Welcome to the April 26, 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."
|
|
AI's Carbon Footprint Is Big, But Easy to Reduce, Google Researchers Say
Fortune Jeremy Kahn April 21, 2021
Researchers at the University of California, Berkeley and Google have released the most accurate estimates to date for the carbon footprint of large artificial intelligence (AI) systems. They determined OpenAI's powerful language model GPT-3, for example, produced the equivalent of 552 metric tons of carbon dioxide during its training. The researchers found the carbon footprint of training AI algorithms depends on their design, the computer hardware used to train them, and the nature of electric power generation in the location where the training occurs; changing all three factors could lower that carbon footprint by a factor of up to 1,000. A reduction by a factor of 10 could be achieved through the use of "sparse" neural network algorithms, in which most of the artificial neurons are connected to relatively few other neurons.
|
Chemical Code Used to Store Jane Austen Quote in Plastic Molecules
New Scientist Christa Lesté-Lasserre April 21, 2021
University of Texas at Austin (UT Austin) researchers encoded a quote from Jane Austen's novel Mansfield Park in software-readable plastic molecules. UT Austin's Eric Anslyn was trying to generate complex molecules to boost the efficacy of pharmaceuticals and dishwasher detergents, and realized those compounds could be used to represent symbolic values for data storage. Molecules assembled from these could become their own code language based on a hexidecimal (16-character) code, while the liquid chromatography-mass spectroscopy analytical system Anslyn was using could easily analyze and sequence such substances. The Austen quote was encoded within a hexadecimal molecular language, using software developed by Anslyn's team. Said Anslyn, "We always write in symbols, and molecules are just another set of symbols that we can assemble—not just for building molecules analogous to those found in nature, but to create our own inventions."
|
Researchers Say Changing Simple iPhone Setting Fixes Long-Standing Privacy Bug
USA Today Mike Snider April 24, 2021
Scammers could exploit a bug in iPhones and MacBooks' AirDrop feature to access owners' email and phone numbers, according to researchers at Germany's Technical University of Darmstadt (TU Darmstadt). AirDrop allows users with both Bluetooth and Wi-Fi activated to discover nearby Apple devices, and share documents and other files; however, strangers in range of such devices can extract emails and phone numbers when users open AirDrop, because the function checks such data against the other user's address book during the authentication process. The researchers said they alerted Apple to the vulnerability nearly two years ago, but the company "has neither acknowledged the problem nor indicated that they are working on a solution." They recommend users disable AirDrop and not open the sharing menu, and to only activate the function when file sharing is needed, then deactivate it when done.
|
Newer Planes Providing Airlines a Trove of Useful Data
The New York Times Christine Negroni April 20, 2021
The retirement of older aircraft during the pandemic has resulted in a fleet equipped with digital technologies that can collect more information about emissions, safety, and other factors. Kevin Michaels at aerospace consultancy AeroDynamic Advisory notes that the latest Airbus airliner, the A350, usually records 800 megabytes of data per flight, double the amount recorded by the Airbus A380. As the numbers of modern aircraft in airline fleets grow, so will the amount of data available. New broadcast tracking signals are flight-specific, but can provide information useful for navigation services and arrival planning to help manage the stream of traffic in the air and at airports.
*May Require Paid Registration
|
From Individual Receptors Towards Whole-Brain Function
Ruhr-Universität Bochum (Germany) Julia Weiler April 23, 2021
Teams of researchers at Germany's Ruhr-Universität Bochum, Spain's Pompeu Fabra University, and the U.K.'s Oxford University developed concepts to measure receptor-specific modulations of brain states, and a computer model for predicting the impact of individual receptor types on brain activity. The researchers simulated the impact of individual receptor types on whole-brain dynamics by compiling data using three imaging methods: diffusion-weighted magnetic resonance imaging (MRI) to record information on the brain's anatomical connectivity; functional MRI to obtain information about resting-state activity of participants; and positron emission tomography-recorded distributions of receptor type. From these, the researchers were able to construct an individualized receptome for each subject, reflecting the overall distribution of receptor types in their brain. The receptome model enabled the simulation of interactions between neurons dependent on activations of individual receptor types, which the researchers hope to apply to diagnosing and treating mental disorders.
|
Algorithm Uses Online Ads to Identify Human Traffickers
Carnegie Mellon University School of Computer Science Aaron Aupperle April 22, 2021
Researchers at Carnegie Mellon University (CMU) and Canada's McGill University hope to identify human trafficking by adapting an algorithm, originally used to spot data anomalies, to detect similarities across escort ads. CMU's Christos Faloutsos said the InfoShield algorithm scans public datasets and clusters textual similarities, and could help law enforcement direct probes and better identify human traffickers and their victims. Said Faloutsos, "Our algorithm can put the millions of advertisements together and highlight the common parts. If they have a lot of things in common, it's not guaranteed, but it's highly likely that it is something suspicious." When tested on a set of escort listings in which experts had already identified trafficking, the algorithm flagged them with 85% precision, while producing no false positives.
|
Which City Rates as Best Place to Work in Tech?
IEEE Spectrum Tekla S. Perry April 22, 2021
Professional social network Blind surveyed 1,085 users at technology firms in Austin, Chicago, New York City (NYC), the San Francisco Bay Area, and Seattle, in order to determine which cities they think treat tech workers best. Although most respondents in each city rated that location high in a majority of categories, differences did emerge. Austin was rated the top city where local government helps tech workers and companies prosper. NYC ranked highest as a place where fellow techies can be found to socialize with, while respondents indicated they saw the San Francisco Bay Area as having the most tech career opportunities.
|
Gold Digger: Neural Networks at the Nexus of Data Science, Electron Microscopy
Max Planck Florida Institute for Neuroscience April 20, 2021
New software developed by researchers at the Max Planck Florida Institute for Neuroscience (MPFI) uses a deep learning algorithm to identify gold particles bound to specific proteins in electron microscopy (EM). The open source Gold Digger software automates the process of analyzing protein distribution in EM images. The adaptable, deep learning-based algorithm can identify different sizes of gold particles, which will speed the counting process and generate more precise location information for protein distributions across a membrane. MPFI's Michael Smirnov said, "We found that by feeding enough training data and correcting errors that pop up in our algorithms, our software could distinguish gold particles from these shadow artifacts with near-human-level accuracy."
|
Multiple Agencies Breached by Hackers Using Pulse Secure Vulnerabilities
The Hill Maggie Miller April 20, 2021
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) said hackers had infiltrated federal agencies and other critical organizations by exploiting flaws in products from Utah-based software company Ivanti Pulse Connect Secure (PCS). The CISA alert followed cybersecurity group FireEye's Mandiant Solutions' publication of a blog post attributing some breaches to a Chinese state-sponsored hacking group and another Chinese advanced persistent threat group. CISA said that hackers had installed webshells in PCS products, which enabled them to circumvent security features. The agency said Ivanti was developing a patch, adding that it "strongly encouraged" all users to update to the latest version of the software and to look for signs of breaches. CISA issued an emergency directive requiring all federal agencies evaluate how many PCS products they and third-party organizations used, and to update them by April 23.
|
ML Model Generates Realistic Seismic Waveforms
Los Alamos National Laboratory News April 22, 2021
The SeismoGen machine learning model can generate high-quality synthetic seismic waveforms, according to researchers at the U.S. Department of Energy's Los Alamos National Laboratory (LANL). The team designed SeismoGen based on a generative adversarial network. Once trained, the SeismoGen model can produce realistic seismic waveforms of multiple labels. The LANL researchers applied the model to actual Earth seismic datasets in Oklahoma. LANL's Youzuo Lin said, "Through a sequence of qualitative and quantitative tests and benchmarks, we saw that our model can generate high-quality synthetic waveforms and improve machine learning-based earthquake detection algorithms."
|
Robot Taught Table Etiquette Can Explain Why It Won't Follow the Rules
New Scientist Ibrahim Sawal April 21, 2021
Researchers at Italy's University of Palermo (UP) programmed a humanoid robot from Japanese manufacturer SoftBank Robotics with software that models human cognitive processes, along with a text-to-speech processor, so it could vocalize its decision-making process while completing tasks. The software enabled the robot, Pepper, to retrieve relevant data from its memory and determine the correct way to respond to human commands. After encoding etiquette rules into Pepper, the UP scientists asked it to set a dinner table, and either enabled or disabled its inner speech to observe the effects. When inner speech was disabled, the robot refused to perform tasks that contradicted the programmed rules, but could not explain its reasoning to the researchers. UP's Arianna Pipitone says hearing a robot voice its decision-making process boosts the transparency between humans and robots, which could have ramifications for cooperative tasks.
|
Stanford Researchers Use AI to Empower Environmental Regulators
Stanford News Rob Jordan April 19, 2021
Stanford University researchers have demonstrated how artificial intelligence combined with satellite imagery creates a low-cost, scalable method for finding and monitoring otherwise hard-to-oversee industries, which environmental regulators could employ to spot violators. Previous research has tended to focus on wealthy countries, while the Stanford-led work concentrated on Bangladesh, where localization and enforcement of environmental regulations related to highly pollutive brick kilns is difficult. In partnership with the International Center for Diarrheal Disease Research, Bangladesh, the team devised a deep learning algorithm that not only identifies whether satellite images contain such kilns, but also learns to locate kilns within the image. The algorithm can reconstruct kilns fragmented across multiple images, identify multiple kilns within a single image, and differentiate between sanctioned and illegal kiln technologies based on shape classification.
|
|