Welcome to the July 27, 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."
|
|
|
Could ML Fuel a Reproducibility Crisis in Science?
Nature
Elizabeth Gibney
July 26, 2022
Princeton University's Sayash Kapoor and Arvind Narayanan warn of a "brewing reproducibility crisis" driven by increasing reliance on machine learning (ML) to base predictions on patterns in data. The researchers say ML is overhyped as an easily learned tool, and peer reviewers lack the time to vet these models, leaving academia with no mechanisms to eliminate irreproducible papers. Kapoor and Narayanan analyzed 20 reviews in 17 research fields, citing 329 papers whose results could not be fully replicated due to flawed ML application. They highlighted data leakage as the most significant problem, and have developed guidelines for avoiding such traps. The guidelines instruct researchers to include with their manuscripts evidence that their models do not have each of eight types of leakage, as well as suggest a template for such documentation.
|
Researchers Create 3D Print Sensors for Satellites
MIT News
Adam Zewe
July 27, 2022
Scientists at the Massachusetts Institute of Technology (MIT) created the first completely 3D-printed retarding potential analyzers (RPAs)—plasma sensors used by satellites to determine the chemical composition and ion energy distribution of the atmosphere. The sensors performed as well as state-of-the-art semiconductor plasma sensors manufactured in a cleanroom, which makes them expensive and requires weeks of intricate fabrication. By contrast, MIT's 3D-printed sensors can be produced for tens of dollars in a matter of days, making them ideal for miniaturized CubeSat satellites. The researchers developed the sensors using a glass-ceramic material that is more durable than traditional sensor materials like silicon and thin-film coatings. Said MIT's Luis Fernando Velásquez-García, “Additive manufacturing can make a big difference in the future of space hardware. Some people think that when you 3D-print something, you have to concede less performance. But we've shown that is not always the case."
|
Using AI to Train Teams of Robots to Work Together
University of Illinois Grainger College of Engineering
Debra Levey Larson
July 19, 2022
University of Illinois Urbana-Champaign (UIUC) scientists have created a reinforcement learning framework to train multiple agents such as robots to collaborate. UIUC's Huy Tran said the researchers envisioned a decentralized scheme, while also focusing "on situations where it's not obvious what the different roles or jobs for the agents should be." He explained that they produced a utility function "that allows us to identify when an individual agent contributes to the global team objective." The algorithms also can detect when an agent or robot is performing a task that is not useful, and the researchers tested them on simulated games like Capture the Flag and StarCraft. Tran said such algorithms can be applied to real-world scenarios, like military surveillance or coordinated autonomous vehicle delivery.
|
Sensing Platform Deployed at Controlled Burn Site, Could Help Prevent Forest Fires
Argonne National Laboratory
Jared Sagoff
July 26, 2022
Scientists from the U.S. Department of Energy's Argonne National Laboratory deployed the Sage platform to conduct a controlled burn on a Kansas prairie, as part of research that could be applied to forest fire prevention. Sage integrates multiple sensors with edge computing and embedded machine learning (ML) algorithms, enabling on-the-spot detection, monitoring, and analysis of the burned area. Sage uses the open-source Waggle wireless sensor platform; Argonne's Pete Beckman said, "It's basically as if Waggle is a cellphone, and Sage is the network that the phone uses to communicate plus the apps that run on it." The controlled burn generated a dataset about the progression of smoke and fire, which can be used to teach an ML algorithm to ascertain the behavior of other fires in real time.
|
Dreaming Up Proteins, AI Churns Out Possible Medicines, Vaccines
Science
Robert F. Service
July 22, 2022
University of Washington (UW) researchers have unveiled artificial intelligence (AI) software that can design nonexistent proteins, including potential cancer treatments and a vaccine candidate for respiratory syncytial virus. The AI taps innovations in predicting three-dimensional protein structures from amino acid sequences, using the RoseTTAFold software to form proteins from scratch via inpainting and constrained hallucination. Inpainting gives the AI a starting point from which it can fill in additional protein components based on its understanding of protein-folding. Meanwhile, constrained hallucination establishes a goal, and the AI evolves a virtual protein to meet the goal. Both approaches yielded successful laboratory-cultured proteins.
|
Quantum Researchers Advance Error Handling
ComputerWeekly.com
Cliff Saran
July 26, 2022
A team led by researchers at the Flatiron Institute discovered a new phase of matter that has the potential to act as long-term quantum information storage. The researchers subjected a quantum computer’s qubits to “quasi-rhythmic laser pulses” based on the Fibonacci sequence, demonstrating a way of storing quantum information that is less prone to errors. Said Flatiron's Philipp Dumitrescu, "With this quasi-periodic sequence, there's a complicated evolution that cancels out all the errors that live on the edge. Because of that, the edge stays quantum-mechanically coherent much, much longer than you'd expect."
|
Imaging the Brain with Ultrasound Waves
ETH Zurich (Switzerland)
Simone Ulmer
July 26, 2022
Researchers at ETH Zurich in Switzerland are exploring ultrasound waves as a tool for imaging the brain. The researchers are working to produce high-resolution brain imagery by overcoming ultrasound waves' inability to penetrate hard tissues like bone. The team is developing algorithms that perform calculations over a "mesh" to model wave propagation through the brain using the Salvus software package; Salvus simulates full wave propagation across scales ranging from a few millimeters to thousands of kilometers. The researchers use magnetic resonance imaging (MRI) of the brain as a reference model, then execute calculations with different parameters on a supercomputer until the simulated image matches that of the MRI. This yields a quantitative image superior to grayscale images typical of conventional ultrasound, which can presumably distinguish healthy from diseased tissue.
|
Solar-Powered Army Drone Has Been Flying for 40 Days Straight
Popular Science
Bob Verger
July 25, 2022
The U.S. Army's solar-powered Zephyr drone has been flying for 40 days straight since its launch from an Arizona installation on June 15. The Airbus-manufactured drone is designed to fly in the stratosphere and run on a tiny amount of electricity, described in a 2021 Army release as the same as "a single commercial light bulb." Zephyr's current flying time exceeds its 14-day record in 2010 and its nearly 26-day record in 2018. The Army said the current mission has thus far "demonstrated Zephyr's energy storage capacity, battery longevity, solar panel efficiency, and station-keeping abilities that will further the Army's goal to implement ultra-long endurance stratospheric UAS [unmanned aerial systems] capabilities."
|
Researchers Use Wearable Tech to Detect COVID-19 Before Onset of Symptoms
McMaster University (Canada)
Jesse Dorey
July 19, 2022
Researchers in Canada and Europe detected COVID-19 prior to symptom onset by pairing wrist-worn health devices with machine learning. The researchers analyzed more than 1,100 participants wearing a fertility tracker that monitors respiration, heart rate, heart rate variability, skin temperature, and blood flow at night while sleeping. The tracker was synchronized to a mobile application that recorded activity that might affect the central nervous system, as well as potential COVID-19 symptoms. Over 100 participants tested positive for the virus, and the tracker detected changes in all physiological markers during infection. The team trained an algorithm to detect COVID-19 symptoms in 70% of positive-testing participants, and 73% of confirmed cases were detected as early as two days before symptom onset; tests on the remaining cohort yielded a 68% detection rate up to two days ahead.
|
Experts Uncover 'CosmicStrand' UEFI Firmware Rootkit Used by Chinese Hackers
The Hacker News
Ravie Lakshmanan
July 25, 2022
Researchers at the Kaspersky cybersecurity company have attributed a new Unified Extensible Firmware Interface (UEFI) firmware rootkit called CosmicStrand to unknown Chinese-speaking hackers. The researchers said CosmicStrand resides "in the firmware images of Gigabyte or ASUS motherboards, and we noticed that all these images are related to designs using the H81 chipset. This suggests that a common vulnerability may exist that allowed the attackers to inject their rootkit into the firmware's image." Attacks aim to interfere with the operating system loading process to implement a kernel-level implant into a Windows machine whenever it is booted, and to use this access to launch shellcode that connects to a remote server to retrieve the malware to be deployed on the system. The researchers noted CosmicStrand appears to have been used in the wild since the end of 2016, before UEFI rootkit exploits began to be publicly detailed.
|
Team Scripts Breakthrough Quantum Algorithm
City College of New York
July 25, 2022
A multi-institutional team of scientists headed by the City College of New York's Pouyan Ghaemi has developed a quantum algorithm that can simulate the evolving state of many interacting quantum particles over time. Ghaemi said, "Our quantum computing algorithm opens a new avenue to study the properties of materials resulting from strong electron-electron interactions. As a result it can potentially guide the search for useful materials such as high-temperature superconductors." Ghaemi also believes the research could enable researchers to consider using quantum computers to explore phenomena that result from the interaction of electrons in solids.
|
Robot that Learns Social Cues Could Feed People with Tetraplegia
New Scientist
Matthew Sparkes
July 26, 2022
Cornell University researchers have trained an artificial intelligence (AI) model to learn social cues to feed people with tetraplegia. The researchers fed the Social Nibbling Network (SoNNET) model videos of 30 groups of three people dining together in a social setting, teaching it to recognize behavior that people exhibited immediately before eating. SoNNET could be used to instruct a robotic arm when to feed a paralyzed person without interrupting the meal's social element. Experiments applied one of three approaches to initiate the robot to provide a mouthful—a set time interval, feeding when the diner opened their mouth and kept it open, and SoNNET. The AI model scored the highest rating among participants.
|
Coding Error Caused Outage That Left Millions Without Service
The Globe and Mail (Canada)
Alexandra Posadzki
July 25, 2022
Millions of Canadians lost their cellphone, Internet, or home phone service for at least a day due to a coding error on July 8, when Rogers Communications was upgrading its wireless/broadband infrastructure. The telecommunications company has one core network that supports all its services, and company documents indicated a piece of code deleted a routing filter during the sixth phase of the seven-phase infrastructure upgrade. The deletion caused all possible channels to the Internet to pass through the routers, resulting in several devices exceeding their memory and processing capacities, inducing a network shutdown. Rogers uses equipment from different manufacturers in its network, and its router suppliers have different traffic management and overload safeguards, which the documents identified as the source of the outage.
|
|
