Welcome to the October 18, 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."
|
|
|
Duke Computer Scientist Receives $1-Million AI Prize, a 'New Nobel'
Duke University Pratt School of Engineering
Ken Kingery
October 12, 2021
Duke University computer scientist Cynthia Rudin was awarded the $1-million Squirrel AI Award for Artificial Intelligence for the Benefit of Humanity by the Association for the Advancement of Artificial Intelligence (AAAI). Rudin was honored for "pioneering scientific work in the area of interpretable and transparent AI systems in real-world deployments, the advocacy for these features in highly sensitive areas such as social justice and medical diagnosis, and serving as a role model for researchers and practitioners." Said Duke’s Jun Yang, Rudin “is changing the landscape of how AI is used in societal applications by redirecting efforts away from black box models and toward interpretable models by showing that the conventional wisdom—that black boxes are typically more accurate—is very often false.”
|
Researchers Show Facebook's Ad Tools Can Target Single Users
TechCrunch
Natasha Lomas
October 15, 2021
Spanish and Austrian researchers have showed that Facebook can target ads to a single individual, given sufficient knowledge of that person’s assigned interests. The researchers described a data-driven model that characterizes a metric indicating the likelihood a Facebook user can be identified based on interests attached to them by the social media giant's ad platform. The model found that "the four rarest interests or 22 random interests from the interests set FB assigns to a user make them unique on FB with a 90% probability," the researchers wrote. These findings raise issues about potentially harmful uses of Facebook's ad-targeting tools, and about the legality of the platform's personal data processing system.
|
These Neural Networks Know What They're Doing
MIT News
Adam Zewe
October 14, 2021
Massachusetts Institute of Technology (MIT) researchers have demonstrated that a specific neural network can learn the cause-and-effect structure of a navigation task it is taught. The researchers observed that a Neural Circuit Policy (NCP) system assembled by liquid neural network cells can autonomously control a self-driving vehicle using just 19 control neurons. They determined that when an NCP is being trained to complete a task, the network learns to interact with the environment and factor in interventions, or to recognize if an intervention is altering its output, and then it can relate cause and effect together. Tests put NCPs through various simulations in which autonomous drones performed navigation tasks. MIT's Ramin Hasani said, "Once the system learns what it is actually supposed to do, it can perform well in novel scenarios and environmental conditions it has never experienced."
|
Researchers Develop Model to Assess for Flood Hazards
Concordia University (Canada)
Patrick Lejtenyi
October 14, 2021
Canadian and U.K. scientists led by a team at Canada's Concordia University have developed a new technique for generating a watershed-scale flood risk model based on Light Detection and Ranging (LiDAR) data. LiDAR can yield a precise, dated, time-stamped water surface reading, which the researchers paired with river discharge data from the Ministry of Environment and Fight Against Climate Change to calculate riverbed depth. Awareness of a river’s carrying capacity can provide public safety officials enough time to warn local residents when models predict conditions that can lead to flooding. The initiative is part of Quebec's Project INFO-Crue, an effort to update flood maps in 50 watersheds (drainage basins) that were prioritized following recent floods.
|
ML Uncovers 'Genes of Importance' in Agriculture
National Science Foundation
October 13, 2021
New York University (NYU) researchers found machine learning (ML) can identify crop growth-promoting genes through a U.S. National Science Foundation (NSF)-funded study. NYU's Gloria Coruzzi said, "Focusing on genes whose expression patterns are evolutionarily conserved across species enhances our ability to learn and predict 'genes of importance' to growth performance for staple crops, as well as disease outcomes in animals." Experiments validated eight master transcription factors as genes of importance to nitrogen use efficiency. The researchers demonstrated that altered gene expression in Arabidopsis and in corn could boost plant production in low-nitrogen soils. NSF's Diane Okamuro said, "This is an excellent example of how NSF-supported scientists lead the way in using AI [artificial intelligence] and cutting-edge computational approaches to accelerate translation of basic plant genomic research and discoveries to the field."
|
Molecular Atlas of Small Cell Lung Cancer Could Explain Its Aggressiveness
Memorial Sloan Kettering Cancer Center
October 14, 2021
Memorial Sloan Kettering Cancer Center (MSK) investigators have generated a high-resolution map of small cell lung cancer (SCLC), the first to be produced by the National Cancer Institute-funded Human Tumor Atlas Network. MSK's Charles Rudin said the molecular atlas revealed "a rare population of stem-like cells within these tumors that is closely correlated with patient outcomes." MSK's Dana Pe'er and colleagues applied single-cell RNA sequencing (scRNAseq) to SCLC tumors to find the cells, whose presence in many otherwise distinct tumors might explain their highly metastatic behavior. Pe'er said, "What we really want to do is try to stop metastasis in its tracks. But to do that, we need to better understand these rare cell populations that seem to be driving it. That's the goal of this atlas."
|
Computational Approach Predicts Autism Diagnosis in Children
University of Chicago Medicine
Matt Reyer
October 13, 2021
University of Chicago researchers have developed a computational approach for predicting eventual diagnoses of autism spectrum disorder (ASD) in young children using only diagnostic codes from previous doctor's visits. The algorithm uses sequences of International Classification of Diseases diagnostic codes ICD9 and ICD10 from previous doctor's visits to generate an autism comorbid risk score (ACoR) for the patient, which estimates the risk that a child with a given timeline of diagnoses will be diagnosed as being on the Autism spectrum in the future. The researchers found this method cut in half the number of false-positive ASD diagnoses generated by traditional screening methods. University of Chicago's Peter J. Smith said, "Using the information already being gathered and being able to harness it for this kind of exploration and clinical use is exciting, and it really has the potential to be a game changer."
|
Data Mining the Past
University at Buffalo News Center
Kevin Manne
October 13, 2021
Researchers at the State University of New York at Buffalo and India's International Institute of Information Technology Bangalore have developed an algorithm to convert old newspapers into searchable data by identifying and ranking people's names in order of their importance. The researchers worked with the New York Public Library to analyze over 14,000 articles from The Sun published in November and December of 1894. The algorithm keys on attributes exclusively from text produced by optical character recognition (OCR) software, like name context, title before the name, article length, and how often the name is mentioned in an article. Because the OCR text was garbled, the researchers modeled the attributes statistically, and tested the algorithm on raw OCR-generated text and articles cleaned up manually by schoolchildren. They found it could rank names very precisely, even from the OCR text, when compared to the cleaned-up versions.
|
Search Engine Could Help Researchers Scour Internet for Privacy Documents
Penn State News
Matt Swayne
October 13, 2021
Pennsylvania State University (Penn State) scientists have designed a search engine that uses artificial intelligence to sift through millions of online documents, which could help privacy researchers find content related to online privacy. The PrivaSeer engine identifies relevant documents using natural language processing (NLP). Penn State's Mukund Srinath said the NLP approach focuses on certain words in text to distinguish between privacy and non-privacy policy documents. PrivaSeer has compiled approximately 1.4-million English language Website privacy policies. Said Penn State’s Shomir Wilson, “This can be a resource for researchers both in natural language processing and privacy, who are interested in this domain of text.”
|
A Computationally Quick Approach to Predict Molten Droplet Solidification on Solid Surface
Tokyo University of Science (Japan)
October 12, 2021
Japanese scientists have modeled the solidification of a molten droplet on a cooler flat surface using a technique that could potentially improve turbine efficiency. The model can be used to program simulations for predicting the deposition process in jet engines, in which small organic particles are sucked in, then melt inside the very hot engine and solidify onto cooler parts of the engine, like the turbines. The method diverges from previous models that assume a constant surface temperature by considering droplet behavior and heat transfer between the hotter droplet and the cooler surface. The mesh-less moving particle semi-implicit technique does not require multiple calculations on each grid, while the grid-based method computes temperature change within the substrate.
|
How to Move More Goods Through America's Clogged Infrastructure? Robot Trains
The Wall Street Journal
Christopher Mims
October 9, 2021
Autonomous trains increasingly are seen as a solution to U.S. truck driver shortages, as well as a way for companies to reduce carbon emissions. Florida A&M University's Maxim A. Dulebenets predicts that "trains are going to reach full autonomy faster than vehicles," especially since hundreds of passenger trains worldwide already operate autonomously as part of urban transportation systems. However, most autonomous trains are built on newer, dedicated tracks that are not shared with human-controlled trains and generally do not include hazards like highway crossings. Dulebenets said completely automating the U.S. rail network, in which multiple private rail companies share many lines, "could take decades." There also are concerns about safety and cybersecurity.
*May Require Paid Registration
|
Researchers Build $400 Self-Navigating Smart Cane
Stanford Institute for Human-Centered Artificial Intelligence
Andrew Myers
October 13, 2021
Stanford University researchers have unveiled a $400 self-navigating robotic cane that can help visually impaired users detect and avoid obstacles. Using tools similar to those used in autonomous vehicles, the cane incorporates LiDAR, global positioning system sensors, accelerometers, magnetometers, and gyroscopes with artificial intelligence-based wayfinding and robotic algorithms. On its tip is a motorized, omnidirectional wheel that maintains contact with the ground, and can steer users by tugging and nudging the user around obstacles. Said Stanford’s Mykel Kochenderfer, “We wanted to optimize this project for ease of replication and cost. Anyone can go and download all the code, bill of materials, and electronic schematics, all for free. Solder it up at home. Run our code. It’s pretty cool.”
|
|
