT officials said the supercomputers will absorb up to 35 terabytes daily from satellites and other instruments. As a result, said National Weather Service director Ken Graham, "We can deploy higher-resolution models to better capture small-scale features like severe thunderstorms, more realistic model physics to better capture the formation of clouds, precipitation, and a larger number of individual model simulations to better quantify model certainty."
|
Robots Play with Play Dough
MIT News
Rachel Gordon
June 23, 2022
Researchers at the Massachusetts Institute of Technology (MIT), Stanford University, and the University of California, San Diego have developed a system that allows robots to manipulate play dough, a rubbery mixture of water, salt, and flour. The RoboCraft system, which learns directly from visual input, enabled a robot with a two-fingered gripper to learn to shape play dough into letters after receiving just 10 minutes of data. RoboCraft relies on cameras to collect visual sensor data from the environment that is transformed into graphs of tiny particles, which are used by a graph-based neural network to learn how the object moves. The robot's behavior is then planned using algorithms that teach it how to shape the object based on training data from numerous pinches of the material.
|
Open Source Language AI Challenges Big Tech Models
Nature
Elizabeth Gibney
June 22, 2022
A natural language processing model developed by an international team of about 1,000 researchers will rival in scale those developed by Google and OpenAI, but will be open source and multilingual. The BigScience collaboration's BLOOM language model was trained with $7 million-worth of publicly funded computing time, and granted free access to France's national Jean Zay supercomputer facility. The project's goal is to minimize the human biases and other flaws of artificial intelligence (AI) systems that are difficult to address because big technology companies have largely closed their models to researchers. While the model will not be bias-free, HuggingFace's Yacine Jernite said the team aimed to improve on existing models through the use of multicultural, high-quality sources.
|
Extra 'Eye' Movements Key to Better Self-Driving Cars
RIKEN (Japan)
June 27, 2022
Researchers at Japan's RIKEN Center for Brain Science have created artificial neural networks that learn to identify objects faster and more accurately by focusing on unnoticed eye movements. RIKEN's Andrea Benucci designed a convolutional neural network (CNN) to optimize the classification of objects in visual scenes while one’s eyes are moving, first by training it to classify 60,000 black-and-white images into 10 categories. Training the CNN with shifted images or motor copies enabled the model to better accommodate visual noise, provided the direction and size of the eye movements causing the shift were also included. Said Benucci, "This advancement will help avoid dangerous mistakes in machine vision," such as those caused by pixel alterations or adversarial attacks, which could affect the safety of self-driving vehicles.
|
Tech Behind Crypto Could Save Luxury Brands Billions
CNN
Rebecca Cairns
June 27, 2022
Luxury brands are using blockchain technology to shield their products from counterfeiting. Luxury brand conglomerate LVMH partnered with fashion company Prada and jewelry conglomerate Cartier on the Aura Blockchain Consortium, a nonprofit platform that generates digital twins for designer products. The consortium employs blockchain to produce digital product identifiers for more than 20 brands, including more than 17 million registered products, according to general secretary Daniela Ott. The software, which amasses a ledger of data like material type and source, where and when products were made, and how many were fabricated, provides customers a digital certificate of authentication that cannot be counterfeited, Ott said.
|
Supernumerary Virtual Robotic Arms Can Feel Like Part of Our Body
University of Tokyo (Japan)
June 27, 2022
Scientists at Japan's University of Tokyo, Keio University, and Toyohashi University have developed supernumerary virtual robotic arms that can be operated by users’ feet. Users wore a head-mounted display providing a first-person view of their own arms in virtual reality, plus the supernumerary limbs; they then performed tasks using the virtual robotic arms only, controlled by toe movements. Study participants said after training that the virtual robotic limbs felt like part of their own bodies, while their peripersonal space (the area around the body perceived as one's personal space) felt like it had been extended to include the vicinity of the robotic arms.
|
NIST Update to Software Reference Library Will Aid Criminal Investigations
NIST
June 27, 2022
The U.S. National Institute of Standards and Technology (NIST) updated its National Software Reference Library (NSRL) to make it easier to search through data on electronic equipment seized in police raids during criminal investigations. In the database's first major update in 20 years, the number and type of records in the database were broadened to reflect the growing diversity of digital files that law enforcement might find on a device. NIST also has revised the records' format to make the NSRL more searchable. The SQLite format makes it easier for users to create tailored filters to screen files for desired content. Said NIST's Doug White, “The update should make it easier for police to separate the wheat from the chaff."
|
Scientists Develop Algorithms to Study Genomic Data
UCLA Samueli School of Engineering
June 28, 2022
A multi-institutional team of researchers led by the University of California, Los Angeles' Sriram Sankararaman has developed algorithms that can analyze genomic data up to 1,800 times faster than previous methods, enabling the potential analysis of information from 1 million people in a single day. The researchers designed the SCalable pOPulation structure inferencE (SCOPE) technique to accelerate and scale the inference of genetic variability within a population to reveal patterns that direct conclusions or avoid false associations in research. SCOPE reduces the volume of computational resources needed to process large datasets, and lowers the cost of running calculations. In one experiment, the researchers used just 250 gigabytes (GB) of memory for computation, rather than the approximately 2,000 GB required by a previous research tool.
|
The Lost Masterpieces Being Revealed
BBC News
Jo Lawson-Tancred
June 21, 2022
Advanced technologies are rediscovering iconic art works by revealing previously hidden masterpieces. George Cann and Anthony Bourached from the U.K.'s University College London (UCL) employed artificial intelligence (AI) to render x-ray images of paintings as three-dimensional prints that recreate concealed images’ colors and texture. London's National Gallery partnered with UCL and Imperial College London to produce clearer images of a portrait underneath Francisco de Goya's Doña Isabel de Porcel by having a computer combine multiple scans from different bands of the electromagnetic spectrum. Amsterdam's Rijksmuseum also has used AI to reconstruct lost panels from Rembrandt's The Night Watch from a copy by artist Gerrit Lundens, converting his style into Rembrandt's.
|
|
