Welcome to the July 5, 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."
|
|
|
AI, CRISPR Precisely Control Gene Expression
New York University
July 3, 2023
Researchers at New York University, Columbia University, and the New York Genome Center manipulated human gene expression with a deep learning model combining artificial intelligence and CRISPR (clustered regularly interspaced short palindromic repeats) screens. The Targeted Inhibition of Gene Expression via guide RNA design (TIGER) model forecasts the activity of RNA-targeting CRISPRs using the Cas13 enzyme to maximize CRISPR activity on the intended target RNA, while minimizing activity on other RNAs that could negatively affect the cell. The researchers used TIGER to quantify the activity of 200,000 guide RNAs targeting essential genes in human cells. They showed it could anticipate on-target and off-target activity and demonstrated the model's off-target predictions can be used to modulate gene dosage.
|
Ancient Rock Engravings at Timna Park Reveal Insights into Human Cultures
The Jerusalem Post (Israel)
Judy Siegel-Itzkovich
July 2, 2023
Scientists at Hebrew University (HU) in Israel used a new computer architecture to gain insights into ancient rock-engraving techniques by analyzing carvings in Israel's Timna Park. The researchers used the ArchCUT3-D framework to evaluate three-dimensional (3D) scans of two engravings and modern graffiti to identify the 3D micromorphological properties of the engravings' constituent incisions. The outcomes show distinct methods underlying the incisions' creation, and the researchers said the ArchCUT3-D software enables the reassembly of engraving gestures and production processes. Said HU's Leore Grosman, "By unlocking the technological secrets behind these engravings, we gain valuable insights into the craftsmanship, artistic expression, and cultural context of our ancestors."
|
Mizzou Engineering Team Develops Video Retrieval System Based on Captioning
Mizzou Engineering
June 29, 2023
A team of researchers from the University of Missouri College of Engineering (Mizzou Engineering) created an image captioning-based system for retrieving video clips. Mizzou Engineering's Praveen Rao explained the system "will find you the top most relevant video clips in the database" in response to a query video. The system splits a given video into representative frame segments, then retrieves clips that are ranked according to the user's text query. Rao said the tool applies principles incorporated into the previous QIK image retrieval system, which generates captions for photos via artificial intelligence, for video retrieval.
|
Tour de France Adds ChatGPT, Digital Twin Tech
ZDNet
Sabrina Ortiz
June 29, 2023
Japan-based information technology and services company NTT said this year's Tour de France will incorporate ChatGPT and "the world's largest connected stadium." The connected stadium establishes a digital twin of the race using real-time data, to help the Amaury Sports Organization better understand goings-on in support of seamless operation. NTT will use geolocation and sensors on each bike to capture a continuous flow of data broadcast over radio networks to race vehicles before a microwave signal transmits it to the race's destination, where an edge-computing device will operate a "containerized version" of a real-time analytics platform. ChatGPT also will provide information to fans via an integration within NTT's artificial intelligence-driven Digital Human solution called Marianne.
|
Device Remotely Monitors Breathing, Tested on Cane Toads
The University of Sydney (Australia)
June 30, 2023
Scientists at Australia's University of Sydney (USYD) and the nonprofit New South Wales Smart Sensing Network developed a photonic radar system to monitor breathing both precisely and remotely. USYD's Ziqian Zhang explained the system combines photonic radar with light detection and ranging (LiDAR) to deliver resolution as fine as six millimeters with micrometer-scale accuracy. Former USYD researcher Yang Liu said the inbuilt redundancy of concurrent radar and LiDAR means "if either system encounters a fault, the other continues to function." The researchers said tests on cane toads and devices that simulate human respiration validated the system's use for monitoring multiple patients' vital signs from a centralized station.
|
Robot Takes Podium as Orchestra Conductor in Seoul
Reuters
Daewoung Kim; Jimin Jung
July 3, 2023
An android robot engineered by South Korea's Korea Institute of Industrial Technology recently conducted that country’s national orchestra at Seoul's National Theater of Korea. The humanoid EveR 6 robot directed three of five highlighted pieces, including one co-led with principal conductor Choi Soo-yeou. Choi said EveR 6 exceeded his expectations for performing a conductor's detailed movements, although its inability to listen is a "critical weakness." Audience member Lee Young-ju said although the robot could keep impeccable rhythm, it could not keep the orchestra primed for collective and instant engagement. Said Choi, "It was a recital that showed that [robots and humans] can co-exist and complement each other, rather than one replacing the other."
|
Computer Vision System Weds Image Recognition, Generation
MIT News
Rachel Gordon
June 28, 2023
Scientists at the Massachusetts Institute of Technology (MIT) and Google designed the Masked Generative Encoder (MAGE) computer vision system to unify image recognition and generation. MAGE renders images as compact 16 x 16 abstracts of image sections called "semantic tokens" in a process that self-supervised frameworks can use to pre-train on unlabeled image datasets. The system's "masked token modeling" technique involves randomly concealing certain tokens, then training a neural network to fill in the blanks. MIT's Tianhong Li explained, "MAGE's ability to work in the 'token space' rather than 'pixel space' results in clear, detailed, and high-quality image generation, as well as semantically rich image representations."
|
How Amazon Taught Alexa to Speak in an Irish Brogue
The New York Times
Bernhard Warner
July 1, 2023
Amazon data scientists taught the Alexa digital assistant to speak Irish-accented English using artificial intelligence (AI) and recordings from native speakers. The project is part of an effort to address voice disentanglement, which goes beyond understanding vocabulary and syntax to include what linguists call "prosody," the nuanced meaning and emotional weight given to words by the speaker's pitch, timbre, and accent. The Amazon team trained Alexa to speak Irish English using an existing speech model consisting of mainly British-English accents, but also including American, Canadian, and Australian accents. Irish Alexa built up its speech model with a greater reliance on AI than on voice actors, using only around 24 hours of recordings featuring 2,000 utterances in Irish-accented English.
*May Require Paid Registration
|
Artificially Cultured Brains Improve Processing of Time Series Data
Tohoku University (Japan)
June 29, 2023
Researchers at Japan's Tohoku University (TU) and Future University Hakodate gauged the computational powers of an "artificially cultured brain" composed of rat cortical neurons using reservoir computing. TU's Hideaki Yamamoto explained the researchers "first recorded the multicellular responses of the cultured neuronal network" via optogenetics and fluorescent calcium imaging. Reservoir computing-enabled decoding revealed that the brain "possessed a short-term memory of several hundred milliseconds, which could be used to classify time-series data, such as spoken digits," according to Yamamoto. The researchers found improved classification performance in samples with a higher level of modularity, while a model trained on one dataset could classify another dataset in the same category.
|
Care Homes in Japan Use Big Data to Boost Caregivers, Lighten Workloads
Financial Times
Eri Sugiura
July 2, 2023
Japan-based nursing home operator and insurer Sompo Holdings is using technology developed in collaboration with U.S. data analysis company Palantir to alleviate a caregiver shortage in that nation. The partners' software platform integrates artificial intelligence and analytics with proprietary data on sleep, diet, medical treatment, and exercise. Sompo home care manager Takako Kojima said using such data analysis methods helps to lighten caregivers' loads and to personalize care provision. Last year, Kojima's facility equipped its beds to evaluate sleep conditions with sensors that quantify body movements, respiration, and heart rate so caregivers can anticipate when residents might need help at night. Sompo said the technology can cut 15% of the workload in a typical 60-person capacity care facility, saving as much as $60,000 annually.
*May Require Paid Registration
|
'Ghost Particle' Image of Milky Way Galaxy Captured
National Science Foundation
Jason Stoughton
June 29, 2023
For the first time, scientists have captured an image of the Milky Way made with "ghost particles," particles of matter also known as neutrinos. Researchers at Drexel University and Germany's TU Dortmund University generated the image using the IceCube Neutrino Observatory at the National Science Foundation's Amundsen–Scott South Pole Station in Antarctica, which uses thousands of networked sensors under a cubic kilometer of clear ice to detect high-energy neutrinos from space. A machine learning algorithm developed by the researchers compared the relative position, size, and energy of over 60,000 cascades of light generated by neutrinos and recorded by the observatory over the course of a decade. The algorithm generated an image indicating bright spots that corresponded to parts of the Milky Way believed to emit neutrinos.
|
Foldseek Gives AlphaFold Protein Database a Rapid Search Tool
Nature
Matthew Hutson
June 29, 2023
Researchers at South Korea's Seoul National University have developed an open source software system that allows scientists to search databases like Google DeepMind's AlphaFold 2 for proteins with similar shapes and functions. The Foldseek system uses a "structural alphabet" to represent a protein's shape, which enables researchers to take advantage of the sensitivity of shape-based searches at the speed of sequence-based searches. Seoul National University's Martin Steinegger said Foldseek's three-dimensional interaction alphabet is better able to represent a protein's global structure by assigning each amino acid one of 20 letters based on its distance from and orientation relative to the amino acid closest in the folded-up protein. Foldseek outperformed the structure-based search tools TM-align and Dali, and the structural-alphabet-based tool CLE-SW, in tests that asked it to rank the most similar proteins for 100 protein shapes that had been predicted using AlphaFold 2.
|
Counting Africa's Largest Bat Colony
Max Planck Gesellschaft (Germany)
June 30, 2023
A team of researchers in Germany and the U.S. integrated GoPro cameras with artificial intelligence to produce the most accurate count of the largest bat colony in Africa to date. The tool monitored the massive November migration of straw-colored fruit bats to a small area of Zambia's Kasanka National Park over five nights, calculating an overall population of 750,000 to 1 million bats per night. The researchers deployed nine GoPro cameras throughout the colony to record bats as they took flight at dusk. Ben Koger at Germany's Max Planck Institute of Animal Behavior trained deep learning models to detect and count bats in the GoPro videos. Tests showed the model's bat tally in video clips matched humans' count with 95% accuracy, even under dark conditions.
|
|
