Welcome to the September 15, 2023, 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."
|
|
DNA-Based Computer Can Run 100 Billion Different Programs
New Scientist Karmela Padavic-Callaghan September 13, 2023
Researchers at China's Shanghai Jiao Tong University (SJTU) and Chinese Academy of Sciences developed a liquid computer that can run more than 100 billion different programs using DNA strands. The researchers combined short DNA molecules into circuits that function as wires and instructed them to form different configurations. Adding different short molecules into tubes filled with DNA strands enable each DNA-based programmable gate array (DPGA) to implement different circuits. One experiment involved designing a DPGA for classifying and selecting small RNA molecules related to a type of renal cancer as a forerunner of a natural method for "intelligent diagnostics of different kinds of diseases," said SJTU's Fei Wang.
|
Machine Learning Tames Huge Datasets
Los Alamos National Laboratory September 11, 2023
A machine learning algorithm developed at the U.S. Department of Energy (DOE)'s Los Alamos National Laboratory (LANL) was able to identify and split a vast dataset's key features into manageable batches. Researchers tested the algorithm on the Summit supercomputer at DOE's Oak Ridge National Laboratory. LANL's Ismael Boureima said, "We developed an 'out-of-memory' implementation of the non-negative matrix factorization method that allows you to factorize larger datasets than previously possible on a given hardware." The algorithm efficiently transfers data between computers by speeding up computation and fast interconnect using hardware like graphics processing units (GPUs), while performing multiple tasks simultaneously. The LANL researchers used the algorithm to process a 340-terabyte dense matrix and an 11-exabyte sparse matrix with 25,000 GPUs.
|
Kotlin Rises to Tiobe Top 20
InfoWorld Paul Krill September 12, 2023
JetBrains' Java language alternative Kotlin has entered Tiobe's monthly index of programming language popularity, ranking 20th with a 0.90% rating in September. The rankings are calculated using a formula that factors in the number of engineers, courses, and third-party vendors using each language based on popular search engine data. Tiobe’s September top 10 are, in order Python, C, C++, Java, C#, JavaScript, Visual Basic, PHP, Assembly language, and SQL. Meanwhile, the top 10 in the September edition of the Pypl Popularity of Programming Language index, which measures Google searches of language tutorials, are Python, Java, JavaScript, C#, C/C++, PHP, R, TypeScript, Swift, and Objective-C.
|
Canceling Noise to Improve Quantum Devices
MIT News Peter Reuell September 6, 2023
A team of physicists at the Massachusetts Institute of Technology borrowed a concept from noise-canceling headphones' ability to filter out surrounding noise using specific frequencies to extend the coherence times of quantum bits (qubits) 20-fold. The researchers' "unbalanced echo" approach defines how heat impacts quadrupole interactions in the quantum system to counter nuclear-electron interactions and lengthen coherence times from 150 microseconds to as long as 3 milliseconds. The experiment involved ensembles of approximately 10 billion nitrogen vacancy (NV) centers in diamond, each existing in a specific quantum spin state for the nitrogen-14 nucleus and with a localized electron nearby. The new system removes the need to "flip" or reverse the NV centers' spin to extend coherence time, preventing data loss.
|
Brain Surgery Visualization using NavTech from Self-Driving Cars
The Johns Hopkins News-Letter Jason Han September 14, 2023
Scientists at Johns Hopkins University (JHU) and medical device company Medtronic created a real-time endoscopic neurosurgery guidance method similar to the navigational technologies used in autonomous cars. The researchers conceived of an endoscopic camera to help map the X, Y, and Z coordinates of target structures within the human brain. JHU's Prasad Vagdargi assessed the navigation system's ability to visualize locations inside the brain using rubbery material that he designed and modeled after the human brain. The team used that material to compare two computer vision algorithms for analyzing three-dimensional (3D) reconstructions, and found the Structure from Motion and Simultaneous Localization and Mapping (SLAM) computer vision algorithm to be significantly faster than the Structure from Motion (SfM) algorithm.
|
Drones Tackle Shark Safety on New York's Beaches
Forbes Melissa Cristina Marquez September 11, 2023
Safety officials at New York's Jones Beach are using drones to monitor for potential human-shark interactions. The New York State Park Police maintains a fleet of 19 lifeguard-operated drones stationed along the beach, and officers can provide augmented monitoring via a command center van. Officers in the van can review livestreamed drone footage to see if a shark is in the vicinity and determine whether swimmers should be evacuated. The drones' cameras can penetrate the water's surface from high overhead and have previously captured footage of sharks swimming alone and feeding on large schools of fish. Park Police Captain Rishi Basdeo said, "The more drones that are flying in the air, there's more of a chance of seeing these animals in their natural habitat."
|
Tool Skewers Socially Engineered Attack Ads
Georgia Tech Research September 8, 2023
Trident, developed by researchers at Georgia Institute of Technology (Georgia Tech), is, a Google Chrome-compatible add-on that can block socially engineered online ads with what the researchers describe as nearly total efficacy. Georgia Tech's Zheng Yang said, "The goal is to identify suspicious ads that often take users to malicious websites or trigger unwanted software downloads. Trident operates within Chrome's developer tools and uses a sophisticated AI [artificial intelligence] to assess potential threats." The researchers built Trident using a dataset amassed from over 100,000 websites, which helped identify 1,479 attacks covering six common types of Web-based social engineering exploits. Trident realized a near-perfect detection rate of malicious ads over the course of a year, yielding a mere 2.57% false-positive rate.
|
Japan's SLIM Moon Lander Is Carrying a Transforming Ball Robot
Space.com Josh Dinner September 11, 2023
On Sept. 6, the Japan Aerospace Exploration Agency (JAXA) launched the Smart Lander for Investigation Moon (SLIM) probe with a small, spherical lunar explorer aboard. The Lunar Excursion Vehicle 2 (LEV-2), about the size of a tennis ball and with around 2 hours of battery power, will detach from the spacecraft once it reaches the moon. The two halves of the roboball will separate, serving as legs and wheels so it can image SLIM's landing and the surrounding area. LEV-2 features two cameras and a stabilizer, and it will use the LEV-1 probe, also attached to SLIM, to transmit data back to Earth. JAXA's Hirano Daichi said, "We adopted the robust and safe design technology for children's toys, which reduced the number of components used in the vehicle as much as possible and increased its reliability" while also minimizing its size.
|
Revealing Secrets of Protein Evolution Using the AlphaFold Database
EMBL News Vicky Hatch September 13, 2023
Scientists at Germany's European Molecular Biology Laboratory (EMBL), Switzerland's ETH Zurich, and South Korea's Seoul National University have identified unique evolutionary patterns by analyzing more than 200 million predicted protein structures with the AlphaFold database. The researchers created and applied the Foldseek Cluster algorithm to the predicted protein structures, which identified over 2 million unique groups of structures whose three-dimensional configurations resemble each other. One-third of the clusters had never been previously described or categorized. EMBL's Sameer Velankar said Foldseek Cluster has "revolutionized" the search for predicted protein structures in the AlphaFold database.
|
'Computer Vision' Reveals Physical, Chemical Details of How Lithium-Ion Battery Works
SLAC National Accelerator Laboratory Glennda Chui September 13, 2023
Researchers at the U.S. Department of Energy's SLAC National Accelerator Laboratory, Stanford University, the Massachusetts Institute of Technology, and the Toyota Research Institute used computer vision to expose the physical and chemical workings of lithium-ion batteries in unprecedented detail. The method enables the observation of lithium iron phosphate (LFP) particles in the electrodes absorbing and releasing lithium ions, as recorded via nanoscale X-ray. The researchers conducted a pixel-by-pixel analysis of 62 X-ray movies of LFP particles charging or discharging, training the computer model on about 180,000 pixels to generate equations describing lithium insertion reactions; these processes closely aligned with reactions forecast by earlier simulations. The discovery that variations in the thickness of an LFP particle's carbon coating affect the rate of lithium-ion inflows and outflows could clear a path toward more efficient charging and discharging.
|
Microflier Robots Use the Science of Origami to Fall Like Leaves
Popular Science Andrew Paul September 13, 2023
Origami-inspired robots designed by University of Washington (UW) researchers can dynamically change their shape while descending through the air. UW's Vikram Iyer said the "microfliers" initially fall "chaotically" from drones in a flat, unfolded position similar to elm leaves, then fold in midair so the airflow around them can enable a more stable descent akin to that of maple leaves. Researchers can program solar-powered actuators to activate at specific times to control how and when the microfliers' shapes interact with the air. Iyer explained, "This highly energy efficient method allows us to have battery-free control over microflier descent, which was not possible before."
|
Using Topology, Researchers Advance Understanding of Cell Organization
News from Brown September 14, 2023
Biomedical engineers and applied mathematicians at Brown and Purdue universities designed a machine learning algorithm that uses computational topology to characterize how cells self-organize into tissue-like architectures. In 2021, the researchers demonstrated how their technique can profile the topological characteristics of one cell type that assembles into different spatial structures, and base predictions on that analysis. The latest research applied persistence images to address the algorithm's hours-long topological computation. The researchers trained other algorithms on those images to produce "digital fingerprints" that record the data's key topological traits, accelerating computation time from hours to seconds and enabling scientists to compare thousands of cell-organization models. They say the goal is to infer the rules governing how different cell types self-assemble into final patterns by working backward.
|
Why Japan Is Building Its Own Version of ChatGPT
Nature Tim Hornyak September 14, 2023
The Japanese government, big Japanese technology firms, and researchers in Japan are working to build versions of ChatGPT with underlying large language models (LLMs) that use the Japanese language. LLMs trained on datasets in other languages do not take account for differences in alphabet systems, sentence structure, and culture. The Tokyo Institute of Technology, Tohoku University, Fujitsu, and the government-funded RIKEN group of research centers are collaborating on a Japanese LLM using the Fugaku supercomputer. The LLM, slated for release next year, could have at least 30 billion parameters. Meanwhile, an LLM funded by Japan's Ministry of Education, Culture, Sports, Science and Technology could start with 100 billion parameters and expand over time. Keio University School of Medicine's Shotaro Kinoshita said the development of an accurate, Japanese version of ChatGPT could have "a positive impact on international joint research."
|
|